COX-1–derived thromboxane A₂ plays an essential role in early B-cell development via regulation of JAK/STAT5 signaling in mouse

B-cell development is a stepwise process that requires tight coordination between cytokines and induced transcription factors, as well as cross-talk between hematopoietic progenitors and the bone marrow (BM) microenvironment in which they reside. The earliest step in B-cell commitment is the generation of common lymphoid progenitors (CLPs) from hematopoietic stem cells (HSCs) through multiple steps. CLPs give rise to pre-pro-B cells, pro-B cells, pre-B cells, immature B cells, and finally develop into mature B cells in BM. Mature B cells migrate to peripheral lymphoid tissues, where they undergo activation and produce specific antibodies in response to antigen exposure.^1,2 

The molecular mechanism dictating early B-cell development within the BM has been extensively investigated, and a small set of transcription factors has been identified as essential regulators of this process.^3,4  Among these, the role of interleukin (IL)-7–induced activation of the Janus kinase/signal transducer and activator of transcription 5 (JAK/STAT5) pathway has been well documented.^5,6  Specifically, CLPs express the receptor for IL-7 (IL-7R),⁷  and IL-7 can be produced by both stromal cells and some progenitor cells in BM environment.⁸  Binding of IL-7 to IL-7R on CLPs activates JAK/STAT5 signaling, which induces the transcription of genes essential for B-cell development, including Pax5 and Ebf1.^9-11  The necessity of the JAK/STAT5 pathway in B-cell development has been firmly established; however, the upstream mechanism regulating JAK/STAT5 activity remains poorly understood.

Cyclooxygenases (COXs) are enzymes that metabolize arachidonic acid and produce prostanoids, including prostaglandins (PGs) and thromboxanes (TXs), in combination with specific PG and TX synthases.¹²  Prostanoids, acting through G-protein–coupled receptors, elicit intracellular signaling pathways and play crucial roles in a variety of (patho)physiological processes, depending on specific metabolite and also on the cell type.^13,14  Two isoforms of COX enzymes, COX-1 and COX-2, have been identified. They differ in terms of gene structure, tissue distribution, and protein functions.¹⁵  COX-1 is constitutively expressed in most cell types, whereas COX-2 is induced in response to certain stimuli.¹⁶  Nonsteroidal anti-inflammatory drugs, including aspirin, exhibit clinically antithrombotic and anti-inflammatory efficacies by inhibition of COX-derived prostanoid production.^17,18 

In this study, we wished to investigate the role of COX-1 in the development of immune cells. We found that COX-1 deficiency specifically impairs early B-cell development through downregulation of thromboxane A2 (TxA2)-TxA2 receptor (TP)–mediated JAK3/STAT5 signaling via the cyclic adenosine monophosphate (cAMP)-protein kinase A pathway (PKA) axis. Furthermore, administration of low-dose aspirin reduced B-cell levels in humans. Thus, our observations demonstrated that COX-1–derived TxA₂ represents a novel regulator of early B-cell development.

All experiments performed on mice were approved by the Institutional Animal Care and Use Committee of Sun Yat-Sen University. COX-1^−/− mice were maintained on a mixed C57BL/6 × Sv129 genetic background (50%:50%) for >20 generations,¹⁹  TP^−/− mice were maintained on a C57BL/6 background. All mice were housed in a specific pathogen-free facility, and age-matched littermates (6-8 weeks) were used as controls in this study.

Single-cell suspensions were prepared and were stained with fluorochrome-conjugated antibodies. Data were collected on a BD LSRII flow cytometer (BD Biosciences, San Jose, CA) and analyzed with FlowJo software (Tree Star, Ashland, OR). Data were acquired as the fraction of labeled cells within a live-cell gate set for 50 000 events. For flow cytometric sorting, cells were stained with specific antibodies and isolated on a BD FACSAria cell sorter (BD Bioscience).

