Depression of tryptophan may contribute to adverse effects of naproxen

Ghosh, Soumita; Lahens, Nicholas F.; Barekat, Kayla; Tang, Soon-Yew; Theken, Katherine N.; Ricciotti, Emanuela; Sengupta, Arjun; Joshi, Robin; Hu, Weiming; Bushman, Frederic D.; Weljie, Aalim; Grosser, Tilo; FitzGerald, Garret A.

doi:10.1038/s41467-026-69684-7

Download PDF

Article
Open access
Published: 14 February 2026

Depression of tryptophan may contribute to adverse effects of naproxen

Nature Communications volume 17, Article number: 2776 (2026) Cite this article

3495 Accesses
13 Altmetric
Metrics details

Subjects

Abstract

Non-steroidal anti-inflammatory drugs are popular choices for the mitigation of pain and inflammation; however, they are accompanied by adverse effects in the gastrointestinal and cardiovascular systems. Identifying biomarkers and mechanisms for early gastrointestinal or cardiovascular disease detection is desirable. Here we compare the effects of placebo, naproxen, a traditional NSAID, and celecoxib, a cyclooxygenase 2 inhibitor and find a decrease in tryptophan and kynurenine levels in the plasma of healthy volunteers who receive naproxen. We further validate this result in mice. Depression of tryptophan is independent of inhibition of either cyclooxygenase but rather, is due to the displacement of bound tryptophan by naproxen. Supplementation of tryptophan in naproxen-treated mice rescues fecal blood loss and inflammatory gene expression driven by IL-1β in the heart. Furthermore, tryptophan also rescues changes in genes that are reflective of inflammation and tissue damage in the gut.

Theaflavin mitigates acute gouty peritonitis and septic organ injury in mice by suppressing NLRP3 inflammasome assembly

Article 23 May 2023

Geniposide alleviates heart failure with preserved ejection fraction in mice by regulating cardiac oxidative stress via MMP2/SIRT1/GSK3β pathway

Article 26 July 2024

Exploring the anti-inflammatory mechanism of geniposide in rheumatoid arthritis via network pharmacology and experimental validation

Article Open access 10 July 2025

Introduction

Non-steroidal anti-inflammatory drugs (NSAIDs) are antipyretic¹, analgesic², and anti-inflammatory drugs; they are amongst the most consumed drugs worldwide. They act by inhibiting the cyclooxygenase (COX) enzymes that metabolize arachidonic acid (AA) to prostaglandins (PGs) and thromboxane (Tx)³. These prostanoids are mediators of inflammation and fever, playing a role in the pathophysiology of many diseases⁴. Although NSAIDs offer a non-addictive option in the management of pain, NSAIDs cause gastrointestinal and cardiovascular adverse effects^5,6,7.

Here, we compared the effects of two NSAIDs, celecoxib, selective for inhibition of COX-2 and naproxen, an inhibitor of both COX-1 and COX-2, which differ in their propensity for GI adverse effects⁸. We analyzed the metabolomic response to these drugs to seek discriminant biomarkers that might reflect and predict NSAID induced toxicity. Using studies in both humans and mice, we found that naproxen, but not celecoxib, depressed levels of tryptophan and its metabolites and that naproxen evoked fecal blood loss and cardiac inflammation in the mouse is mitigated by tryptophan supplementation.

Results

Naproxen reduces tryptophan and kynurenine in humans

A double-blind three-way crossover study compared placebo, celecoxib and naproxen, each taken for 7 days (Fig. 1A). The baseline characteristics of study participants are detailed elsewhere⁹. No one was on medication for hypertension. As expected, serum TxB₂ and ex vivo LPS evoked PGE₂ in plasma, reflective of inhibition of COX-1 and COX-2, respectively, were both decreased by naproxen, while only PGE₂ was reduced by celecoxib compared to placebo (Fig. 1B, C)^9,10.

**Fig. 1: Metabolomic response to NSAID administration to healthy volunteers.**

The LC/MS/MS metabolomics data, obtained at T0 and T4, i.e., 0 hr and 4 hr after final drug administration as detailed in Supplementary Data 1 were analyzed using Orthogonal Partial Least Squares-Discriminant analysis (OPLS-DA), using VIP (variable importance on projection) screening where appropriate. The OPLS-DA scores plot (CV ANOVA of OPLS-DA plots for the T0 and the T4 time points are 0.0002, 0.0002, respectively) segregated the naproxen group from the celecoxib and placebo-treated groups (Fig. 1D, E). We used the loading data from OPLS-DA analysis to identify metabolic pathways altered by naproxen treatment. The top list of the OPLS-DA plots is presented in Supplementary Data 2. The top metabolites corresponding to OPLS-DA between placebo and naproxen treatment were selected based on a Variable Importance in Projection (VIP) > 1.5 and were subjected to pathway analysis in Metaboanalyst. The pathway analysis shows that the false discovery rate (FDR) for tryptophan metabolism was 0.024, and the pathway impact was 0.24 (Fig. 1F), consistent with this being significantly altered by naproxen. Next, a 2-way repeated measures ANOVA, combined with Tukey’s HSD post hoc tests, was used to perform univariate analyses of the metabolites obtained in the global metabolomics platform. This revealed significant changes in the levels of tryptophan and kynurenine in the plasma of naproxen-treated volunteers compared to those in the placebo and celecoxib groups (Supplementary Fig. S1A, B). A targeted metabolomics experiment confirmed these findings and allowed us to obtain quantitative measures of the metabolites in the tryptophan pathway (Fig. 1G). The quantitative measurements of the metabolites demonstrate a decrease in the tryptophan levels at both T0 and T4 in the naproxen group (P < 0.0001, P < 0.0001; Fig. 1H). The median tryptophan level in plasma on naproxen, celecoxib, and placebo were 68.0 µmol/l, 85.0 µmol/l, and 91.1 µmol/l respectively at T0 while at T4 it was 57.9 µmol/l, 83.4 µmol/l, and 86.0 µmol/l. The major tryptophan metabolite, kynurenine, was also depressed by naproxen at T0 and T4 (P < 0.0001, P < 0.0001; Fig. 1I). The median level of kynurenine on naproxen, celecoxib, and placebo were 1.9 µmol/l, 3.1 µmol/l and 3.0 µmol/l at T0, while at T4 were 2.1 µmol/l, 2.9 µmol/l, and 3.2 µmol/l, respectively. Other tryptophan metabolites, like kynurenic acid (KA), xanthurenic acid (XA), and anthranilic acid (AA), did not exhibit significant differences between the groups (Supplementary Fig. S1C–E). The Receiver Operating Characteristics curve (ROC) was used to evaluate the specificity and sensitivity of the marker molecules. The area under the curve (AUC) for tryptophan in placebo vs. naproxen was 0.84 (P < 0.0001; Fig. 1J), celecoxib vs naproxen was 0.8 (P < 0.0001; Fig. 1K), while placebo vs. celecoxib was 0.5 (P = 0.24; Fig. 1L). The AUC for kynurenine was 0.79 for placebo vs naproxen (P < 0.0001; Fig. 1M); 0.78 for celecoxib vs naproxen (P = 0.0002; Fig. 1N), and 0.5 for placebo vs. celecoxib (P = 0.83; Fig. 1O). Some of the other features discriminated at T0 vs. T4 in naproxen treatment are nicotinamide and N-methyl-Nicotinamide along with sarcosine, arginine, proline, alanine, xanthine, and methyladenosine. Nicotinamide and N methyl nicotinamide also reflected alterations in tryptophan metabolism (Supplementary Fig. S1F,–H). Isobutyrate, 3-hydroxybutyrate, 2-hydroxybutyrate, 2-oxoisocaproate, and succinate differed between T0 and T4 in the celecoxib group (FDR ≤ 0.2; Supplementary Fig. S2). Apart from this, we also found alterations in the intermediates of the TCA cycle in urinary metabolic profiles from volunteers receiving both celecoxib and naproxen (Fig. S3A). Consistent with these data, plasma alanine and urinary malonate are also altered by naproxen, reflecting an alteration in the TCA cycle (Supplementary Fig. S3B, C). In addition, pathway analysis obtained from the correlated urinary metabolites in celecoxib and naproxen-treated volunteers shows an alteration in the glycine, serine and threonine metabolic pathway (Supplementary Fig. S3D).

