Hemodynamic shock and multiple organ failure are hallmarks of the systemic inflammatory response syndrome (SIRS), which may be caused by bacterial infections (Cohen, 2002). Lipopolysaccharide (LPS), which is released from the outer membrane of gram-negative bacteria lysed by immune cells (Raetz and Whitfield, 2002), plays a key role in the pathophysiology of endotoxic sepsis (Beutler and Rietschel, 2003). By interacting with toll-like receptor 4 (TLR4), LPS stimulates the production of tumor necrosis factor-α (TNF-α) (Carswell et al., 1975; Hoshino et al., 1999) and interferon-γ (IFN-γ, which in turn induce the enzyme indoleamine 2,3-dioxygenase (IDO) in a variety of cells, including immune and endothelial cells (Okuda et al., 1996; Wei et al., 2000; Mizutani et al., 2002; Jung et al., 2009; Wang et al., 2010). Tryptophan is metabolized by IDO or tryptophan 2,3-dioxygenase (TDO) into formylkynurenine, which is spontaneously converted into L-kynurenine (KYN). KYN in transaminated by kynurenine aminotransferases (KATs) into kynurenic acid (KYNA), a neuroactive compound that interacts with NMDA receptors and α₇ nicotinic receptors. Alternatively, KYN is converted by kynurenine monooxygenase (KMO) into 3-hydroxykynurenine (3-HK), which can be transaminated into xanthurenic acid (XA) by KATs or, alternatively, metabolized into 3-hydroxyanthranylic acid (3-HANA) by kynureninase. 3-HANA is transformed into quinolinic acid (QUINA), which is further metabolized to produce nicotinamide (Figure 1).

A growing body of evidence suggests that metabolites generated within the kynurenine pathway are involved in the regulation of blood pressure. Polymorphic variants of the gene encoding for kynureninase are associated with changes in systolic blood pressure (SBP), and the transcript of kynureninase is about threefold higher in spontaneously hypertensive rats than in normotensive control rats (Mizutani et al., 2002). Regulation of vascular tone by kynurenine metabolites may become prominent during inflammation, i.e., under conditions in which the kynurenine pathway is strongly induced (Wang et al., 2010). Accordingly, pharmacological blockade of IDO protects mice against LPS-induced shock (Jung et al., 2009), and KYN has been shown to act as an endothelium-derived relaxing factor causing hypotension in mice treated with LPS or infected by Plasmodium berghei (Wang et al., 2010). Because the action of KYN on blood vessels was not mimicked by KYNA, 3-HK, 3-HANA or QUINA, it was concluded that KYN itself, rather than any of its metabolites, causes vascular relaxation under inflammatory conditions (Wang et al., 2010). However, the analysis of kynurenine metabolites did not include XA, which has long been considered as a mere detoxification product of 3-HK (Okuda et al., 1996; Wei et al., 2000). The recent demonstration that XA is biologically active, interacts with a not-yet identified G-protein-coupled receptor, and its plasma levels are largely reduced in patients affected by schizophrenia (Fazio et al., 2015) fueled interest in examining the role of XA in physiology and pathology. To our knowledge, the cardiovascular effects of XA have never been investigated with the exception of an old study showing that XA and other kynurenine metabolites cause bradycardia in isolated frog hearts (Rudzit et al., 1986).

We now report that XA displays a greater potency and efficacy than KYN in causing relaxation of a mouse resistance artery, and that its action required the presence of the endothelium and was mediated, at least in part, by an enhanced production of nitric oxide (NO·) and endothelium-derived hyperpolarizing factor (EDHF). More important, the hypothensive effect of LPS was abrogated by systemic treatment with selective inhibitors of KMO, the enzyme that converts KYN into the XA precursor, 3-HK.