Healthy volunteers who had no reported history of chronic illness, no drug allergy, or no bleeding tendency were enrolled. They are administrated daily oral aspirin (100 mg/day) for 10 days. Peripheral blood mononuclear cells were isolated by Ficoll centrifugation before and at the end of administration. The study was approved by Ethics Review Board of Sun Yat-Sen University, and all participants gave written informed consent in accordance with the Declaration of Helsinki.

Statistical tests were performed using GraphPad Prism version 5.0a software and SPSS 170. Unpaired Student t tests and analysis of variance (ANOVA) were used to confirm most comparisons. Statistical analysis was performed using the paired Student t test in the human aspirin experiment. Correlations between different parameters were analyzed using the Spearman rank test. P < .05 was considered significant.

To investigate the physiological role of COX-1 during hematopoiesis, we analyzed the effect of COX-1 deficiency on the hematopoietic system by comparing COX-1^−/− mice with wild-type (WT) littermates. The absence of COX-1 expression in the BM and spleen from the COX-1^−/− mice was confirmed by immunoblotting (supplemental Figure 1 available on the Blood Web site).

We first evaluated the frequencies of B cells and other immune cells in several tissues, including BM, spleen, peripheral blood, and lymph nodes, by flow cytometric analysis. The total numbers of nucleated cells in multiple tissues remained largely unchanged in COX-1^−/− mice compared with WT controls (data not shown). We found that COX-1^−/− mice displayed a twofold reduction in the percentage of B220⁺CD19⁺ B cells (Figure 1A-B); the absolute number of B cells was consistently reduced in various tissues of COX-1^−/− mice, reaching only ∼50% of the level present in BM of WT mice (Figure 1C). In contrast, the proportions and absolute cell counts of T lymphocytes were not noticeably different between COX-1^−/− and WT mice (Figure 1A-C). The proportion of different types of myeloid cells, including immature myeloid cells, dendritic cells, and macrophages, did not display any changes between WT and COX-1^−/− mice (supplemental Figure 2). Taken together, these observations indicated that COX-1 is specifically essential for B-cell homeostasis.

Generation of B cells is coordinated by B-cell development and B-cell survival.²⁰  The B-cell reduction in COX-1^−/− mice prompted us to investigate the underlying mechanism. We first examined whether it was due to enhanced B-cell apoptosis caused by COX-1 deficiency. We failed to observe any difference in the frequency of apoptotic cells in the presence or absence of COX-1, in either B cells or T cells, by annexin V staining (Figure 2A).

Flow cytometric analysis revealed that the frequency of hematopoietic progenitors, including HSCs, CLPs, and common myeloid cells, did not differ between WT and COX-1^−/− mice (supplemental Figure 3), indicating that the B-cell defect caused by COX-1 deficiency may occur at a stage later than CLPs.

To gain insights into the relationship between COX-1 and early B-cell development, we first evaluated the expression of COXs in developing B cells using quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR). In contrast to the steadily low expression of COX-2 in these B-cell compartments, the expression of COX-1 was found to be dynamic in developing B cells, peaking at the pro-B stage. The mRNA expression level of COX-1 in pro-B cells was comparable to that of IL-7R and Pax5, 2 crucial regulators of early B-cell development (Figure 2B). These results indicate a potential role for COX-1 in the stage transition of developing B cells.

We next examined B cells in the mouse BM, the primary tissue in which early B-cell development occurs. The frequencies and absolute numbers of pre-B and immature B cells were markedly reduced in COX-1^−/− mice. The proportion of pro-B cells displayed a moderate but significant increase in COX-1^−/− mice, whereas the level of pre-pro-B cells remained essentially unchanged (Figure 2C). These observations collectively indicated that COX-1 deficiency impacts the transition from pro-B to pre-B stage.

