Selectivity and specificity of sphingosine-1-phosphate receptor ligands: caveats and critical thinking in characterizing receptor-mediated effects.

Receptors for sphingosine-1-phosphate (S1P) have been identified only recently. Their medicinal chemistry is therefore still in its infancy, and few selective agonists or antagonists are available. Furthermore, the selectivity of S1P receptor agonists or antagonists is not well established. JTE-013 and BML-241 (also known as CAY10444), used extensively as specific S1P(2) and S1P(3) receptors antagonists respectively, are cases in point. When analyzing S1P-induced vasoconstriction in mouse basilar artery, we observed that JTE-013 inhibited not only the effect of S1P, but also the effect of U46619, endothelin-1 or high KCl; JTE-013 strongly inhibited responses to S1P in S1P(2) receptor knockout mice. Similarly, BML-241 has been shown to inhibit increases in intracellular Ca(2+) concentration via P(2) receptor or α(1A)-adrenoceptor stimulation and α(1A)-adrenoceptor-mediated contraction of rat mesenteric artery, while it did not affect S1P(3)-mediated decrease of forskolin-induced cyclic AMP accumulation. Another putative S1P(1/3) receptor antagonist, VPC23019, does not inhibit S1P(3)-mediated vasoconstriction. With these examples in mind, we discuss caveats about relying on available pharmacological tools to characterize receptor subtypes.

Receptors for sphingosine-1-phosphate (S1P) were discovered and characterized in the late nineties. Ten years ago, there was no selective S1P receptor agonist or antagonist. Suramin, an old anti-protozoal drug, was used as S1P3 receptor antagonist (Ancellin and Hla, 1999; Salomone et al., 2003), but its usefulness was limited because it lacked specificity (Voogd et al., 1993). Recently, a number of ligands for S1P receptors have been screened as agonists and antagonist (Im, 2010). Most of these newly developed agents are commercially available and are increasingly being used to characterize the S1P receptor subtypes involved in specific biological mechanisms and functions. However, they have often not been screened for specificity against a wide array of targets, nor have they been systematically studied in vivo, as is required for drugs intended for human use.

In contrast, more detailed pharmacological data is available for compounds that show significant therapeutic potential. Among these, particularly worthy of mention is fingolimod (FTY720), an FDA- and EMA-approved drug, that, once phosphorylated, binds all S1P receptors but S1P2 in vitro. In vivo, phospho-fingolimod down-modulates lymphocytic S1P1 to inhibit S1P-S1P1-signaling and lymphocyte egress (Matloubian et al., 2004; Martin et al., 2010), which induces remission in patients with multiple sclerosis (Kappos et al., 2010). Fingolimod has been screened against a large number of other targets (including about 100 receptors) to insure S1P receptor specificity (V. Brinkmann, Novartis Institutes for BioMedical Research, personal communication).

At variance with most other available S1P receptor agents, FTY720 has been used in a large number of publications (nearly 1000 PubMed-indexed publications as of January 2011). Possibly because it has been the object of such an intense scrutiny, other, non-S1P receptor-mediated effects of FTY720 have been documented. For instance, FTY720 has been shown to inhibit ceramide synthase (Lahiri et al., 2009), cytosolic phospholipase A2 (Payne et al., 2007), S1P lyase (Bandhuvula et al., 2005), sphingosine kinase (SPK) 1 (Vessey et al., 2007; Tonelli et al., 2010), and acid sphingomyelinase (Dawson and Qin, 2011), stimulate 27-hydroxycholesterol production (Blom et al., 2010), bind to 14-3-3 proteins and inhibit their pro-survival signaling (Woodcock et al., 2010). Because most of these effects were observed at relatively high FTY720 concentration, they may not be relevant to the in vivo effects of the drug. But investigators should be cautious when interpreting the results of in vitro studies, particularly when these studies use high FTY720 concentrations. Incidentally, the fact that fingolimod does not discriminate between S1P1, S1P3, S1P4, and S1P5 receptors may not matter in vivo. Because blood levels of FTY720 are low (<3 nM), the drug does not occupy S1P3 receptors to a significant extent. However, after down-regulation of S1P1 receptors, homeostatic activation of S1P2 and S1P3 by endogenous plasma S1P (>200 nM) becomes predominant, and the S1P signaling balance is shifted from S1P1 to S1P2/S1P3, and thus from Gi to G12/13/Rho/Rho kinase activation. The observed functional consequence of FTY720 administration may therefore be a sum of drug- and endogenous S1P-mediated effects, and be very similar with S1P1 selective agents and non-subtype selective S1P modulators.