Naproxen reduced tryptophan and kynurenine in mice

Eight-week-old mice were randomized into naproxen and chow control dietary groups (n = 8/sex/diet), as shown in Fig. 2A. After 3 weeks of dosing, naproxen had lowered tryptophan in the plasma compared to the controls in both sexes (P < 0.0001; Fig. 2B). In males, tryptophan levels were 121.6 µmol/l (IQR: 106.8-131.9) with naproxen and 142.8 µmol/l (IQR: 130.1-151.0) in controls. In females, the naproxen group had levels of 121.5 µmol/l (IQR: 92.77–136.2 µmol/l), while the control group had levels of 153.2 µmol/l (IQR: 146.4–165.4 µmol/l). Kynurenine was also reduced in the Naproxen group compared to controls in both sexes (P < 0.0001; Fig. 2C). In males, the kynurenine levels were 0.46 µmol/l (IQR: 0.42-0.58 µmol/l) with vs. 0.78 µmol/l (IQR: 0.58-1.31 µmol/l) in controls. For females, the naproxen group had levels of 1.27 µmol/l (IQR: 1.21-1.44 µmol/l), while the control group showed levels of 1.99 µmol/l (IQR: 1.38-2.71 µmol/l). The other metabolites in the tryptophan pathway, KA, XA, and AA, were not depressed by naproxen (Fig. 2D–F), consistent with our observations in humans. The AUC of the ROC was 0.91 for tryptophan, with a 95% confidence interval (CI) of (0.80-1.00), (P < 0.0001; Fig. 2G). The AUC for kynurenine was 0.86 with a 95% CI of (0.73-0.98), (P < 0.0005; Fig. 2H). Both tryptophan and kynurenine were also depressed by naproxen after only 5 days of treatment (Supplementary Fig. S4A–D). Weight and food intake were measured for three weeks to assess the potential effect of feeding on tryptophan levels. However, no significant differences were found between the groups (Supplementary Fig. S5A–E). To address the possibility that depression of tryptophan reflected platelet COX-1 inhibition by naproxen, we measured plasma tryptophan and kynurenine in platelet-specific COX-1 knockout (KO) mice (Pf4-Δ Cre ^+/-/COX-1^fx/fx)¹¹. Despite the suppression of the major COX-1 product, Tx, in the KOs, tryptophan was unaltered (Supplementary Fig. S6A–C). Similarly, we assessed the impact of naproxen on tryptophan in mice following postnatal depletion of COX-1 and COX-2 globally (Cox DKO) and controls, and there was no impact of genotype on drug effect (Fig. 2I, J). There was no significant correlation between urinary enzymatic metabolites of prostaglandins (indices of in vivo systemic biosynthesis of prostanoids) with either tryptophan or kynurenine in naproxen-treated humans (Fig. 2K). In addition, consistent with these findings, we also did not observe any significant changes in plasma tryptophan or kynurenine levels in global COX-1 knockout mice compared to controls, nor in global COX-2 knockout mice compared to controls (Fig. 2L, M). This further supports the conclusion that COX suppression and tryptophan depletion by naproxen are unrelated. We sought to address the possibility that tryptophan was differentially metabolized via the serotonin or melatonin pathways during naproxen administration. While melatonin was not detected in mouse plasma, consistent with previous findings¹², the serotonin level was unaltered by naproxen (Supplementary Fig. S6D). We also found no evidence that naproxen altered kynurenine aminotransferase (KAT) activity as a potential cause for the change in plasma tryptophan concentration in both mice and humans (Supplementary Fig. S7A–D). R-2-phenylpropionic acid (PPA), a structural analog of naproxen, devoid of COX inhibitory activity¹³, had no impact on tryptophan (Fig. 2N–P).

**Fig. 2: Plasma metabolic profiles of mice are altered by naproxen.**

Naproxen alters tryptophan binding

Tryptophan was 73.8% bound to albumin in the placebo group (IQR: 69.90-78.66), while in the celecoxib group, it was 70.9% (IQR:67.64 -75.20) and in the naproxen group 57% bound (IQR: 48.47- 63.15). The % bound tryptophan was significantly decreased by naproxen compared to placebo and celecoxib (P < 0.01 and P < 0.05, respectively; Fig. 3A). Free tryptophan was reduced by naproxen in the plasma of volunteers (P < 0.05; Fig. 3B). Further, we found a negative correlation between tryptophan and naproxen levels in plasma (P < 0.05; Fig. 3C). We also found that bound tryptophan was decreased dependently by naproxen in mouse plasma, investigated in one set of pooled mouse plasma samples. Importantly, both naproxen and tryptophan bind to the Sudlow 2 site of albumin with an association constant of 1.2–1.8 × 10⁶M⁻¹ and 4.88 × 10⁴M⁻¹, which explains the lower binding of tryptophan to albumin in the presence of naproxen. Similar to tryptophan, kynurenine binds to albumin with an association constant of 2.8 × 10⁵(M⁻¹)¹⁴, which is lower than the association of naproxen, explaining lower binding to albumin in the presence of naproxen. Our data is consistent with the displacement of tryptophan and kynurenine by naproxen.

**Fig. 3: Naproxen displaces tryptophan from albumin binding.**

Potential role of naproxen-induced changes in the microbiome in tryptophan metabolism

Naproxen treatment was associated with alterations in the gut microbiome. Bacteria of the genus Coprococcus and family Ruminococcaceae showed a trend toward an increase in abundance following naproxen treatment (P = 0.018, P = 0.022, FDR = 0.25, 0.25; Fig. 4A, B). In humans, both taxa have been suggested to influence tryptophan metabolism^15,16. This assessment represents an analysis conducted at the operational taxonomic unit (OTU) level. We report results for these, and other bacterial lineages compared over multiple taxonomic levels for tests of the drugs against placebo in Supplementary Table 1. For this analysis, we combined individual OTU relative abundances into taxonomic groups before conducting differential abundance testing at the taxon level. This analysis summarizes overall changes in the microbial community caused by both Naproxen and Celecoxib. The term “estimate” in Supplementary table 1 refers to estimated regression coefficients in the model.

**Fig. 4: Naproxen alters the microbiome and potential microbiome-derived metabolites.**

A meta-analysis of 16S sequencing datasets shows that Coprococcuses and the family Ruminococcaceae are commonly present in fecal samples from both humans and mice¹⁷. These lineages may contribute to the increase of urinary indole 3 acetic acid and indole lactic acid in humans (Fig. 4C, D) and in mice (Fig. 4E, F) associated with naproxen treatment, as previous reports have linked Coprococcuses and Ruminococcaceae to indole metabolites^15,18,19,20. Furthermore, although many microbial genera are not shared between mouse and human²¹ it has previously been estimated using gene-level functional analyses that up to 95% of gut microbial functions are shared between mouse and human²², which may influence metabolite concentrations reported here. Here, a modest sample size can reduce the statistical power to identify differential microbial characteristics, particularly for taxa with low abundance, which may also contribute to greater variability in parameter estimates. Furthermore, this analysis employed a threshold of P < 0.1 and a relatively liberal false discovery rate (FDR < 0.25) to balance sensitivity and specificity in its exploratory framework; so, this approach may increase the likelihood of false positive findings. It will be thus valuable to follow up with further analysis of these associations in validation cohorts.

Functional consequences of tryptophan depletion by naproxen

Depression of plasma tryptophan by NSAIDs was negatively correlated with a rise in Mean Arterial Pressure (MAP) in healthy normotensive volunteers (https://clinicaltrials.gov/study/NCT02502006)⁹, suggesting a possible link between the change in tryptophan and cardiovascular function (Supplementary Fig. S8A). L-tryptophan can cause a dose-related decrease in blood pressure, which could be due to the enhanced uptake of tryptophan in the brain and its conversion to serotonin²³. To investigate the role of tryptophan in the response to naproxen, we supplemented mice on the naproxen diet with tryptophan (50 mg/kg i.p. daily for 7 days) (Fig. 5A), sufficient to restore plasma levels from their drug induced depression (Fig. 5B). While body weight was unaltered in these experiments, many DEGs caused by naproxen were rescued by tryptophan supplementation (Fig. 5C). Of the 212 (out of 17093) genes that were rescued (Supplementary data 3), 42% are involved in biological regulations, 45% in cellular processes, 24% response to stimulus and 18% in metabolic processes (PANTHER.db). Further, we identified 275 genes that differ between control mice and those given tryptophan alone (Supplementary Data 4 A). Of these, 9 genes exhibited changes in the same direction to those rescued by tryptophan on naproxen, while 3 changed in the opposite direction (Supplementary Data 4B). Several inflammatory pathways are restrained by tryptophan supplementation during naproxen treatment (Fig. 5D). We find negative z activation scores of IL-13 activation pathway, cytokine storm signaling, and cardiac hypertrophy signaling (− 0.447 − 2.53 and – 2.71, respectively), signifying inhibition of these inflammatory pathways in the heart by tryptophan supplementation. Further, causal network and upstream regulator analyses demonstrate that IL1β has a z activation score of − 3–6 (BH corrected p-value = 0.000093), suggesting that IL1β might also be inhibited by tryptophan supplementation (Fig. 5E); this was confirmed by RT PCR (Fig. 5F). The nucleotide-binding domain, leucine-rich containing family, and pyrin domain containing 3 (NLRP3) was also decreased by tryptophan supplementation (Fig. 5G). Notably, NLRP3 activates Caspase-1, which in turn releases IL1β. Causal network analysis also revealed inflammasome inhibition by tryptophan supplementation (activation score of − 4.12 (p = 5.82e-8)). The genes rescued by tryptophan supplementation of naproxen treated mice which are also targets of IL1β are listed in Fig. 5H. These genes participate in the inflammation-mediated chemokine and cytokine signaling pathway as well as the interleukin signaling pathway, as identified in PANTHER.db (Fig. 5H). In addition, a subset of IL1β mediated rescue genes is involved in various inflammatory pathways in the heart. The pathways, including cytokine storm signaling, IL-13 signaling, cardiac hypertrophy, cardiac dysfunction, and myocardial infarction, are illustrated in Fig. 5I. Furthermore, The IL-1β protein was also elevated in response to Naproxen and was rescued by tryptophan (Fig. 5J). In addition, 1L-10 (Supplementary Fig. S9A) and the elevation of IL17 receptor and IL4 receptor expression by naproxen (Supplementary Fig. S9B, C), related to adverse cardiovascular health^24,25 were rescued by tryptophan supplementation. Tryptophan may also have mitigated energy demand in the heart by rescuing FNIP2, PGC1α, ITGB3 and TGM2^26,27,28,29. In addition, naproxen raises gut inflammation, shown by increased stool hemoglobin and higher proline levels. Tryptophan concentration is inversely related to hemoglobin content (Supplementary Fig. S10A–D). Furthermore, tryptophan supplementation prevented the increase in fecal hemoglobin caused by naproxen, but not platelet function (Supplementary Fig. S10E, F). In addition to this, we conducted RNA sequencing of jejunal wall samples from five groups: Control, naproxen-treated mice, celecoxib-treated mice, tryptophan-only treated mice, and naproxen + tryptophan-treated mice. Supplementary Fig. S11A lists the number of genes impacted on different groups. Supplementary Fig. S11B, C displays the gene changes induced by naproxen that were mitigated by tryptophan in the gut. The supplementation of tryptophan rescued naproxen induced changes in adiponectin, CD19, Cd79b, COL7A1, Fndc1, Pi15, SELL, and Vwa3b. Lower adiponectin levels during naproxen use may affect gut inflammation, as previous studies have linked high adiponectin levels with anti-inflammatory phenotypes in macrophages³⁰. Adiponectin directly inhibits pro-inflammatory pathways, including those regulated by Toll-like receptors (TLRs), by suppressing nuclear factor-kappa B (NF-κB) activity in various cell types³¹. CD19 is associated with B cell development and activation, which is associated with wound healing. CD19 overexpression has been implicated in enhanced wound healing and cytokine expression³². Similarly, Cd79b is important for B cell function and regulates wound healing. COL7A1 encodes the α1 chain of collagen type VII, which interacts with the extracellular matrix to form anchoring fibrils. These fibrils play a crucial role in the formation and adhesion of epithelial basement membranes and the integrity of the intestine wall³³. A decrease in COL7A1 by naproxen is consistent with damage to the intestine which was rescued by tryptophan. Rescue of SELL may also reflect tissue repair, and Fndc1 is reported to play an anti-inflammatory role in previous studies³⁴.