In an aim to further specify the effect of COX-1 on early B-cell development, an in vitro B-cell differentiation experiment was performed. As expected, the capability of progenitors in BM to generate pre-B cells was significantly lower in COX-1^−/− mice than in WT controls (Figure 2D). Consistently, purified pro-B cells from COX-1^−/− mice also displayed impairment in differentiating into pre-B cells compared with the corresponding controls (Figure 2D). Taken together, our data demonstrated that COX-1 is required for the transition from pro-B to pre-B stage during early B-cell development.

To investigate whether the B-cell defect in COX-1^−/− mice was cell intrinsic, we performed BM transplantation experiments. We mixed BM cells from WT or COX-1^−/− (CD45.2⁺) with BM from syngenic (CD45.1⁺), at a 1:1 ratio, and cells were then transplanted into lethally irradiated syngenic mice.^21,22  After 6 weeks, flow cytometry analysis was used to evaluate the relative percentages of B cells expressing the CD45.2 (donor)- or CD45.1 (recipient)-derived population. Results showed that the ability of COX-1^−/− BM to generate B220⁺CD19⁺ cells was significantly decreased compared with the BM cells from WT mice (Figure 2E). The levels of pre-B and immature B cells were decreased almost twofold in the COX-1^−/− BM compared with their WT control (Figure 2F-G).

These data indicated that the defect of B-cell development in COX-1–deficient mice is cell autonomous rather than being associated with the BM microenvironment.

To explore the molecular mechanism underlying the impaired B-cell development caused by COX-1 deficiency, we first profiled the expression of some transcription factors that are crucial for B-cell commitment.⁴  qRT-PCR showed a marked reduction in mRNA levels of Pax5 and Ebf1 in the purified pro-B and pre-B cells from COX-1^−/− mice compared with those from WT controls. In contrast, the expression levels of other transcription factors did not display marked differences (Figure 3A). The downregulation of Ebf1 and Pax5 by COX-1 deletion was further confirmed by immunoblotting (Figure 3B).

To specify the stage at which gene downregulation occurs, the expression levels of Pax5 and Ebf1 in developing B cells were measured by qRT-PCR. A dramatic reduction in both Pax5 and Ebf1 expression levels was observed in pre-B and immature B cells from COX-1^−/− mice in comparison with those from WT controls (Figure 3C).

Because both Pax5 and Ebf1 are transcriptional targets of IL-7–induced JAK/STAT5 signaling in B cells,¹¹  we therefore proposed that COX-1 deficiency may downregulate JAK/STAT5 signaling. Flow cytometric analysis showed that the pro-B and pre-B cells from COX-1^−/− mice failed to show much difference in STAT5 phosphorylation (p-STAT5) levels before and after IL-7 stimulation, whereas B cells from WT control expectedly displayed enhanced STAT5 phosphorylation on IL-7 treatment (Figure 3D). Moreover, we also observed a failure in the induction of STAT5 target genes in these B-cell populations from COX-1^−/− mice in response to IL-7 treatment (Figure 3E). Collectively, these results demonstrated that COX-1 deletion impaired IL-7–induced JAK/STAT5 signaling in B cells.

To investigate whether JAK/STAT5 signaling plays an essential role in perturbed B-cell development of COX-1–deficient mice, we modulated STAT5 signaling, either pharmacologically or genetically, in B cells. First, we used the JAK inhibitor WHI-P131 to block phosphorylation of STAT5 in differentiating B cells.²³  We found that WHI-P131 blocks the development of B cells to the same extent as cells treated with SC-560, a specific inhibitor of COX-1, although WHI-P131 did not enhance the inhibitory effect of SC-560. These results indicated that JAK/STAT5 signaling mediates the regulatory effect of COX-1 on B cells (Figure 3F). Second, lentiviral plasmid expressing constitutively activated STAT5 was used to infect BM cells cultured for B-cell differentiation. The transduction efficiency was >20% (data not shown), and B-cell phenotype was evaluated among green fluorescent protein (GFP)⁺ cells. Results showed that transduction of constitutively activated STAT5 almost completely rescued the impaired B-cell differentiation in COX-1^−/− mice to the level of WT control (Figure 3G). However, overexpression of Pax5 only partially recovered the defect (Figure 3H). Our observations therefore demonstrated that the JAK/STAT5 signaling pathway plays an essential role in mediating the effect of COX-1 on B-cell development.