We chose to focus this Perspective on JTE-013 and BML-241 because they are commonly used despite reports of their lack of specificity. We also discuss VPC23019, for which evidence of S1P3 receptor antagonism activity seems weak. JTE-013 was developed by Central Pharmaceutical Research Institute, Japan Tobacco Inc. Its patent (Patent WO 01/98301; December 27, 2001) stated that JTE-013 inhibited the specific binding of radiolabeled S1P to membranes of CHO cells transfected with human and rat S1P2 receptors, with IC50 values of 17 ± 6 and 22 ± 9 nM, and did not affect S1P binding to S1P3 and S1P1, at concentrations up to 10 μM (Osada et al., 2002; Ohmori et al., 2003). Based on these data, JTE-013 has been considered since then a specific S1P2 receptor ligand and, following further experimental data, a S1P2 antagonist. Because JTE-013 had been used to characterize the S1P receptor mediating canine coronary artery contraction (Ohmori et al., 2003), we used it in rodent isolated vessels to assess the role of S1P2 receptors in S1P-induced vasoconstriction of basilar arteries (Salomone et al., 2008). We knew that S1P-induced vasoconstriction was abolished in mice and were therefore surprised to see that JTE-013 inhibited vasoconstriction to S1P in arteries from wild type mice, because this suggested that S1P2 instead of, or in addition to, S1P3 was mediating S1P effect. Investigating further, however, we realized that JTE-013 inhibited vasoconstriction not only to S1P, but also to the prostanoid analog U46619, endothelin-1 and high KCl (Salomone et al., 2008; of note, vasoconstriction induced by high KCl is not receptor-mediated but related to L-type Ca2+ channels). We then performed a critical genetic control experiment, and found that JTE-013 inhibited S1P-induced vasoconstriction in mice, demonstrating that this effect was not related to S1P2 receptors. Yet, JTE-013 is still widely used to characterize S1P2-dependent effects (Table 1). Of note, most studies that did validate S1P2-blocking effects of JTE-013 used concentrations equal to or lower than 1 μM, while we found evidence of non-specificity at 10 μM. It is therefore possible that JTE-013 retains sufficient selectivity and remains a useful antagonist in the submicromolar range, but S1P2 receptor involvement should be validated by other means.

Table 1

Some published data obtained with JTE-013 (among more than 50 PubMed-indexed citations).