**Fig. 5: Tryptophan Supplementation in mice attenuates naproxen induced cardiac inflammation.**

Discussion

NSAIDs are amongst the most consumed drugs because of their non-addictive efficacy in relieving pain and inflammation. Despite this, they can cause gastrointestinal and cardiovascular adverse effects which can, on occasion, be life threatening³⁵. While metabolomic analyses of the response to NSAIDs have been reported³⁶, here we study the comparative effects of a traditional NSAID, naproxen, and one specifically designed to target COX-2, celecoxib³⁷ in humans. Gastroduodenal adverse effects of NSAIDs have been attributed largely to the inhibition of COX-1³⁸, while their risk of heart attack, hypertension and stroke reflects particularly suppression of COX-2³⁹.

Our most striking finding was that naproxen, but not celecoxib, depressed tryptophan and its major metabolite, kynurenine. We replicated this drug effect in mice. A surprise was that this seems unrelated to the suppression of PGs and Tx by naproxen. Using a variety of mutant mice^11,40,41, we found that depletion of COX-1 or COX-2, alone or together had no impact on the depression of tryptophan by naproxen, uncoupling this effect from drug induced suppression of PGs and Tx.

The next question was how naproxen might depress tryptophan. Although some effects on other metabolites – elements of the TCA cycle, for example – were observed, the impact on the tryptophan pathway was the most striking. Measurement of free and bound tryptophan provided us with evidence consistent with naproxen, but not celecoxib, displacing tryptophan from its protein binding site, permitting its degradation and elimination under steady state dosing conditions. Indeed, this may interact with the perturbation of the microbiome by naproxen to accelerate tryptophan disposition. Thus, we found that the drug increased the abundance of two fecal strains, Ruminococcaceae and Coprococcuses, in humans which have been shown to influence tryptophan metabolism^16,20,42,43 by its conversion to indole derivatives. The microbiome was not altered by tryptophan supplementation in naproxen treated mice (Supplementary Fig. S12A–C). This is also reflected in the indole metabolites, ILA and IAA. (Supplementary Fig. S12D, E). The data is consistent with drug induced microbial perturbation preceding the microbial breakdown of tryptophan, rather than restructuring the microbiome resulting from the indole metabolites.

The next question was whether this depressive effect of naproxen on tryptophan might have contributed to its adverse effect profile. To address this possibility, we examined the impact of tryptophan on gene expression in the mouse heart. Here, using a variety of approaches, we found that naproxen induced expression of inflammatory genes, including those driven by IL-1β, culminating in activation of the inflammasome. Supplementation of this naproxen treated mice with tryptophan, sufficient to restore its plasma levels, rescued much of the inflammatory signature of drug induced gene expression.

Correspondingly, we performed a preliminary study to determine if this mitigating effect of tryptophan supplementation might influence other aspects of the adverse effect profile of naproxen. Naproxen increases fecal occult blood, reflective of the gastroenteropathy and decreases Tx-M consistent with inhibition of platelet function in mice (Supplementary Fig. S10F). Here again, tryptophan supplementation appeared to attenuate the marker of gastroenteropathy but not of platelet inhibition. Naproxen induced changes in the expression of genes reflective of inflammation and tissue damage in the gut, many of which were mitigated by tryptophan supplementation.

In summary, we report that naproxen, in humans and mice, depresses tryptophan and its major metabolite by a mechanism unrelated to COX inhibition and reflecting displacement from albumin and accelerated disposition by the gut microbiome. Naproxen induces inflammatory gene expression in the mouse heart and gut that may contribute to the heart failure and gastroenteropathy that complicate NSAID administration. These gene signatures and the fecal blood loss evoked by naproxen are both mitigated by tryptophan supplementation.

Thus, depletion of tryptophan may contribute to the cardiovascular and gastrointestinal adverse effects of naproxen and perhaps other NSAIDs in humans. Our studies in mice raise the possibility that tryptophan supplementation may attenuate adverse effects and widen the therapeutic index of this commonly consumed NSAID.

Methods

The research for the human clinical trial and mouse experiments complies with ethical regulations. The study protocol for the clinical trial was approved by the University of Pennsylvania Institutional Review Board (IRB#820715). The full clinical trial protocol is provided as Supplementary note 1. Experimental protocols for the mice experiments were reviewed and approved by the Institute for Animal Care and Use Committee at the University of Pennsylvania (Protocol #804445).

Study design of clinical trial

Comprehensive details regarding clinical trial registration, study protocol, data collection, and outcomes are provided by Theken et al.⁹ The participants provided written informed consent to participate in the study. The age and the gender of the volunteers are mentioned by Theken et al.⁹. The participants received $525 compensation for completing the study. This includes each week of study drug treatment plus each whole-day visit to the Clinical and Translational Research Center (CTRC) and each post-treatment visit.

The metabolomic response to naproxen and celecoxib was evaluated as a secondary analysis of a double-blind, randomized, crossover study in healthy volunteers (N = 16; 9 men, 7 women). The trial design is shown in (Fig. 1A). Healthy men and women aged 18 or older, as determined by medical history, physical exam, vital signs, and lab tests, were enrolled. Subjects received celecoxib (100 mg bid), naproxen (250 mg bid), or placebo (bid) for 7 days, then returned to the clinical research unit for a 12 h visit for pharmacokinetic-pharmacodynamic sampling. There was a washout period of at least 14 days between each study phase. The participants were asked to abstain from analgesics, high-dose vitamins and nutritional supplements until study completion. All the experiments were performed in a blinded manner by the analyst. The study protocol was approved by the University of Pennsylvania Institutional Review Board (IRB#820715; ClinicalTrials.gov: NCT02502006). The original study was designed to investigate the clinical, genetic, and environmental factors that contribute to variability in pharmacologic and blood pressure response to NSAIDs at steady state, with a plan to evaluate the effects of a low dose (naproxen 250 mg bid vs. celecoxib 100 mg bid vs. placebo) and a high dose regimen (naproxen 500 mg bid vs. celecoxib 200 mg bid vs. placebo). The planned enrollment was N = 288 to ensure a diverse study cohort with sufficient statistical power to evaluate the effects of age, race, and dose on NSAID response. The study was terminated after the first 16 subjects had completed the study (all in the low dose arm) due to insufficient funding to continue enrollment⁹. Given the limited sample size, analyses based on sex or gender were not pursued; as stratifying or powering sub-analyses by these variables would result in insufficient subgroup sizes, thereby reducing the statistical power necessary to detect differences, thus increasing the likelihood of Type II errors. The primary endpoint was COX-1 and COX-2 activity as assessed ex vivo using whole blood assays and in vivo by measuring urinary excretion of prostaglandin metabolites after 7 days of treatment⁹. Secondary endpoints included pharmacokinetic profiles of celecoxib and naproxen, ambulatory blood pressure, renal function, transcriptomics in peripheral blood mononuclear cells and whole blood, plasma proteomics, gut microbiome analysis, urinary and plasma metabolomics at peak and trough drug concentrations, DNA sequencing, and nutritional assessments after 7 days of treatment. The comparison of COX-1 and COX-2 activity, pharmacokinetics, and blood pressure response to naproxen and celecoxib are reported in Theken, et al.⁹. The urinary and plasma metabolomics and microbiome analysis are reported in this manuscript.

Metabolomics analysis in plasma

Plasma samples taken at T0 (T = 0 h after 7 days of dosing) and T4 (T = 4 h after 7 days of dosing) for (N = 16; 9 men, 7 women) were subjected to metabolomics analysis. The time points represent the crest and the trough of the pharmacological activity of the drugs. Metabolomics data was collected in compliance with previously reported community standard⁴⁴. Briefly, samples were stored in -80 °C and thawed on ice before extraction. The total number of biological replicates for each treatment group at the T0 time point is as follows: Placebo, 16; Celecoxib, 16; Naproxen, 15. For the T4 time point, the number of biological replicates is: Placebo, 16; Celecoxib, 16; Naproxen, 16. The controls group consisted of subjects that received placebo treatment. The plasma samples from these subjects were subjected to metabolite extraction. Polar metabolites were extracted from 50 µL of plasma using a modified Bligh-Dyer protocol⁴⁵. Briefly, 2:2:1 cold methanol/chloroform/water was added to 50 µL of plasma. The samples are vortexed and centrifuged 18,787 × g for 7 min at 4 °C. The upper fraction was carefully collected and dried in a vacuum concentrator. The dried sample was reconstituted in 1:1 cold acetonitrile/water. 200 μl of each sample was added into vials (Phenomenex, Torrence, CA) and used for UPLC-MS/MS data acquisition. The order of sample preparation and data collection is completely randomized. The samples were injected into an XBridge BEH Amide column (2.5 mm × 100 mm x 2.1 mm) installed on a Acquity H-class UPLC system (Waters Corp. Milford, MA). Mobile phase A consisted of (95:5 water/acetonitrile, pH adjusted to ~ 9 using 20 mM ammonium acetate), and mobile phase B consisted of Acetonitrile. The gradient used in the experiment is (0 min, 85%B; 5 min 30%B; 15 min 30%B, 16 min 2%B, 22 min 2%B; 23 min 85%B; 24 min 85%B; 25 min 85%B, 30 min 85%B). Mass spectrometry was performed using a Waters Xevo TQ-S micro spectrometer (Waters Corp., Milford, MA.), using a desolvation and temperature 450 ^OC). Data was acquired in multiple reaction monitoring (MRM) mode with random injections to remove run order bias. The MRMs were curated from METLIN and HMDB databases. The run started with a blank injection of reconstituting solvent, followed by diluted (1:10) pooled QC samples were injected prior to starting the runs to condition the column with the sample matrix. quality control (QC) samples. Pooled QC samples were injected before, during and after the run to correct any instrumental drift. Data was integrated in Targetlynx 4.1. (Waters Corp., Milford, MA.). No carryover from prior runs in blank samples confirmed the suitability of beginning the new analytical batch. Every metabolite feature from QC samples was fitted to a LOESS function. Features appearing in < 50% of QC samples and/or relative standard deviation > 40% were excluded from further analysis. Data was exported for further analysis. In addition, probabilistic quotient normalization was applied to the data. Multivariate statistical analysis was performed in Simca-P 18.0 (Sartorius Stedim, Germany). Data was first checked for quality and/or outlier samples using unsupervised principal component analysis followed by a supervised orthogonal partial least square discriminant analysis (OPLS-DA). The metabolites of interest were validated using a targeted MRM-based assay with labeled internal standards, as described below in the section ‘Targeted Metabolomics of the Tryptophan Pathway’.