The kynurenine pathway is activated in response to pro-inflammatory cytokines in various cell types, including endothelial cells (Wang et al., 2010). Wang and colleagues have shown that activation of IDO was required for inflammation-induced vasodilation, and that KYN was the only tested metabolite of the pathway that caused arterial relaxation. KYN was active in vitro at relatively high concentrations (≥300 μM), but large amounts of KYN can be produced in endothelial cells under inflammatory conditions (Wang et al., 2010). KYN acted directly on smooth muscle cells by enhancing both cAMP and cGMP formation, and its action persisted in the absence of the endothelium (Wang et al., 2010). Our data offer a complementary scenario suggesting that XA is a new player in mechanisms of inflammation-induced hypotension. As opposed to KYN (Wang et al., 2010), XA cannot be considered as a putative endothelium-derived relaxing factor, because the action of exogenous XA required the presence of an intact endothelium. In addition, exogenous XA enhanced activating phosphorylation of eNOS and its relaxing activity was blunted by pharmacological inhibition of NOS, and by apamine and Ctx, which block the action of EDHF (Doughty et al., 1999) by inhibiting SK Ca²⁺-activated K⁺ channels and BK, IK, and some voltage-dependent potassium channels, respectively. Because the vasorelaxing action of XA was not entirely blocked by apamin and Ctx, it cannot be excluded that additional types of potassium channels are involved. Of note, the vasorelaxing action of L-kynurenine appears to be mediated by activation of Kv7 voltage-sensitive potassium channels (Sakakibara et al., 2015), although, as opposed to XA, the action of L-kynurenine on blood vessels does not require the integrity of the endothelium (Wang et al., 2010; Sakakibara et al., 2015). It is possible that XA and KYN cause arterial relaxation acting on different types of vessels and via different mechanisms. Accordingly, KYN caused relaxation of porcine coronary arteries, aortic rings, and mouse abdominal aorta (Wang et al., 2010), whereas XA was more efficacious than KYN in relaxing mouse mesenteric artery, but it failed to relax mouse aorta. The selectivity of XA for mesenteric artery with respect to aortic rings might reflect the different endothelium-dependent relaxing mechanisms in the two vessels. For example, endothelium-dependent aortic relaxation is mainly mediated by NO production, whereas endothelium-dependent mesenteric artery relaxation involves additional mechanisms, such as prostanoids and EDHF (Runnie et al., 2004). The vascular bed-specificity of endothelium-mediated mechanisms of relaxation/contraction is further supported by studies examining the response of different types of vessels, including aorta and mesenteric arteries, to mineralocorticoids (Mueller et al., 2015) or insulin/insulin-like growth factor-I (Wu et al., 1994).

How precisely XA acts to activate eNOS and the EDHF in endothelial cells remains to be determined. Experiments with tetrodotoxin excluded that the action of XA was mediated by the release of a vasoactive molecule from perivascular nerve endings. Because the vasorelaxing action of XA was concentration-dependent and a saturating effect was visible at concentrations >2 mM, it is likely that XA binds to a specific and saturable recognition site localized on endothelial membranes. The identity of the endothelial XA receptor is unknown. Arterial relaxation was not mediated by mGlu2 metabotropic glutamate receptors, which are involved in some behavioral effects of XA (Fazio et al., 2015). XA binds to specific and saturable recognition sites and enhances GTP-γ-S binding in brain membranes (Taleb et al., 2012). The elucidation of the molecular mechanism underlying XA-induced arterial relaxation awaits the identification of the putative G-protein coupled receptor activated by XA. The evidence that XA was able to cause arterial relaxation raised the possibility that XA could lower blood pressure in living animals and mediate, at least in part, SIRS-induced hypotension. Exogenous administration of XA (100 mg/kg) significantly reduced blood pressure at 2–8 h, when the corresponding serum XA levels were higher than 100 ng/ml (corresponding to 487 nM). Blood pressure returned back to normal at 24 h, when serum XA levels fell below 50 ng/ml (Figures 4A,B).