System	Effect	Concentration	Inferred significance	Validation	Conflicting data/caveat	Reference
Smooth muscle cells/HUVEC	Cell migration	Up to 10 μM	S1P inhibits cell migration via S1P2	None		Osada et al. (2002)
Hepatocytes	DNA synthesis	Up to 10 μM	S1P inhibits hepatocyte proliferation via S1P2	None		Ikeda et al. (2003)
Coronary artery smooth muscle cells	Contraction	Up to 10 μM	S1P contracts coronary muscle via S1P2	None		Ohmori et al. (2003)
Transfected CHO cells	Ca2+ mobilization	Up to 1 μM	S1P2 negatively regulates cell motility	S1P2 overexpression		Arikawa et al. (2003)
Melanoma B16 cells	ERK activation Cell migration
Human coronary smooth muscle cells	cAMP production	Up to 10 μM	S1P2 induces cAMP production through PGI2	siRNA-mediated S1P2 knockdown		Damirin et al. (2005)
Mouse spiral modiolar artery	Contraction	Up to 3 μM	S1P2 receptors mediate modiolar artery contraction			Kono et al. (2007)
HUVEC	Permeability	0.2 μM	S1P2 increases vascular permeability	S1P2 overexpression		Sanchez et al. (2007)
Rabbit bladder smooth muscle	Contraction	1 μM	S1P contracts bladder smooth muscle via S1P2		Phospho-FTY720 also contracted	Watterson et al. (2007)
U373 glioblastoma cells	uPAR and PAI-1 mRNA expression	1 μM	S1P induces uPAR and PAI-1 mRNA expression via S1P2	siRNA-induced S1P2 knockdown		Bryan et al. (2008)
WiT46 cells	CTGF expression	1 μM	S1P induces CTGF expression via S1P2	S1P2 overexpression		Li et al. (2008a)
Rat vascular smooth muscle cells	Cell migration and Rac activation	1 μM	S1P inhibits cell migration via S1P2	S1P2 overexpression		Takashima et al. (2008)
Rat/mouse basilar artery	Contraction	10 μM	S1P3 mediate basilar artery constriction JTE-013 is not specific for S1P2	S1P2 and S1P3 knockout mice		Salomone et al. (2008)
Rat vascular smooth muscle cells	Proliferation	1 μM	S1P inhibits cell proliferation via S1P2			Wamhoff et al. (2008)
Hamster resistance arteries	Contraction	1 μM	S1P induces contraction via S1P2	S1P2 knockdown by antisense oligonucleotide	No proof of S1P2 mRNA or protein decrease	Peter et al. (2008)
Mouse neural progenitors	Cell migration	0.25 nmol/h, icv	S1P inhibits cell migration via S1P2	shRNA-mediated S1P2 knockdown		Kimura et al. (2008)
Human mast cells	Cytokine/chemokine secretion	100 nM	S1P degranulates mast cell and modulates anaphylaxis via S1P2	siRNA-induced S1P2 knockdown		Oskeritzian et al. (2010)
Mouse	Passive anaphylaxis	20 μg/mouse		S1P2 knockout mice
Hamster resistance artery	Myogenic tone	1 μM	S1P induces myogenic tone via S1P2			Lidington et al. (2009)
Isolated perfused mouse lung	Vasoconstriction	10 μM	S1P modulates pulmonary vascular tone in vivo via S1P2	S1P2 knockout mice		Szczepaniak et al. (2010)
Mouse	Proatherosclerotic cytokine release	1.2 mg/kg	S1P2 receptors regulate macrophage retention and inflammatory cytokine secretion	S1P2 knockout mice		Skoura et al. (2011)