Measurements of prostanoids in ex-vivo assays and prostaglandin metabolites in humans

The ex vivo measurements of COX-1 and COX-2 activity was performed as described in the literature previously⁹. Briefly, COX-1 activity ex vivo was assessed by measuring serum thromboxane B₂. Whole blood was collected in vacuum tubes containing a clot activator, incubated at 37 °C for one hour, and subsequently centrifuged to separate the components serum. COX-2 activity ex vivo was assessed by measuring plasma PGE₂ after lipopolysaccharide (LPS) stimulation in whole blood. The biological replicates included 16 for celecoxib, 16 for naproxen, and 15 for placebo. LC/MS/MS was used to quantify prostanoids as described⁴⁶. Briefly, 5 ng of and d⁴ TXB₂ were added to 50 µl of serum, respectively. To measure TxB₂, 200 μl of water and 125 μl of methoxamine (MO) HCl solution were added to the mixture. The sample was kept at room temperature for 30 minutes before 575 μl of milliQ water was added. This was subjected to solid-phase extraction (Strata-X, 33 µm Polymeric Reversed Phase, Phenomenex) dried, reconstituted in 15 μl Acetonitrile and 165 μl water and measured by LC/MS/MS^47,48 using a Waters AQUITY UPLC system (Waters Corp., Milford, MA.), Separation of TxB₂MO and d⁴TxB₂MO was achieved using a Waters ACQUITY UPLC system equipped with an ultra-performance liquid chromatography (UPLC) column (2.1 × 150 mm, 1.7 μm particles size; Waters ACQUITY UPLC BEH C18) and a mobile phase consisting of water containing 0.5% ammonium acetate at pH 5.7 (mobile phase A) and acetonitrile–methanol mixture at a ratio of 95:5 (mobile phase B). The flow rate was 350ul/min. The gradient used was (0 min 20%B, 10 min-27% B; 10.5 min 100% B, 12 mins 100%B) with a total time of 15 mins. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.). The desolvation temperature used was 500 °C. The MRM ion transitions and optimized parameters (collision energy, cone voltage) from standards were used for the characterization of TxB₂MO and d⁴TxB₂MO. These were 398.2/169.15 (17 V, 30 V) and 402.20/173.15 (17 V, 30 V). Data was integrated into Targetlynx 4.1 software (Waters Corp., Milford, MA). There were 16 biological replicates each for celecoxib, naproxen, and placebo. The placebo group served as the control. To measure PGE₂, 5 ng each of d⁴ PGE₂ was added to 100 µl of plasma. This was followed by the addition of 900 μl of water and 20 μl of formic acid. This was subjected to solid-phase extraction (Strata-X, 33 µm Polymeric Reversed Phase, Phenomenex) dried, reconstituted in 15 μl Acetonitrile and 165 μl water and measured by LC/MS/MS Separation of PGE₂ and d⁴ PGE₂ was achieved using Waters ACQUITY UPLC system equipped with an ultra-performance liquid chromatography (UPLC) column (2.1 × 150 mm, 1.7 μm particles size; Waters ACQUITY UPLC BEH C18) and a mobile phase consisting of water containing 0.5% ammonium acetate at pH 5.7 (mobile phase A) and acetonitrile–methanol mixture at a ratio of 95:5 (mobile phase B). The flow rate used was 350ul/min. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.). The MRM ion transitions and optimized parameters (collision energy, cone voltage) obtained from standards were applied to characterize PGE2 and d4 PGE2: 351.20/271.20 (15 V, 25 V) and 355.20/275.20 (15 V, 25 V), respectively. The desolvation temperature was set to 500 °C. Data acquisition was performed using the Waters TQ-S AQUITY UPLC system (Waters Corp., Milford, MA). Data were integrated into Targetlynx 4.1 software (Waters Corp., Milford, MA.). Measurement of urinary metabolites of prostanoids was achieved by LC/MS/MS measurements in as described earlier⁴⁶ in 16 biological replicates. Briefly, d⁶-PGEM (25 ng), d⁶-PGDM (25 ng), d³-PGIM (5 ng), d⁴-TxM (5 ng) were added to 500 µl of human urine. This was followed by the addition of methoxamine (MO) HCl solution, and the mixture was allowed to equilibrate for 30 min. The mixture was further brought to a total volume of 1 ml. This was followed by solid phase extraction (Strata-X 33 μm polymeric reversed phase cartridges, Phenomenex, 8B-S100-TAK). The elute was dried, reconstituted in 9 μl of acetonitrile and 171 μl of millQ water and measured by LC/MS/MS^47,48. In this method, separation of the metabolites were achieved using a UPLC system equipped with an ultra-performance liquid chromatography (UPLC) column (2.1 × 150 mm, 1.7 μm particles size; Waters ACQUITY UPLC CSH C18) and a mobile phase consisting of water containing 0.5% ammonium acetate at pH 5.7 (mobile phase A) and acetonitrile–methanol mixture at a ratio of 95:5 (mobile phase B). The flow rate used was 350 μl/min. The gradient used in this method was follows: 0 min 0%B, 1 min 0%B, 17 min 10%B, 17.5 min 10.5%B, 32 min 11.5%B, 35 min 20%B, 40 min 34.4%B, 40.5 min 100%B, 42 min 100%B, 42.5 min 100%B. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.).The MRM ion transitions and parameters (collision energy, cone voltage) used for the characterization of TxM (11deHTxB₂) and d⁴TxM (d⁴11deHTxB₂) were 367.2/305.2 (16 V, 30 V) and 371.2/309.2 (16 V, 30 V) respectively; PGIM (d6KMO) and d³PGIM (d³d6KMO) are 370.2/232.2 (17 V, 40 V) and 373.2/235.2 (17 V, 40 V) respectively; PGDM/PGEM MO and d⁶PGDM/PGEM MO 385.25/336.2 (16 V, 40 V) and 391.25/342.2 (16 V, 40 V) respectively. The desolvation temperature used was 500 °C. Data were integrated into Targetlynx 4.1 software (Waters Corp., Milford, MA). The metabolites were normalized with respect to urinary creatinine, also quantified by LC/MS/MS. Briefly, 1 ml of stable isotope-labeled internal standard (10 μg/mL d³-creatinine in 3% H₂O/acetonitrile) was added to 10 μL of each urine sample, and the 2 μl from the mixture was diluted with 200 μL acetonitrile. Separation was performed using a UPLC column measuring 2.1 × 50 mm, 2.5 μm particles (Waters XBridge BEH HILIC). The mobile phase comprised 100% acetonitrile (mobile phase A) and a 5 mM ammonium formate aqueous solution at a pH of 3.98 (mobile phase B), with a flow rate of 350 μL/min. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.), The MRM ion transitions, and the parameters (collision energy, cone voltage) used for the characterization of Creatinine were 114/86 (10 V, 30 V) and d³ Creatinine 117/89 (10 V, 30 V). The data collected were analyzed for correlations with plasma tryptophan and kynurenine.

Measurements of urinary prostaglandin metabolites in mice

Urinary prostaglandin metabolites were analyzed using liquid chromatography/mass spectrometry according to the described method⁴⁶. Briefly, d⁶-PGEM (25 ng), d⁶-PGDM (25 ng), d³-PGIM (5 ng), ¹⁸O-TxM (5 ng) were added to 100 μl of mouse urine. To this was added 75 μl of MO/HCl. The sample was kept at room temperature for 30 min. The mixture was further brought to a total volume of 1 ml. This was followed by solid phase extraction (Strata-X 33 μm polymeric reversed phase cartridges, Phenomenex, 8B-S100-TAK). This was subjected to solid-phase extraction (Strata-X, 33 µm Polymeric Reversed Phase, Phenomenex), dried, reconstituted in 9 μl of acetonitrile and 171 μl of milliQ water, and measured by LC/MS/MS. In this method, separation of the metabolites were achieved using a UPLC system equipped with an ultra-performance liquid chromatography (UPLC) column (2.1 × 150 mm, 1.7 μm particle size; Waters ACQUITY UPLC CSH C18) and a mobile phase consisting of water containing 0.5% ammonium acetate at pH 5.7 (mobile phase A) and acetonitrile–methanol mixture at a ratio of 95:5 (mobile phase B). The flow rate used was 350ul/min. The gradient used in this method was follows: 0 min 0%B, 1 min 0%B, 17 min 10.4%B, 17.5 min 11.8%B, 20.5 min 20%B, 28.5 min 43%B, 28.6 min 90%B, 30.10 min 90%B, 30.2 min 0%B. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.).The MRM ion transitions and parameters (collision energy, cone voltage) used for the characterization of TxM (dTxMO) and ¹⁸O₂TxM (18O₂dTxMO) were 370.2/155.15 (12 V, 30 V) and 374.2/155.15 (12 V, 40 V) respectively; PGIM (d6KMO) and d³PGIM (d³d6KMO) are 370.2/232.2 (17 V, 40 V) and 373.2/235.2 (17 V, 40 V) respectively; PGDM/PGEM MO and d⁶PGDM/PGEM MO 385.25/336.2 (16 V, 40 V) and 391.25/342.2 (16 V, 40 V) respectively. The desolvation temperature used was 500 °C. Data were integrated into Targetlynx 4.1 software (Waters Corp., Milford, MA).