In an attempt to investigate the mechanism underlying COX-1–mediated B-cell development, we first examined the prostanoid profiles of sorted pro-B and pre-B cells from COX-1^−/− and WT mice using mass spectrum assays. Results showed that the content of thromboxane (Tx)B₂, the stable metabolite of TxA₂, in B cells from COX-1^−/− mice was reduced to ∼50% of the corresponding control; PGE₂ and PGD₂ levels were also reduced to some extent, whereas the amounts of PGF₂ and 6-keto-PGF_1α remained largely unchanged (Figure 4A). We further measured the amount of Leukotriene B₄ in B220⁺ B cells to determine whether there was compensatory overproduction of leukotrienes through an alternative arachidonic acid metabolic pathway in COX-1^−/− mice.¹²  No significant difference was observed between COX-1^−/− and WT mice (supplemental Figure 4).

Further in vitro B-cell differentiation experiments showed that administration of PGE₂ or PGD₂ did not cause any effect on B-cell development (supplemental Figure 5), which excluded their possible participation in the COX-1 effect on B cells.

TxA₂ exerts its action through the specific G-protein–coupled TxA₂ receptor (TP).²⁴  qRT-PCR results showed that TP was abundantly expressed in both pro-B and pre-B cells from mouse BM (Figure 4B). Interestingly, TP expression was dynamic in developing B cells, peaking at the pro-B cell stage (Figure 4C), raising the possibility that TxA₂-TP signaling may play a potential role in B-cell development. Flow cytometric analysis showed that the frequencies of total B, pre-B, and immature B cells were consistently reduced in BM from TP^−/− mice compared with that in WT controls. The proportions of pro-B and pre-pro-B cells in TP^−/− BM were largely unchanged (Figure 4D). The total B-cell numbers in spleen, peripheral blood, and lymph nodes were also clearly reduced in TP^−/− mice (supplemental Figure 6). A further mixed BM transplantation experiment confirmed that the B-cell defect in TP^−/− mice was cell autonomous (Figure 4E, 4F). These observations demonstrated that TP deficiency caused a disturbed early B-cell development.

We next investigated whether TxA₂ mediates the effect of COX-1 on B-cell development. TP agonist U46619 and TP antagonist SQ29548 were used to modulate TP signaling in differentiating B cells. As expected, administration of SQ29548 in BM from WT mice significantly suppressed B-cell differentiation, whereas activation of TP by U46619 almost completely rescued the impaired B-cell differentiation in COX-1^−/− mice to that seen in WT littermates, as evidenced by the frequencies of pre-B cells in BM cultures in vitro (Figure 5A).

The rescue effect of TP agonist on B-cell differentiation in COX-1^−/− mice was further confirmed in vivo. U46619 was administered to COX-1^−/− and WT mice by intraperitoneal injection twice weekly. B-cell phenotype was evaluated by flow cytometry 3 weeks after injection. Results showed that the frequencies of total B cells and pre-B cells in BM from COX-1^−/− mice were recovered to the levels of those in phosphate-buffered saline–treated WT mice, after administration of TP agonist U46619 (Figure 5B). These observations strongly indicated that TxA₂ mediates the effect of COX-1 on B-cell development through the TP receptor.