BML-241 was developed at the same time as JTE-013, by rational drug design using the structure of S1P to interrogate a three-dimensional database. Two novel compounds were identified that showed antagonist activity. When tested at 10 μM, “Compound 2” inhibited by 37% S1P-induced increases in [Ca2+]i in HeLa cells expressing S1P3 receptors and by about 7% [Ca2+]i increases in cells expressing S1P1 receptors (Koide et al., 2002). Strikingly, this study was based on measurements with a single BML-241 concentration and the comparison of only two potential targets (S1P1 and S1P3 receptors) in one assay; furthermore, an inhibition by less than 40% by a 10-μM antagonist concentration in cell culture might be considered less than impressive. Nevertheless, at least 20 studies (Table 2) were published using BML-241 (also known as CAY10444), most of them after the publication of an article showing that BML-241 inhibits [Ca2+]i increases via purinergic P2 receptor or α1A-adrenoceptor stimulation and α1A-adrenoceptor-mediated contraction, while not affecting the S1P3-mediated decrease of forskolin-induced cAMP accumulation (Jongsma et al., 2006). In a third of these studies, BML-241/CAY10444, tested at 1 or 10 μM, had no effect, leading the authors to conclude on a lack of involvement of S1P3 receptors in their paradigm. It is however possible that the BML-241/CAY10444 concentration used in these studies was too low to block S1P3 receptors to a significant extent; indeed, using a β-arrestin recruitment assay, Wetter et al. (2009) showed that their S1P3 cell line response to S1P was inhibited to 78% of the receptor response by 100 μM BML-241/CAY10444, with an IC50 of 4.6 μM. A less likely, but possible explanation for negative studies with BML-241 is that a non-specific action of the drug may have counteracted the partial blockade of S1P3 receptors. In light of the low affinity of BML-241/CAY10444 for S1P3 receptors, it does not appear that the few studies that documented an effect of the antagonist at 1 or 5 μM were justified in implicating S1P3 receptors. Two studies that used higher BML-241/CAY10444 concentrations (50 and 100 μM) did observe an antagonist effect. In one of these studies, the S1P1 agonist SEW2871 had no effect, while the S1P1/S1P3 receptor antagonist VPC23019 reproduced the effect of BML-241/CAY10444 (Lichte et al., 2008). In the other study, B-cell migration was promoted by a mixed S1P1/S1P3 receptor agonist (VPC24191), but not by the S1P1 agonist SEW2871. Furthermore, B-cells from mice migrated to S1P, whereas cells from S1P3 knockout mice were unable to migrate to S1P at all concentrations of S1P tested (Donovan et al., 2010). Interestingly, in this study B-cell migration was only slightly, and not significantly, inhibited by BML-241/CAY10444, tested at concentration as high as 100 μM.

Table 2

Published data obtained with BML-241/CAY10444.

System	Effect	Concentration	Inferred significance	Validation	Conflicting data/caveat	Reference
Transfected HeLa cells	[Ca2+]i increase	10 μM	Blocks S1P3 receptors	No effect on S1P1-Hela		Koide et al. (2002)
Endothelial cells	Expression of adhesion molecules	1–10 μM	S1P3 receptors modulate adhesion molecule expression	Antisense oligonucleotides		Kimura et al. (2006)
Pancreatic Islet β Cells	Cell survival	10 μM	S1P3 receptors mediate protective effects on β-cells against cytokine-induced apoptosis.	None		Laychock et al. (2006)
B lymphocytes	Rap1 activation	10 μM	S1P3 receptors activate Rap1 and might promote B-cell adhesion and migration	None		Durand et al. (2006)
Transfected CHO cells	[Ca2+]i increase	10 μM	BML-241 is not selective			Jongsma et al. (2006)
Astrocytes	Cell migration	10 μM	No role of S1P3	none		Sato et al. (2007)
Fibroblast-like synoviocytes	Cell migration Cytokine/chemokine secretion Cell survival	5 μM	S1P stimulates FLS migration through S1P1 and S1P3, induces cytokine/chemokine secretion through S1P2 and S1P3, and protects from cell apoptosis via S1P1.	None		Zhao et al. (2008)
Multiple myeloma cells	Cell survival	10 μM	S1P3 receptors mediate S1P-induced STAT3 phosphorylation and Mcl-1 upregulation	None		Li et al. (2008b)
Keratinocytes	[Ca2+]i increase	50 μM	S1P3 receptors mediate [Ca2+]i increase	No effect of SEW2871 (S1P1 agonist), inhibition by VPC23019	Specificity of SEW2871 and VPC23019 non-fully characterized	Lichte et al. (2008)
Transfected HEK-293 cells	[Ca2+]i increase	50 μM	S1P3 receptors mediate [Ca2+]i increase	No effect of SEW2871 (S1P1 agonist), inhibition by VPC23019	Specificity of SEW2871 and VPC23019 non-fully characterized	Lichte et al. (2008)
Mouse cardiomyocytes	Erk activation	0.1–10 μM	No role of S1P3	none		Tao et al. (2009)
B35 neuroblastoma cells	Cell migration	10 μM	No role of S1P3	No effect with other S1P3 antagonists (VPC23019, VPC25239)		Hans et al. (2009)
Embryonic stem cells	Erk activation	1–10 μM	S1P5, not S1P3, activate Erk1/2	None		Rodgers et al. (2009)
Tango EDG3 cell lines	GPCR β-arrestin recruitment and deactivation	1 nM–0.1 mM	The EDG3 response to S1P is inhibited by CAY-10444 with an IC50 of 4.6 μM	None		Wetter et al. (2009)
Bronchial smooth muscle	Contractility	10 μM	No role of S1P3	none		Chiba et al. (2010)
Granulosa lutein cells	Cell migration	1 μM	S1P3 receptors are involved in FF-HDL- and S1P-stimulated granulosa cell migration	Effect of suramin	Suramin is non-specific	Becker et al. (2010)
Mouse cardiomyocytes	Cell viability	10 μM	S1P3 receptors play a role in Akt activation and cardiomyocyte viability	none		Tao et al. (2010)
Coronary artery	Vasorelaxation	10 μM	S1P3 receptors mediate endothelium dependent coronary relaxation	No effect of W146 (S1P1 antagonist), but inhibition by VPC23019 (mixed S1P1/S1P3 antagonist)	Specificity of VPC23019 not fully characterized	Mair et al. (2010)
Immature B lymphocytes	Cell migration	100 μM	S1P3 receptors are responsible for immature B-cell chemotaxis to S1P	S1P3−/− immature B-cells unable to migrate to S1P		Donovan et al. (2010)
Estrogen receptor-positive breast cancer cells	Cell migration	10 μM	S1P3 stimulates the accumulation of phosphor-ERK-1/2 and promotes cell migration	siRNA knockdown		Long et al. (2010)
Ovarian cancer cells	Cell proliferation and survival	1 μM	S1P3 receptors do not play a role in cell proliferation and survival	None		Illuzzi et al. (2010)
Mesenteric venules	Microvessel permeability	10 μM	S1P3 receptors do not play a role in the protection by S1P of PAF-induced permeability	None		Zhang et al. (2010)