The data was acquired in a Waters ACQUITY UPLC system (Waters Corp., Milford, MA). Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.). The data was normalized by urinary creatinine concentration.

Targeted metabolomics of the tryptophan pathway

For the targeted assay of the metabolites, plasma samples were processed as described⁴⁹. In this experiment, 10 µl of an internal standard, consisting of 260 ng of d⁵-Tryptophan was added to 40 μl of plasma samples. To this solution 10 µl of 0.1% formic acid, was added followed by 1 ml of milli-Q water. The mixture was vortexed and centrifuged at 10,000 × g for 5 mins. Solid Phase extraction in the reversed-phase cartridge (Strata-X, 33 µm Polymeric Reversed Phase, Phenomenex, Torrance, CA). was carried on these samples. The cartridges were conditioned with 1 mL methanol followed by 1 ml milli-Q water. Samples were loaded to the cartridges and washed with 1 ml milli-Q water. The cartridges were dried in vacuum for 15 mins. The samples are eluted in 1 ml of 95:5 methanol: water containing 0.1% formic acid. The samples are dried under nitrogen and reconstituted in 40 µL acetonitrile and 360 µL water. Separation of the metabolites was performed using an ultra-performance liquid chromatography (UPLC) column, 2.1 × 150 mm with 1.7 μm particles (Waters ACQUITY UPLC BEH C18), following this gradient: Mobile phase A consisted of 0.1% formic acid, 95% water, 5%B, and mobile phase B was acetonitrile: methanol in 95:5 containing 0.1% formic acid. Linear gradient was run as follows: 0 min 98% A; 5 min 95% A; 5.5 min 90% A; 20 min 40% B; 20.1 min 90% A; 23 min 98% A at a flow rate of 0.3 mL min⁻¹ with a total run time of 25 min. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.). The MRM ion transitions, and the parameters (collision energy, cone voltage) used for the characterization of tryptophan 204.79/145.88 (18 V, 18 V), d⁵ Tryptophan 210.03/149.9 (14 V, 14 V) and Kynurenine 209/93.7 (12 V, 2 V). Desolvation temperature was 600 °C. Calibration curves were used for the quantitation of the metabolites.

Measurement of plasma drug concentrations

For the analysis of drug concentration, 50 µl of an internal standard containing 500 ng of d³ Naproxen was added to plasma samples before solid-phase extraction (Strata-X, 33 µm Polymeric Reversed Phase, Phenomenex). The samples were dried, reconstituted in 9 μl of acetonitrile and 171 μl of milliQ water, and measured by LC/MS/MS. Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.), equipped with a UPLC system (Waters Corp., Milford, MA), was used to measure drug concentrations. Naproxen was profiled using a 2.1 × 150 mm column, with 1.7 μm particles (Waters ACQUITY UPLC CSH C18). The mobile phase consisted of 0.5% ammonium acetate at pH 5.7 (mobile phase A) and acetonitrile (mobile phase B). The following gradient was then used with a flow rate of 350 µl/min; The gradient used in the method was 0 min 0%B; 15 min 67.5%B; 15.2 min 90%B; 17 min 90%B; 17.5 min 0%B. The MRM ion transitions, and the parameters (collision energy, cone voltage) used for the characterization of Naproxen were 231/185.1(8 V, 40 V) and d³Naproxen 234/188 (8 V, 40 V) respectively. The desolvation temperature was 550 °C. Data was again integrated into Targetlynx 4.1 software.

Measurement of bound and free tryptophan concentrations in blood

Free and bound tryptophan were separated by equilibrium dialysis (RED plate, Thermo Fisher). Briefly, 100 µl of plasma was inserted into the inner vial and 300 µl of the buffer (provided in the kit) was added in the outer vial. The plate was shaken for 4 hrs. The samples were removed from the chambers and stored in separate vials at − 80 °C. Before the LC/MS/MS analysis, the samples were thawed in ice, and d⁵-tryptophan internal standard was added. To understand the direct role of naproxen on tryptophan displacement, pooled mouse plasma was aliquoted to 10 equal parts of 100 µL. Duplicate samples were treated with (0 µM, 2.5 µM,10 µM, 25 µM, 100 µM) of Naproxen in vitro and were subjected to Equilibrium Dialysis in RED plate as mentioned before. The samples were further processed for the tryptophan measurement as mentioned in the section above.

Analysis of Indole metabolites in urine

Urine was collected from mice and humans for analysis of Indole 3 Acetic acid, Indole lactic acid, and Indole sulfate by LC/MS/MS. 10 µl of the urine sample was added to 100 µl of milli-Q water containing d⁵-Tryptophan as internal standard. The mixture was centrifuged at 10,000 × g at 4 °C for 10 mins. The supernatant was loaded to an autosampler, and separation of the metabolites was achieved by using a Waters ACQUITY UPLC system (Waters Corp., Milford, MA) as using a mobile phase A of 95:5 water and B, with 0.1% formic acid, and mobile phase B consisting of 95:5 acetonitrile: water with 0.1% formic acid. The gradient used in the method is 0 min 0%B; 5 min 0%B; 5.5 min 10%B; 20 min 60%B; 20.1 min 0%B, 23 min 0%B. Mass spectrometry was performed using a Waters Xevo TQ-S spectrometer (Waters Corp., Milford, MA.). The desolvation temperature was 550 °C. The MRM ion transitions, and the parameters (collision energy, cone voltage) used for the characterization of Indole 3 lactic acid were 205.84/117.77 (20 V, 2 V), while that of Indole Acetic acid ware 175.96/129.81 (22 V, 40 V) respectively. Data was again integrated into Targetlynx 4.1 software. The metabolites were normalized with respect to urinary creatinine, also quantified by LC/MS/MS.

Global metabolomics profile of urine by nuclear magnetic resonance

Urine samples were centrifuged at 16612 × g at 4 °C. The samples were aliquoted, and 180 µL of the aliquot was added to 20 µL 1 M phosphate buffer containing 2.5 M DSS and 0.03% (v/v) sodium azide. The samples were transferred to 3 mm NMR tubes (Bruker Biospin, Billerica, MA). ¹H NMR spectra were acquired in a 700 MHz Bruker Avance III HD NMR spectrometer (Bruker Biospin, Billerica, MA) fitted with a 3 mm triple resonance inverse (TXI) probe. All spectra were acquired using a NOESYPR1d pulse program with relaxation delay of 1 s, 0.1 s mixing time, 76 k data points and 14ppm spectral width. A total of 256 scans were acquired per sample. Water was suppressed using the presaturation technique during relaxation delay and mixing time. Raw spectral data were imported into Chenomx v8.0 (Chenomx Inc. Edmonton, Alberta, Canada) for further processing. The spectra were Fourier transformed after zero filling to 128 k, and Raw spectral data were imported into Chenomx v8.0 (Chenomx Inc., Edmonton, Alberta, Canada) for further processing. Fourier transformed the spectra after zero filling to 128 k, and linear broadening of 0.1 Hz was applied. All spectra were referenced to an internal standard followed by targeted profiling of metabolites of interest⁵⁰. The peaks of the spectra are profiled using Chenomx software.

Calculation of sample size for experiments in mice

We determined the sample size for the animal experiments using online software (clincalc.com/stats/samplesize.aspx). Considering the human trial outcome of tryptophan concentrations reflecting effect size (Cohen’s d) to be 0.94 between placebo and naproxen treatment, and type I error set at α = 0.05 and a power (1-β) of 80%, a total of 12 samples were needed in each group. However, considering a dropout of mice due to chronic naproxen dosing during the experiment, we decided to start with N = 16 in each group. In addition, we have considered literature for considering sample sizes for different experiments^36,51,52,53.

Tryptophan metabolism in naproxen treated mice

Male and Female C57BL/6 mice were obtained from the Jackson Laboratory (Bar Harbor, ME) at six weeks of age. After two weeks of acclimatization, they were randomly assigned to two groups, n = 16 per group: control chow (Male = 8, Female = 8) and a naproxen diet (Male = 8, Female = 8). Animals were kept in a 12 h day/night cycle with lights on at 7 A.M. and lights off at 7 P.M. The temperature and the humidity were 70 F and 40% respectively. Experimental protocols for the mice experiments were reviewed and approved by the Institute for Animal Care and Use Committee at the University of Pennsylvania (Protocol #804445). The mice had ad libitum access to food and water. Food was provided as pellets: naproxen diet (5001; 1323 ppm for females, 1610 ppm for males) and control diet (Purina Lab Chow, Envigo, Indianapolis, IN). Weights were recorded weekly for all the mice for three weeks, after which they were sacrificed. Blood was collected, and plasma was immediately separated after centrifugation at 3000 × g for 5 minutes (Eppendorf Centrifuge 5424). The plasma obtained was frozen at − 80 °C for analysis. We utilized various murine models to investigate the role of the COX enzyme in tryptophan metabolism. All experimental mice were maintained on the C57BL/6 J genetic background. Sex was not considered as a variable in later experiments using different knockout (KO) models, since male and female wild-type mice showed similar tryptophan depletion responses to a naproxen-containing diet (Fig. 2B).