Based on our observation that JAK/STAT5 signaling mediated the impaired B-cell development caused by COX-1 deletion, we were then interested to establish whether JAK/STAT5 signaling was regulated by COX-1–derived TxA₂. Results showed that the level of p-STAT5 in sorted pro-B and pre-B cells from TP^−/− mice was markedly reduced on IL-7 stimulation compared with the corresponding control cells from WT mice (Figure 5C). Expression of the STAT5 target gene Pax5 showed a similar trend (data not shown). The TP agonist enhanced the p-STAT5 level in B cells from WT mice, whereas a TP antagonist displayed reverse change (supplemental Figure 7). More importantly, the TP agonist could rescue defective JAK/STAT5 signaling in B cells from COX-1^−/− mice (Figure 5D). These observations demonstrated that JAK/STAT5 signaling was subject to regulation by the TxA₂-TP axis in developing B cells.

On binding to TxA₂, TP couples with G-proteins and elicits the activation of intracellular effector molecules and signaling pathways, including at least adenylyl cyclase (AC), phosphatidylinositol-specific phospholipase C, and small G-protein ρ-specific guanine nucleotide exchange factor.^25,26  AC catalyzes the production of the second messenger cAMP, which activates downstream kinases, primarily PKA; phosphatidylinositol-specific phospholipase C leads to elevation of intracellular Ca²⁺ concentration and activation of phospholipase C (PKC); and small G-protein ρ-specific guanine nucleotide exchange factor activates the ρ-mediated signaling.^27,28  To investigate which molecule mediates the regulatory effect of TxA₂-TP on JAK/STAT5 signaling, we used specific inhibitors for these molecules in WT BM cells pretreated with the TP agonist U46619. Flow cytometric analysis showed that the PKA inhibitor H89 clearly counteracted the effect of U46619 on B-cell differentiation, which was almost comparable to the effect of JAK3 inhibitor WHI-P131. Other inhibitors, however, did not display any noticeable effects (Figure 6A). Consistently, administration of H89 abrogated the effect of TP agonist on STAT5 phosphorylation (Figure 6B) and Pax5 expression (data not shown). These observations support the possibility that TP may function through PKA in regulating JAK/STAT5 signaling in developing B cells.

We next investigated whether TP could elicit cAMP-PKA events in B cells, Data showed that the TP agonist U46619 markedly increased the intracellular cAMP level and PKA kinase activity in B cells stimulated with arachidonic acid, which could be virtually completely abrogated by the AC inhibitor MDL12330A; the AC activator forskolin stimulated the activation of the cAMP-PKA axis as expected (Figure 6C). In line with this, B cells from COX-1^−/− mice displayed marked downregulation of cAMP-PKA signaling, whereas the TP agonist could recover this defect (Figure 6D). These observations suggested that TP functionally couples with the Gs α protein in B cells.

We next asked whether PKA phosphorylated STAT5 directly or through the upstream kinase JAKs. Data showed that the PKA activator 6-Bnz-cAMP enhanced the level of p-STAT5 on IL-7 stimulation compared with the corresponding control cells, whereas pretreatment with the JAK3 inhibitor WHI-P131 abrogated this effect (Figure 6E). These results indicated that PKA phosphorylates STAT5 through JAK3. Further immunoblotting experiments demonstrated that the PKA activator led to an increase in phosphorylation of JAK3 (tyrosine 980/981), but not JAK1, whereas a PKA inhibitor reversed this effect (Figure 6F). Consistently, we observed a clear reduction in the level of p-JAK3 in IL-7–stimulated BM cells from COX-1^−/− mice compared with that in WT controls. The TP agonist, however, abrogated this difference (Figure 6G). These results indicate that JAK3 is a potential substrate of PKA in B cells. Collectively, these observations demonstrated that the JAK/STAT5 pathway is regulated by TxA₂-TP through the cAMP-PKA axis in B cells.