It is increasingly recognized that S1P, generated intracellularly by SPK, can be released into the extracellular space and thereby stimulate membrane S1P receptors, establishing an autocrine loop (Kim et al., 2009; Takabe et al., 2010). When this occurs, even a specific S1P receptor antagonist might inhibit the response to an agonist other than S1P. Autocrine loops might account for the fact that JTE-013 and BML-241 inhibit responses to U46619, endothelin-1 and high KCl, and to purinergic P2 receptor or α1A-adrenoceptor stimulation, respectively. However, autocrine loops cannot explain why an agonist such as JTE-013 has any action in mice lacking S1P2 receptors. Potential S1P autocrine loops should be taken into account when screening S1P receptor antagonists in complex systems (cell or organism level), particularly when both SPK inhibitors and S1P receptor blockade (pharmacologically or by preventing receptor expression) attenuate the response (Peter et al., 2008; Schnitzer et al., 2009; Salomone et al., 2010).

VPC23019 was initially described as an S1P1/3 receptor antagonist, with pKB values of 7.5 and 6.0 for the S1P1 and S1P3 receptors, respectively (Davis et al., 2005). Since then it has been mainly used for characterizing S1P1-receptor mediated responses. In a few studies, however, it has been also used as an S1P3 receptor antagonist. For example, as mentioned above, Lichte et al. (2008), reported that S1P-induced calcium signaling in human keratinocytes is mainly mediated by S1P3, because it can be blocked by the putative S1P3 antagonists BML-241, and VPC23019. Recently, Jongsma et al. (2009), using three different assays, have shown that several compounds of the VPC series, including VPC23019, behave as full or partial agonists at S1P3 receptors. Although obtained in vitro, in a system expressing high S1P3 receptor density, these data suggest that VPC23019 is a less than ideal tool to characterize S1P3-mediated responses. When studying the effects of S1P on vascular tone, we found that VPC23019 potentiated S1P-induced contractile response in both rat and mouse basilar arteries with intact endothelium, while it failed to do so in preparations without endothelium (Salomone et al., 2008). We interpreted this finding as a result of S1P1 antagonism exerted by VPC23019, because S1P1 receptors located on vascular endothelium are known to stimulate nitric oxide release and induce vasodilatation (and may therefore counteract vasoconstriction induced by S1P through S1P3 receptors). However, we were surprised not to observe inhibition of S1P-induced vasoconstriction by VPC23019, at concentrations as high as 10 μM, when solid evidence (S1P3 knockout mice) indicated that S1P-induced vasoconstriction was mediated by S1P3 receptors. Indeed, Murakami et al. (2010) showed that the novel potent and selective S1P3 antagonist TY-52156 restores coronary blood flow reduced by S1P, but VPC23019 was inactive in this system. Interestingly, the same study also showed that in isolated coronary smooth muscle cells, TY-52156 inhibited both Rho activation and Ca2+ signal, whereas VPC23019 only inhibited Ca2+ signal. Taken together with our data, this study suggests that results obtained with VPC23019 should be interpreted with caution.