Platelet specific COX-1 KO mice (Pf4-Δ Cre ^+/-/COX-1^fx/^fx) were generated as described¹¹. The COX-1^fx/fx mice were kindly provided by Harvey Herschman at UCLA⁵⁴. Briefly, male mice (Pf4-Δ Cre ^+/−/COX-1^fx/fx, 8 weeks old) were mated with female mice (COX-1^fx/fx, 8 weeks old) to yield Pf4-Δ Cre ^+/−/COX-1^fx/fx mice as platelet-specific Cox-1 KO mice. Cox-1^fx/fx mice were used as controls. We used mice from 12–14 weeks old in our studies. The Cox-1^fx/fx mice were kindly provided by Harvey Herschman of UCLA⁵⁴. For global Cox-1 deficient mice, Ind-Cre^+/- mice⁵⁵ were mated with Cox-1^fx/fx mice to generate the Ind-Cre^+/- Cox-1^fx/fx mice. The Cre-/- COX-1^fx/fx mice were used as controls. Mice of 12–14 weeks old mice were used in our studies. Ind-Cre^+/-COX-2^fx/fx were generated as described⁴¹. Female COX-2^fx/fx mice were bred with B6 Cre-ER male mice to produce Cre-ER + /− and COX-2^f/+ offspring. Male Cre-ER + /− mice were then intercrossed with female Cox-2^fx/fx mice to generate Cre^+/−/COX-2^fx/fx (Ind.COX-2 KO) mice, and COX-2^fx/fx mice were used as controls. Mice were used for experiments at 3-4 months of age. To avoid the roles of Coxs during development we generated mice in which both Cox-1 and Cox-2 were depleted postnatally in a tamoxifen inducible manner⁴⁰. These Cox-1^fx/fx Cox-2^fx/fx CMV-Cre^+/- mice are abbreviated as “COX-DKO” mice in this paper. The controls were Cre ^-/- controls. Mice of the age were 14-16 weeks of age.

COX enzymes and tryptophan metabolism in mice on Naproxen

We utilized mice in which Cox enzymes were depleted to address the role of their inhibition in mediating the effects of Naproxen on tryptophan. Cox-DKO and their controls were profiled for their plasma tryptophan response to naproxen administration. In a different experiment, control mice were treated with Phenyl propionic acid (PPA) in the diet (1323 ppm R2 Phenyl propionic acid) and were compared to control mice on naproxen (1323 ppm). The mice were allowed to feed ad libitum for 3 weeks prior to plasma collection. Approval for animal studies: experimental protocols were reviewed and approved by the Institute for Animal Care and Use Committee at the University of Pennsylvania (Protocol #804445).

Hemoglobin and calprotectin measurements

Human and mouse fecal samples were assayed for hemoglobin and calprotectin using ELISA kits from MyBioSource (San Diego, CA) and Alpco (Macedon, NY), respectively. The samples were processed according to the manufacturers’ instructions.

Analysis of the human fecal microbiome

Analysis using 16S amplicon sequencing was carried out as described^56,57,58. Briefly, DNA was isolated from ~ 200 mg of stool using the Qiagen PowerSoil Kit (Lot #154048139) following the manufacturer’s protocol. Isolated DNA was quantified using the Picogreen method. Primers to amplify the bacterial 16S rRNA gene region were barcoded to label each sample, and PCR reactions were carried out in triplicate using Accuprime (Invitrogen, Carlsbad, CA, USA). Each reaction contained 5ul of extracted DNA and 0.4uM of each primer, 50 nanograms of DNA and 10 pM of each primer. Primers annealing to the V1V2 region of the 16S bacterial gene were used for amplification as described⁵⁹. Amplified 16S rDNA was purified using a 1:1 volume of Agencourt AmPure XP beads (Beckman-Colter, Brea, CA, USA). The purified products were pooled in equal amounts and analyzed using Illumina MiSeq sequencing. DNA-free water and blank extraction columns were subjected to the same purification and amplification procedure to allow empirical assessment of environmental and reagent contamination. Positive controls were also included, consisting of synthetic DNA plasmids mimicking sequences from well-studied organisms. Quality control and sample analysis was essentially as described in ref. ⁵⁶. For the statistical analysis of microbiome data, we used QIIME (version 1) with default settings as described⁶⁰. Only those OTUs detected in the majority of samples (relative abundances > 0 in at least 99% of samples) were retained for differential analyses. The log10 of the relative abundances of each remaining OTU were evaluated for drugs vs placebo differences using a mixed-effects linear model that included a random effects term for subject (to account for repeated measures). Each OTU was evaluated against the placebo separately for both Celecoxib and Naproxen. OTUs with P-values < 0.10 were selected as an initial threshold for exploratory analysis. All P-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) procedure, with a cut-off of 0.25, as shown in Fig. 4, panels A and B. Supplementary Table 1 presents results based on taxon-level data. For this analysis, we aggregated OTU relative abundances into taxonomic groups before performing differential abundance testing at the taxon level. Taxa with P-values < 0.1, along with their corresponding FDRs, are shown in Supplementary table 1.

We chose this cut-off to present the significantly differential OTUs in order to balance sensitivity and specificity in this exploratory framework. This approach may increase the likelihood of false positive findings and will benefit from follow-up studies to test these findings in validation cohorts. The 16S rRNA has been uploaded to SRA. The Bioproject ID is PRJNA1150540.

Tryptophan rescue experiment in mice

Female C57BL/6 mice (N = 40; N = 8 per group) from Jackson Laboratory (Bar Harbor, ME) were randomly grouped and treated as control, celecoxib, naproxen, tryptophan, and naproxen + tryptophan for three weeks of the study. The mice in the control group were on normal chow while the celecoxib group received chow supplemented with celecoxib (550 ppm), and the naproxen and naproxen + tryptophan groups received chow supplemented with naproxen (naproxen diet, 5001; 1323 ppm). The mice in the naproxen + tryptophan group received daily intraperitoneal (i.p.) injections of 50 mg/kg tryptophan in phosphate buffered saline (PBS) for 1 week, as shown in Fig. 5A. All mice on the naproxen and control group received daily i.p. injections of PBS for 1 week. At the end of the treatment period, mice were sacrificed via CO₂ asphyxiation and their whole heart or jejunum tissues were dissected, chopped into small pieces, and snap frozen in liquid nitrogen. Fecal matter was flushed from the jejuna with PBS, prior to chopping and freezing. Mice and tissues from these groups were treated and processed in three different experimental batches: Experimental batch 1 consisted of heart tissues from control and naproxen mice. Experimental batch 2 included heart tissues from control, naproxen, and naproxen + tryptophan mice. Experimental batch 3 included heart tissues from control and tryptophan mice. Experimental batch 4 included jejunum tissues from control, celecoxib, naproxen, tryptophan, and naproxen + tryptophan mice. Note, all mice, except for those in experimental batch 1, received i.p. injections of either PBS or tryptophan.

RNA Extraction and RNA-Seq library preparation

The Promega Maxwell RSC simplyRNA kit (AS1340) was used to extract RNA from the whole heart or 3 mg of jejunum tissue. RNA samples were quantified with the NanoDrop 8000 Spectrophotometer and Agilent 21200 Bioanalyzer systems. Libraries for each sample were constructed from 400 ng of total RNA using either the Illumina TruSeq Stranded mRNA library prep kit (heart) or the Illumina Stranded Total RNA Prep with Ribo-Zero Plus (jejunum). Jejuna RNA samples were processed with a ribo-depletion based library prep kit because they had RIN scores < 8. After library prep and QC using Tapestation they were normalized to 1.55 nM and run on the Illumina NovaSeq 6000 sequencer using a v1.5 S2 200 cycle kit aiming for 40 million reads per sample.

RNA-Seq Quantification

We used Salmon (v1.9.0)⁶¹ to quantify transcript read counts from the raw FASTQ files against transcripts from v102 of the Ensembl annotation⁶¹. The Salmon index we used for this data includes transcript sequences from the ‘Mus_musculus.GRCm38.cdna.all.fa’ (all Ensembl transcripts, excluding non-coding RNAs) and ‘Mus musculus.GRCm38.ncrna.fa’ (non-coding RNAs) FASTA files, both of which we downloaded from the Ensembl ftp site. We also included the full DNA sequence from the primary assembly (GRCm38; also downloaded from Ensembl) in the Salmon index as a decoy. When running Salmon, we used the following command line parameters to account for sequence, GC, and positional biases: “--seqBias --gcBias –posBias”. We also used the selective alignment mapping strategy (“--validateMappings”) and the Inward Stranded Reverse LIBTYPE (“-l ISR”). Next, we imported the Salmon output into R (v4.2.2) and summarized these data to gene level quantifications with the tximeta package (v1.16.0)⁶². Unless otherwise indicated, all RNA-Seq analyses and visualizations in this manuscript were prepared from the unnormalized, gene-level read counts estimated by Salmon and tximeta.

Differential gene expression

We used DESeq2 (v1.38.1)⁶³ to conduct differential expression (DE) analyses between our three groups of interest (control, celecoxib, naproxen, tryptophan, and naproxen + tryptophan). For the control vs naproxen DE comparison, we used RNA-Seq data from experimental batches 1 and 2. When fitting the model for the control vs naproxen experiment, we included a term to account for any batch effects between the two experiments. We only used data from experimental batch 2 for the control vs. naproxen + tryptophan and naproxen vs. naproxen + tryptophan, DE analyses, because the first experimental batch did not include any mice treated with tryptophan. For the control vs. tryptophan comparison, we used data from experimental batch 3. Lastly, we used data from experimental batch 4 for all jejunum analyses. The tissues in experimental batches 3 and 4 were collected on different days. DE analyses for these experimental batches also included terms to account for dissection day effects. Data from the following samples were excluded from all DE analyses because the source RNA had RIN values < 4: Heart_Control_Exp1_02, Jejunum_Naproxen_Exp4_04, and Jejunum_Naproxen + Tryptophan_Exp4_07. For all comparisons, we allowed DESeq2 to apply its default minimum expression filter, outlier detection, and Benjamini-Hochberg multiple testing adjustment to p-values (also known as q-values), using a target DE q-value cutoff < 0.4. Next, we merged the results of these DE analyses to identify genes affected by naproxen treatment, where those effects were reversed (or ‘rescued’) by tryptophan treatment. Briefly, we identified genes with evidence of DE (q-value < 0.4) in both the ‘control vs naproxen’ and ‘naproxen vs naproxen + tryptophan’ comparisons. We further reduced this list to those with fold-change values showing opposite signs in these two comparisons. In other words, we identified DE genes that showed a simultaneous increase in response to naproxen treatment and a decrease in response to tryptophan supplementation (or vice versa).

Pathway analyses

We conducted all pathway analyses with the QIAGEN Ingenuity Pathway Analysis tool (IPA). Unless otherwise stated, we used the list of all provided genes as the background gene set for enrichment analysis instead of the list of genes contained in the Ingenuity Pathway Analysis knowledge base.