Low-dose (50-100 mg/day) aspirin has a clinically well-established effect in reducing the incidence of myocardial infarction and stroke,²⁹  which is attributed to the inhibition of COX-1–mediated TxA₂ synthesis. To determine whether low-dose aspirin has any effects on human B-cell levels, we investigated the levels of B cells in peripheral blood from healthy volunteers (n = 8) before and after a 10-day low-dose aspirin regime. As expected, the TxA₂ metabolites in the urine were markedly reduced in all participants after aspirin administration (supplemental Figure 8). Both the frequencies and absolute numbers of total B cells were decreased significantly after taking aspirin compared with those before administration (Figure 7A-B). Interestingly, we found that the B-cell levels correlated with urinary TxA₂ metabolites, both before (r = 0.6154, P = .0332) and after (r = 0.7665, P = .0368) aspirin uptake. The reduction in B cell levels by aspirin was not caused by enhanced B-cell apoptosis (supplemental Figure 8). Instead, the phosphorylation of STAT5, as well as the expression of its target genes, decreased significantly in human B cells after aspirin (Figure 7C). Stimulation of human B cells with arachidonic acid clearly enhanced the activity of STAT5 signaling. In line with the absence of IL7R expression on human B cells, administration of IL-7 failed to display any effect on STAT5 signaling (Figure 7D). These data suggest an important role of arachidonic acid metabolites in the regulation of JAK/STAT5 signaling in human B cells. Collectively, our observations indicated a potential adverse effect of low-dose aspirin on B-cell development.

Besides their well-established effects on inflammation and the cardiovascular system, emerging roles for COXs and their metabolites in the immune system are gaining increasing attention in recent years, especially for COX-2 and its metabolites.^30-33  The roles of COX-1 in immune system, however, are far from being understood.^34,35  In this study, we demonstrated a novel role of COX-1 in early B-cell development and uncovered the underlying molecular mechanism.

IL-7–induced JAK/STAT5 signaling plays fundamental roles in the regulation of B-cell lineage commitment and development.¹⁰  The mechanisms regulating the activity of JAK/STAT5 signaling in B cells remains unclear. In this study, we found that COX-1–derived TxA₂ could regulate JAK/STAT5 signaling via cAMP-PKA, on binding its receptor TP. The expression of both COX-1 and TP is inducible in developing B cells, which causes the activation of the JAK/STAT5 pathway through the intracellular cAMP-PKA event and finally leads to B-cell development. A model for the role of COX-1 in the regulation of early B-cell development is proposed in Figure 7E. The mechanisms regulating the transcription of COX-1 and TP in developing B cells, however, remain to be investigated.

TxA₂ has established platelet aggregating and vessel-contracting activities.²⁷  To date, reports about the role of TxA₂ in the immune system have been limited. TxA₂ plays a key role in neutrophil-mediated T-cell immune responses to infections and vaccinations.³⁶  TxA₂-TP signaling also negatively regulates dendritic cell–T cell interaction and modulates acquired immunity.³⁷  Interaction between TxA₂ with its receptor in thymic microenvironment may be important in the apoptosis of activated T cells.³⁸  The role of the TxA₂-TP axis in B-cell biology remains poorly understood. We found that TP-deficient mice display disturbed early B-cell development and that administration of a TP agonist or antagonist could modulate B-cell differentiation. These observations demonstrated that TxA₂-TP signaling plays essential roles in early B-cell development.

The sources of TxA₂ production and the manner in which TxA₂ exerts its role on developing B cells in vivo is an interesting question. A variety of cell types in the BM microenvironment could produce TxA₂, including dendritic cells, T cells, macrophages, and stromal cells.^37-39  Deletion of COX-1 selectively impairs the production of TxA₂ in B cells but not in T cells (data not shown), supporting the possibility that TxA₂ functions in an autocrine manner in developing B cells. In vitro culture of purified B cells using a stromal-free system would help to clarify whether B cells produce TxA₂ and activate the signaling axis we observed. In contrast with a previous report that B cells from mouse spleen express very low levels of TP,³⁷  we found that TP expression in B cells from mouse BM was abundant and much higher than those from the spleen (data not shown); more interestingly, TP expression was dynamic during the course of B-cell development. These observations raise the possibility that some stimuli within the BM may induce the transcription of genes essential for TxA₂ signaling in B cells, which facilitate B-cell development.