Lipopolysaccharide injection in mice caused the expected drop in blood pressure, which peaked at 24 h and was no longer observed at 48 h. Several mechanisms have implicated in LPS-induced hypotension, involving proinflammatory cytokines, inducible NOS (iNOS), integrin beta-2, superoxide anions, and prostaglandins (Michie et al., 1988; Mukaida et al., 1996; Obermeier et al., 1999; Liu et al., 2003). IDO is expressed in the endothelial cells of experimental animals and humans (Wang et al., 2010; Changsirivathanathamrong et al., 2011), and IDO activity in resistance vessels correlates with the extent of hypotension in human septic shock (Changsirivathanathamrong et al., 2011). That activation of the kynurenine pathway plays a key role in septic shock is suggested by the evidence that inhibition of IDO with 1-methyl-tryptophan prevents the drop in blood pressure in mice injected with LPS (Wang et al., 2010). In mice challenged with LPS, levels of both KYN and XA increased by several fold after 24 h, the time that corresponds to the peak of LPS-induced hypotension. Levels of both compounds continued to increase in the following 24 h, when levels of blood pressure turned back to normal. This suggests that homeostatic mechanisms that control blood pressure overcome the hypotensive effect of kynurenines at 48 h following LPS injection. Interestingly, a substantial reduction of blood pressure was observed 4 h following LPS injection, when KYN levels were unchanged, and XA levels showed a trend to an increase but remained lower than 50 ng/ml. These relatively low levels of XA might reach the threshold to cause hypotension in the presence of other vasoactive molecules produced in response to LPS (compare the two experiments described in Figures 4A–D). Although serum concentrations of XA in mice challenged with LPS (about 0.5–0.75 μM at 24–48 h) were >20-fold lower than the minimal vasorelaxing concentrations in vitro, it is likely that much greater concentrations of XA are reached in blood vessels because of a local synthesis in the endothelium, where the kynurenine pathway is activated by IDO.

Perhaps the most relevant finding of our study was that treatment of mice with the KMO inhibitor, Ro-618048 (Moroni et al., 2005), abrogated the effect of LPS on blood pressure although it caused large increase in serum KYN levels. This evidence strongly suggests that, at least in our model, KYN metabolism downstream of KMO is a critical event in the pathogenesis of LPS-induced hypotension. It is unlikely that Ro-618048 had off-target effects on the regulation of blood pressure because the drug was inactive in the absence of LPS, i.e., when the kynurenine pathway was not activated. In principle, any kynurenine metabolite lying downstream of KMO could have contributed to the hypotensive effect of LPS. Knowing that 3-HK, 3-HANA and QUINA cause no arterial relaxation in vitro (Wang et al., 2010), it is likely that Ro-618048 prevented LPS-induced hypotension by reducing the synthesis of XA from the KYN metabolite, 3-HK. Of note, however, pretreatment with Ro-618048 reduced, but did not abolish, the increase in serum XA levels induced by LPS. This leaves opened the possibility that other kynurenine metabolites that lie downstream of KMO (perhaps different from 3-HK, 3-HANA and QUINA) are involved in inflammation-induced hypotension. This hypothesis warrants further investigation. We could not use KAT inhibitors (Jayawickrama et al., 2015) to directly inhibit XA synthesis from 3-HK because KAT also canalizes the formation of KYNA from KYN (Jayawickrama et al., 2015; see also Figure 1). The two expected effects of KAT inhibitors (shunt of the pathway toward the metabolism of KYN into of 3-HK and inhibition of 3-HK transamination into XA), and perhaps other off-target effects, may confound the study of LPS-induced hypotension.

Our findings demonstrate that XA selectively relaxes resistance vessels, and suggest that induction of XA synthesis is critically involved in the pathogenesis of endotoxin-induced hypotension. This is highly novel in the kynurenine field because, so far, the study of cardiovascular effects has been limited to “conventional” kynurenines, such as KYN, KYNA, 3-HK, 3-HANA, and QUINA. Our findings encourage the use of KMO inhibitors, which limit the production of the XA precursor, 3-HK, in the treatment of endotoxic shock. Again, the potential use of KAT inhibitors is limited by the multiple action of KATs within the kynurenine pathway. The design of drugs that selectively inhibit transamination of 3-HK into XA with limited effects on transamination of KYN into KYNA (a very difficult task) may offer new powerful instruments for the study of the biological activity of XA. If ever available, these drugs are expected to be highly effective in the treatment of endotoxic shock without the potential adverse effects of IDO and KMO inhibitors, which are upstream inhibitors of the kynurenine pathway and may limit the homeostatic actions of kynurenines in the CNS and immune system (Schwarcz et al., 2012).

All the authors have contributed significantly to the submitted work. FF, AC, LL, AD, LC, MA, and MM designed and performed experiments, analyzed data and wrote the manuscript. GB, VB, MS, FN, and CV designed the experiments and analyzed data. CV and FN made critical revision of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.