Quantitative real-time PCR of genes

The RNA was extracted using a Promega kit (AS1340). RNA samples were quantified with the NanoDrop 8000 Spectrophotometer and Agilent 21200 Bioanalyzer systems. Quantitative Real-time PCR was performed using Taqman Gene Expression Assays (ThermoFisher Scientific, catalog # 4331187) for IL-1β (Assay ID- Mm00434228_m1), IL-6 (Assay ID- Mm00446190_m1), IL-10 (Assay ID- Mm00439614_m1), and IL-10ra (Assay ID- Mm00434151_m1) with an Applied Biosystems ViiA 7 real-time PCR system in a 384-well plate. Results were normalized to HPRT (Assay ID- Mm01545399_m1).

Statistical analysis

Measurements were taken from distinct samples. The metabolomics data set from plasma and urine from human volunteers was analyzed in R 4.1.2. Statistical testing was reported after correction for multiple testing using the Benjamini-Hochberg (BH) method. For all comparisons of RNA-seq data, we followed Benjamini-Hochberg multiple testing adjustment to p-values also known as q-value. One-way analysis of variance (ANOVA) was performed in Prism 10.4.2. Statistical significance was reported after performing Tukey’s post hoc tests, unless otherwise specified. Two-way analysis of variance (ANOVA) was applied to the targeted analysis of metabolites in Prism 10.4.2, followed by a pairwise test as appropriate. This was also followed by Tukey’s post hoc test. Mann-Whitney and Wilcoxon tests were applied as described in the figure legends. *p ≤ 0.05, **p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The RNA seq data generated in the study is available from Gene Expression Omnibus (GEO) with accession number GSE301033. The microbiome sequencing data generated in human subjects are available from BioProject (A database maintained by NCBI), with accession number PRJNA1150540. All the microbiome sequencing data generated for the mouse are available from BioProject, with accession number PRJNA1331839. All metabolomics data has been submitted to Metabolomics Workbench⁶⁴ (https://www.metabolomicsworkbench.org) under Study ID ST004263. The dataset is accessible at https://doi.org/10.21228/M8Z24S. Source data are provided in this paper.

References

Simmons, D. L., Wagner, D. & Westover, K. Nonsteroidal anti-inflammatory drugs, acetaminophen, cyclooxygenase 2, and fever. Clin. Infect. Dis. 31, S211–S218 (2000).
Article CAS PubMed Google Scholar
Cashman, J. N. The mechanisms of action of NSAIDs in analgesia. Drugs 52, 13–23 (1996).
Article CAS PubMed Google Scholar
Gunaydin, C. & Bilge, S. S. Effects of nonsteroidal anti-inflammatory drugs at the molecular level. Eurasian J. Med. 50, 116–121 (2018).
Article CAS PubMed PubMed Central Google Scholar
Smyth, E. M., Grosser, T., Wang, M., Yu, Y. & FitzGerald, G. A. Prostanoids in health and disease. J. Lipid Res. 50, S423–S428 (2009).
Article PubMed Google Scholar
Bjarnason, I. et al. Mechanisms of damage to the gastrointestinal tract from nonsteroidal anti-inflammatory drugs. Gastroenterology 154, 500–514 (2018).
Article CAS PubMed Google Scholar
Grosser, T., Ricciotti, E. & FitzGerald, G. A. The cardiovascular pharmacology of nonsteroidal anti-inflammatory drugs. Trends Pharmacol. Sci. 38, 733–748 (2017).
Article CAS PubMed PubMed Central Google Scholar
Kearney, P. M. et al. Do selective cyclo-oxygenase-2 inhibitors and traditional non-steroidal anti-inflammatory drugs increase the risk of atherothrombosis? Meta-analysis of randomised trials. BMJ 332, 1302–1308 (2006).
Article CAS PubMed PubMed Central Google Scholar
Laine, L. Gastrointestinal safety of coxibs and outcomes studies: what’s the verdict? J. Pain Symptom Manage. 23, S5–S10 (2002).
Article CAS PubMed Google Scholar
Theken, K. N. et al. Degree of cyclooxygenase-2 inhibition modulates blood pressure response to celecoxib and naproxen. Hypertension 83, e25516 (2026).
Giuliano, F. & Warner, T. D. Ex vivo assay to determine the cyclooxygenase selectivity of non-steroidal anti-inflammatory drugs. Br. J. Pharmacol. 126, 1824–1830 (1999).
Article CAS PubMed PubMed Central Google Scholar
Tang, S. Y. et al. Differential impact in vivo of Pf4-ΔCre–mediated and Gp1ba-ΔCre–mediated depletion of cyclooxygenase-1 in platelets in mice. Arterioscler. Thromb. Vasc. Biol. 44, 1393–1406 (2024).
Article CAS PubMed PubMed Central Google Scholar
Zhang, Z., Silveyra, E., Jin, N. & Ribelayga, C. P. A congenic line of the C57BL/6J mouse strain that is proficient in melatonin synthesis. J. Pineal Res. 65, e12509 (2018).
Article PubMed PubMed Central Google Scholar
Fournel, S. & Caldwell, J. The metabolic chiral inversion of 2-phenylpropionic acid in rat, mouse and rabbit. Biochem. Pharmacol. 35, 4153–4159 (1986).
Article CAS PubMed Google Scholar
Badawy, A. A.-B. & Guillemin, G. The plasma [Kynurenine]/[Tryptophan] ratio and indoleamine 2,3-dioxygenase: time for appraisal. Int. J. Tryptophan Res. 12, 1178646919868978 (2019).
Article PubMed PubMed Central Google Scholar
Lu, Y. et al. TrpNet: Understanding Tryptophan Metabolism across Gut Microbiome. Metabolites 12, https://doi.org/10.3390/metabo12010010 (2021).
Notting, F., Pirovano, W., Sybesma, W. & Kort, R. The butyrate-producing and spore-forming bacterial genus Coprococcus as a potential biomarker for neurological disorders. Gut Microbiome 4, e16 (2023).
Article PubMed PubMed Central Google Scholar
Nguyen, T. L. A., Vieira-Silva, S., Liston, A. & Raes, J. How informative is the mouse for human gut microbiota research? Dis. Model. Mech. 8, 1–16 (2015).
Article CAS PubMed PubMed Central Google Scholar
Zhao, P., Chen, Y., Zhou, S. & Li, F. Microbial modulation of tryptophan metabolism links gut microbiota to disease and its treatment. Pharmacol. Res. 219, 107896 (2025).
Article CAS PubMed Google Scholar
Wu, P. H. et al. Exploring the relationship between gut microbiome composition and blood indole-3-acetic acid in hemodialysis patients. Biomedicines 12, https://doi.org/10.3390/biomedicines12010148 (2024).
Ley, R. E. et al. Obesity alters gut microbial ecology. Proc. Natl. Acad. Sci. USA 102, 11070–11075 (2005).
Article ADS CAS PubMed PubMed Central Google Scholar
Xiao, L. et al. A catalog of the mouse gut metagenome. Nat. Biotechnol. 33, 1103–1108 (2015).
Article ADS CAS PubMed Google Scholar
Sved, A. F., Van Itallie, C. M. & Fernstrom, J. D. Studies on the antihypertensive action of L-tryptophan. J. Pharmacol. Exp. Ther. 221, 329–333 (1982).
Article CAS PubMed Google Scholar
Robert, M. & Miossec, P. Effects of Interleukin 17 on the cardiovascular system. Autoimmun. Rev. 16, 984–991 (2017).
Article CAS PubMed Google Scholar
Peng, H. et al. Profibrotic Role for Interleukin-4 in cardiac remodeling and dysfunction. Hypertension 66, 582–589 (2015).
Article CAS PubMed PubMed Central Google Scholar
Liang, H. & Ward, W. F. PGC-1α: a key regulator of energy metabolism. Adv. Physiol. Educ. 30, 145–151 (2006).
Article PubMed Google Scholar
Saettini, F. et al. Absent B cells, agammaglobulinemia, and hypertrophic cardiomyopathy in folliculin-interacting protein 1 deficiency. Blood 137, 493–499 (2021).
Article CAS PubMed PubMed Central Google Scholar
Szondy, Z. et al. Tissue transglutaminase (TG2) protects cardiomyocytes against ischemia/reperfusion injury by regulating ATP synthesis. Cell Death Differ 13, 1827–1829 (2006).
Article CAS PubMed Google Scholar
Ata, R. & Antonescu, C. N. Integrins and Cell Metabolism: An Intimate Relationship Impacting Cancer. Int. J. Mol. Sci. 18, https://doi.org/10.3390/ijms18010189 (2017).
Ohashi, K. et al. Adiponectin promotes macrophage polarization toward an anti-inflammatory phenotype. J. Biol. Chem. 285, 6153–6160 (2010).
Article CAS PubMed Google Scholar
Karaskova, E. et al. Role of adipose tissue in inflammatory bowel disease. Int. J. Mol. Sci. 22, 4226 (2021).
Iwata, Y. et al. CD19, a response regulator of B lymphocytes, regulates wound healing through hyaluronan-induced TLR4 signaling. Am. J. Pathol. 175, 649–660 (2009).
Article CAS PubMed PubMed Central Google Scholar
Ma, Y. et al. The COL7A1/PI3K/AKT axis regulates the progression of cholangiocarcinoma. Heliyon 10, e37361 (2024).
Article CAS PubMed PubMed Central Google Scholar
Zhang, R. X. et al. FNDC1 is a myokine that promotes myogenesis and muscle regeneration. EMBO J. 44, 30–53 (2025).
Article CAS PubMed Google Scholar
Davis, A. & Robson, J. The dangers of NSAIDs: look both ways. Br. J. Gen. Pract. 66, 172–173 (2016).
Jung, J. et al. 1H NMR-based metabolic profiling of naproxen-induced toxicity in rats. Toxicol. Lett. 200, 1–7 (2011).
Article CAS PubMed Google Scholar
Chang, I. J. & Harris, R. C. Are All COX-2 Inhibitors Created Equal? Hypertension 45, 178–180 (2005).
Article CAS PubMed Google Scholar
Vonkeman, H. E. & van de Laar, M. A. F. J. Nonsteroidal anti-inflammatory drugs: adverse effects and their prevention. Semin. Arthritis Rheum. 39, 294–312 (2010).
Article CAS PubMed Google Scholar
Grosser, T., Yu, Y. & FitzGerald, G. A. Emotion recollected in tranquility: lessons learned from the COX-2 Saga. Annu. Rev. Med. 61, 17–33 (2010).
Article CAS PubMed Google Scholar
Barekat, K. et al. Concomitant suppression of COX-1 and COX-2 is insufficient to induce enteropathy associated with chronic NSAID use. J. Clin. Invest. e190575. https://doi.org/10.1172/JCI190575 (2026).
Yu, Z. et al. Disruption of the 5-lipoxygenase pathway attenuates atherogenesis consequent to COX-2 deletion in mice. Proc. Natl. Acad. Sci. USA 109, 6727–6732 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Seymour, B. J. et al. Microbiota-dependent indole production is required for the development of collagen-induced arthritis. J. Clin. Invest. 134, e167671 (2023).
Candeliere, F. et al. Indole and p-cresol in feces of healthy subjects: Concentration, kinetics, and correlation with microbiome. Front. Mol. Med. 2, https://doi.org/10.3389/fmmed.2022.959189 (2022).
Chappell, C. L. et al. Tryptophan metabolites and their predicted microbial sources in fecal samples from healthy individuals. Preprint at https://doi.org/10.1101/2023.12.20.572622 (2024).
Kirwan, J. A. et al. Quality assurance and quality control reporting in untargeted metabolic phenotyping: mQACC recommendations for analytical quality management. Metabolomics 18, 70 (2022).
Article CAS PubMed PubMed Central Google Scholar
Malik, D. M., Rhoades, S. & Weljie, A. Extraction and analysis of pan-metabolome polar metabolites by ultra performance liquid chromatography–tandem mass spectrometry (UPLC-MS/MS). Bio. Protoc. 8, e2715 (2018).
Song, W. et al. Noninvasive assessment of the role of cyclooxygenases in cardiovascular health: a detailed HPLC/MS/MS method. in Lipidomics and Bioactive Lipids: Specialized Analytical Methods and Lipids in Disease 433, 51–72 (Academic Press, 2007).
Patrignani, P., Filabozzi, P. & Patrono, C. Selective cumulative inhibition of platelet thromboxane production by low-dose aspirin in healthy subjects. J. Clin. Invest. 69, 1366–1372 (1982).
Article CAS PubMed PubMed Central Google Scholar
Panara, M. R. et al. Dose-dependent inhibition of platelet cyclooxygenase-1 and monocyte cyclooxygenase-2 by meloxicam in healthy subjects. J. Pharmacol. Exp. Ther. 290, 276–280 (1999).
Article CAS PubMed Google Scholar
Westbrook, R. et al. Kynurenines link chronic inflammation to functional decline and physical frailty. JCI Insight 5, e136091 (2020).
Slupsky, C. M. et al. Investigations of the effects of gender, diurnal variation, and age in human urinary metabolomic profiles. Anal. Chem. 79, 6995–7004 (2007).
Article ADS CAS PubMed Google Scholar
Ghosh, S. Metabolomic studies for metabolic alterations induced by non-steroidal anti-inflammatory drugs: mini review. Biomolecules 11, https://doi.org/10.3390/biom11101456 (2021).
Takeuchi, K. et al. Metabolic profiling to identify potential serum biomarkers for gastric ulceration induced by nonsteroid anti-inflammatory drugs. J. Proteome Res. 12, 1399–1407 (2013).
Article CAS PubMed Google Scholar
Um, S. Y. et al. Pattern recognition analysis for the prediction of adverse effects by nonsteroidal anti-inflammatory drugs using 1H NMR-based metabolomics in rats. Anal. Chem. 81, 4734–4741 (2009).
Article ADS CAS PubMed Google Scholar
Mitchell, J. A. et al. Cell-specific gene deletion reveals the antithrombotic function of COX1 and explains the vascular COX1/prostacyclin paradox. Circ. Res. 125, 847–854 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hayashi, S. & McMahon, A. P. Efficient recombination in diverse tissues by a tamoxifen-inducible form of Cre: a tool for temporally regulated gene activation/inactivation in the mouse. Dev. Biol. 244, 305–318 (2002).
Article CAS PubMed Google Scholar
Kim, D. et al. Optimizing methods and dodging pitfalls in microbiome research. Microbiome 5, 52 (2017).
Article PubMed PubMed Central Google Scholar
Lozupone, C. A., Hamady, M., Kelley, S. T. & Knight, R. Quantitative and qualitative beta diversity measures lead to different insights into factors that structure microbial communities. Appl. Environ. Microbiol. 73, 1576–1585 (2007).
Article ADS CAS PubMed PubMed Central Google Scholar
McKenna, P. et al. The macaque gut microbiome in health, lentiviral infection, and chronic enterocolitis. PLoS Pathog. 4, e20 (2008).
Article PubMed PubMed Central Google Scholar
Wu, G. D. et al. Sampling and pyrosequencing methods for characterizing bacterial communities in the human gut using 16S sequence tags. BMC Microbiol. 10, 206 (2010).
Article PubMed PubMed Central Google Scholar
Caporaso, J. G. et al. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7, 335–336 (2010).
Yates, A. D. et al. Ensembl 2020. Nucleic Acids Res. 48, D682–D688 (2020).
CAS PubMed PubMed Central Google Scholar
Love, M. I. et al. Tximeta: Reference sequence checksums for provenance identification in RNA-seq. PLoS Comput. Biol. 16, e1007664 (2020).
Article CAS PubMed PubMed Central Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderateds estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Sud, M. et al. Metabolomics Workbench: An international repository for metabolomics data and metadata, metabolite standards, protocols, tutorials and training, and analysis tools. Nucleic Acids Res. 44, D463–D470 (2016).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