Our investigations showed that PKA plays a dominant role in the regulation of the JAK/STAT5 signal, in the presence of IL-7 stimulation, with JAK3 serving as the substrate of PKA in this axis. It has been reported that the cAMP-PKA pathway could negatively regulate JAK3/STAT5 activation in response to IL-2 stimulation in T lymphocytes, and functionally, inactivation of JAK3 by PKA is the mechanism.^40,41  Our study demonstrated that PKA positively regulates JAK/STAT5 signaling in developing B cells, and PKA can activate JAK3 through tyrosine phosphorylation. It is therefore plausible that distinct cell types use different mechanisms in regulating the same signaling molecule, which may be context dependent.

The immunoregulatory effects of aspirin are gaining increasing attention.^29,36  Aspirin is generally considered to have immune-suppressive capabilities.⁴²  Clinical administration of aspirin has beneficial effects on certain autoimmune diseases,⁴³  as well as allograft rejection.⁴⁴  Aspirin and other nonsteroidal anti-inflammatory drugs were reported to suppress the humoral immune response by impairing antibody synthesis.^45,46  The mechanisms underlying aspirin’s immunomodulatory effects, however, deserve further exploration. Our study indicates that low-dose aspirin may suppress early B-cell development, providing an alternative explanation for the beneficial effects of aspirin in some autoimmune diseases. Given that low-dose aspirin has been routinely prescribed to prevent myocardial infarction and stroke,²⁹  our work raises concerns about the potential adverse effect of this maneuver on B-cell development.

The authors thank Dr Toshio Kitamura (University of Tokyo, Tokyo, Japan) for providing the constitutively activated STAT5 construct (STAT5A1*6).

This work was supported by the following grants to J.Z.: National Key Basic Research Program of China grant 2012CB524900; Guangdong Innovative Research Team Program grant 2009010058; Guangdong Province Universities and Colleges Pearl River Scholar Funded Scheme grant GDUPS, 2014; Key Research Projects of National 12th Five-year Plan for the Prevention and Treatment of Major Infectious Diseases grant 2012ZX10001003; National Natural Science Foundation of China grants 81072397 and 31270921; Natural Science Foundation of Guangdong grant S2011020006072; and Fundamental Research Funds for the Central Universities, 111 Project grant B12003. Y.Y. was supported by the following grants: Ministry of Science and Technology of China grants 2012CB945100, 2011CB503906, and 2011ZX09307-302-01; National Natural Science Foundation of China grants 81030004 and 31200860; National Natural Science Foundation of China–Canadian Institutes of Health Research joint grant 81161120538; and One Hundred Talents Program of the Chinese Academy of Sciences grant 2010OHTP10. Y.Y. is a Fellow at the Jiangsu Collaborative Innovation Center for Cardiovascular Disease Translational Medicine.

Contribution: Q.Y. carried out the experiments, analyzed the data, and participated in figure organization and manuscript writing; M.S. and S.Z. performed the bone marrow transplantation experiment; Y.C. assisted with the low-dose aspirin experiment; Y.S. helped with mouse strain maintaining; C.Z. performed mass spectrometry assay; H.Z. and D.I.G. provided helpful suggestions in the experimental design and manuscript writing; Y.Y. provided all the mouse strains, participated in experimental design, and edited the manuscript; and J.Z. designed the overall study, supervised all the experiments, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jie Zhou, Institute of Human Virology, Zhongshan School of Medicine, Sun Yat-Sen University, 74 Zhongshan 2nd Rd, Guangzhou 510080, China; e-mail: zhouj72@mail.sysu.edu.cn; or Ying Yu, Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, 294 Taiyuan Rd, Shanghai 200031, China; e-mail: yuying@sibs.ac.cn.