While we are emphasizing in this Perspective issues related to three particular receptor antagonists, our purpose is to make a broader methodological point on the distinction between selectivity and specificity. The term “selectivity” should refer to the ability of a drug to discriminate between related targets (e.g., receptors or enzymes), showing a higher binding affinity for one subtype or isoform. Selectivity should be assessed by screening in pure systems (e.g., cell lines transfected with one receptor subtype at a time) and eventually in complex systems, including in vivo wild type and genetically altered models. The term “specificity” should refer not only to the ability of a drug to identify a receptor of interest, but also to its potential for discriminating between negative and positive interactions; i.e., the drug should bind the receptor with appropriate affinity, the drug should have low/no cross-reactivity with other receptors. In order for “specificity” to be assessed, the drug should therefore be screened for its interaction with the target of interest and for its interaction with as many unrelated biological targets as possible. For receptors, ligand binding studies may be considered adequate to quantitatively assess affinity for the binding sites under examination, but require further assessment of function, in the presence and absence of the reference agonist, to categorize a given ligand as agonist or antagonist.

Until the end of 1980s, pharmacological studies were mainly based on the use of agonists and antagonists (Salomone, 2010). More recently, investigators have also manipulated the expression levels of receptors using molecular techniques and genetic alterations. This approach represents, under critical evaluation, the most specific pharmacologic strategy available today. When using a drug as an agonist or antagonist, one has to consider the probability of off-target effects; instead, when using a receptor knock out animal, one can exclude, with great degree of confidence, that an effect is mediated by the receptor product of the deleted gene. We therefore believe that such a molecular genetic control, whenever available, should be considered the best validation in analytical and experimental pharmacology. Unfortunately, investigational pharmacological agents, agonists or antagonists, are increasingly being used to come to pharmacologic conclusions on receptor function, even though data on their selectivity is seldom sufficient and little information on their specificity is available. As this Perspective demonstrates, these conclusions are sometimes drawn despite conflicting evidence (see shaded rows in Tables 1 and 2), such as that coming from gene-deletion studies.

In conclusion, experimental ligands not used in humans (for which extensive pre-clinical characterization has uncovered potential off-target actions) and for which only limited information from simple systems (in vitro, transfected cell lines) is available should be used with caution and pharmacological data obtained with them should be considered as weak evidence, unless supported by consistent stronger evidence (for example genetic knockdown or knockout data, or concurrent pharmacological evidence obtained with other chemically unrelated ligands). Furthermore, negative and/or conflicting data obtained with experimental ligands should be considered and quoted; in other words, negative and/or conflicting data should be used for retrospective analysis and to interpret data already published with these compounds. Finally, potential off-target effects, including non-receptor-mediated effects, should always be considered and might sometimes be suspected by looking at the reversibility (many receptor-mediated effects are reversible, toxic non-specific effects are often irreversible), concentration–response or dose–response relationship, or at the kinetics of the response (some effects are too slow or too fast to be compatible with a given receptor function).

Conflict of Interest Statement

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.