Dr. FitzGerald is the McNeil Professor of Translational Medicine and Therapeutics. We acknowledge the technical support of Fengxiang Wang. 16S sequencing was provided by the CHOP Microbiome Center, Children’s Hospital of Philadelphia. The study was supported by a grant (UL1TR001878) from the National Center for Advancing Translational Sciences of the NIH. We gratefully acknowledge the Metabolomics Workbench for providing data hosting and sharing infrastructure, supported by NIH grants U2C-DK119886 and OT2-OD030544. Supported by a grant (UL1TR001878) from the National Center for Advancing Translational Sciences of the NIH.

Author information

Authors and Affiliations

Institute for Translational Medicine and Therapeutics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA
Soumita Ghosh, Nicholas F. Lahens, Kayla Barekat, Soon-Yew Tang, Robin Joshi, Tilo Grosser & Garret A. FitzGerald
Department of Oral and Maxillofacial Surgery and Pharmacology, School of Dental Medicine, University of Pennsylvania, Philadelphia, PA, USA
Katherine N. Theken
Department of Systems Pharmacology and Translational Therapeutics, University of Pennsylvania, Philadelphia, PA, USA
Emanuela Ricciotti, Arjun Sengupta, Aalim Weljie & Garret A. FitzGerald
Division of Gastroenterology, Hepatology, and Nutrition, The Children’s Hospital of Philadelphia, Philadelphia, PA, USA
Weiming Hu
Department of Microbiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA
Frederic D. Bushman
Department of Translational Pharmacology, EWL School of Medicine, Bielefeld University, Bielefeld, Germany
Tilo Grosser
Department of Medicine, School of Medicine, University of Pennsylvania, Philadelphia, PA, USA
Garret A. FitzGerald

Authors

Soumita Ghosh
View author publications
Search author on:PubMed Google Scholar
Nicholas F. Lahens
View author publications
Search author on:PubMed Google Scholar
Kayla Barekat
View author publications
Search author on:PubMed Google Scholar
Soon-Yew Tang
View author publications
Search author on:PubMed Google Scholar
Katherine N. Theken
View author publications
Search author on:PubMed Google Scholar
Emanuela Ricciotti
View author publications
Search author on:PubMed Google Scholar
Arjun Sengupta
View author publications
Search author on:PubMed Google Scholar
Robin Joshi
View author publications
Search author on:PubMed Google Scholar
Weiming Hu
View author publications
Search author on:PubMed Google Scholar
Frederic D. Bushman
View author publications
Search author on:PubMed Google Scholar
Aalim Weljie
View author publications
Search author on:PubMed Google Scholar
Tilo Grosser
View author publications
Search author on:PubMed Google Scholar
Garret A. FitzGerald
View author publications
Search author on:PubMed Google Scholar

Contributions

Funding: G.A.F. and T.G. Writing: G.A.F., S.G., T.G., N.L., and A.S. Conceptualization: G.A.F., T.G., S.G., E.R., and K.B. Data Interpretation: G.A.F., S.G., T.G., S.Y.T., K.B., and E.R. Clinical trial experiment: K.T., G.A.F. and T.G. Nuclear Magnetic Resonance experiments: A.S. and A.W. Mass Spectrometry: S.G. and R.J. Microbiome data analysis: F.D.B. and W.H. RNA-seq data analysis: N.L. Animal experiment: S.Y.T. and S.G.

Corresponding authors

Correspondence to Tilo Grosser or Garret A. FitzGerald.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Paola Patrignani, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information (download PDF )

Description Of Additional Supplementary File (download PDF )

Supplementary Data 1 (download XLSX )

Supplementary Data 2 (download XLSX )

Supplementary Data 3 (download XLSX )

Supplementary Data 4 (download XLSX )

Reporting summary (download PDF )

Transparent Peer Review file (download PDF )

Source data

Source Data (download XLSX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Ghosh, S., Lahens, N.F., Barekat, K. et al. Depression of tryptophan may contribute to adverse effects of naproxen. Nat Commun 17, 2776 (2026). https://doi.org/10.1038/s41467-026-69684-7

Download citation

Received: 26 November 2024
Accepted: 05 February 2026
Published: 14 February 2026
Version of record: 25 March 2026
DOI: https://doi.org/10.1038/s41467-026-69684-7