Bioluminescence imaging of G protein-coupled receptor activation in living mice

G protein-coupled receptors (GPCRs), a superfamily of cell-surface receptors involved in virtually all physiological processes, are the major target class for approved drugs. Imaging GPCR activation in real time in living animals would provide a powerful way to study their role in biology and disease. Here, we describe a mouse model that enables the bioluminescent detection of GPCR activation in real time by utilizing the clinically important GPCR, sphingosine-1-phosphate receptor 1 (S1P1). A synthetic S1P1 signaling pathway, designed to report the interaction between S1P1 and β-arrestin2 via the firefly split luciferase fragment complementation system, is genetically encoded in these mice. Upon receptor activation and subsequent β-arrestin2 recruitment, an active luciferase enzyme complex is produced, which can be detected by in vivo bioluminescence imaging. This imaging strategy reveals the dynamics and spatial specificity of S1P1 activation in normal and pathophysiologic contexts in vivo and can be applied to other GPCRs.

G protein-coupled receptors (GPCRs) (also known as seven-transmembrane-domain receptors) are the largest and most diverse gene superfamily in the human genome, comprising greater than 3% of the protein coding genes [1][2][3] . GPCRs are widely expressed, initiate cellular signal transduction by a diverse array of extracellular ligands, and are involved in virtually all physiological functions 1,4 . They are also extremely important in clinical medicine as a major drug target class. It is estimated that~30% of approved pharmaceutical drugs are directed to GPCRs 4,5 . Despite their enormous biological and clinical importance, technologies to image GPCR activation in their native physiologic environment in real time are lacking, hampering a full understanding of the spatiotemporal dynamics of their activation in vivo by endogenous or exogenous ligands.
Sphingosine-1-phosphate receptor 1 (S1P 1 ) is one of a fivemember GPCR family with high affinity for the bioactive sphingolipid, sphingosine-1-phosphate (S1P). It is ubiquitously expressed among tissues and is highly enriched in endothelial cells, where it serves as a key regulator of vascular barrier function [6][7][8] . S1P 1 also performs important actions in the immune and nervous systems 7 . Its endogenous sphingolipid ligand, S1P, is produced by all cells and is carried by lipoproteins and albumin in the circulation 7 . S1P-S1P 1 signaling has been linked to diverse disease processes, including infection, multiple sclerosis, atherosclerosis, inflammatory bowel disease, and cancer 6,7,9 . FTY720 (fingolimod/Gilenya TM ), an S1P 1 ligand, is an FDA-approved drug for the treatment of remitting-relapsing multiple sclerosis 10 .
Ligand-activated GPCRs stimulate intracellular signaling pathways via G proteins 11,12 and other effectors, including βarrestins [13][14][15][16] . β-Arrestin binding to activated GPCRs also mediates receptor desensitization, internalization, and recycling. The near-universal interaction between GPCRs and β-arrestins following agonist binding has been employed in assays for quantifying receptor activation [17][18][19][20][21] . We recently developed a mouse model for the detection of activated S1P 1 in which a green fluorescent protein (GFP) reporter gene was activated 22 . This model enabled the identification of cells in which receptor activation has occurred in a cumulative manner, but it is not suited for intravital imaging of real-time activation of GPCR signaling.
Here, we describe a genetic model in mice that enables the imaging of S1P 1 signaling in real time in the native physiological environment of the receptor. These S1P 1 signaling mice were genetically designed to report the interaction between S1P 1 and β-arrestin2 upon receptor activation via firefly split luciferase fragment complementation, which produces an active enzyme complex whose activity can be detected by bioluminescence imaging. The S1P 1 signaling mice respond to synthetic ligands in accordance to the specificity of the native receptor and the stability of pharmacologic agonists. The S1P 1 signaling mice also reveal the timing and anatomical localization of receptor activation by the endogenously synthesized S1P ligand during endotoxin-induced systemic inflammation. This paradigm can be applied to other members of the clinically important GPCR family to enable the study of receptor activation in their in vivo settings.

Results
Generation of S1P 1 luciferase signaling mice. Our strategy for imaging the activation of a GPCR in real time in living mice was to adapt a firefly split luciferase complementation system 21,23,24 for the detection of the interaction between ligand-activated S1P 1 and β-arrestin2 (Fig. 1a). Both S1P 1 and β-arrestin2 were genetically modified as individual fusion proteins to carry inactive, but complementary, fragments of luciferase. The S1P 1 C terminus was linked to the C-terminal fragment of firefly luciferase (CLuc: amino acids #394-550), and the β-arrestin2 N terminus was linked to the N-terminal fragment of firefly luciferase (NLuc: amino acids #1-416). In this scheme, S1P 1 activation will promote its interaction with β-arrestin2, facilitating the association of the inactive luciferase fragments and producing an active enzyme complex that, in the presence of substrates ATP and D-luciferin, generates light that can be detected by bioluminescence imaging (BLI).
The efficacy and specificity of the split luciferase complementation system for detection of S1P 1 activation was tested by cotransfection of expression vectors containing the S1P 1 -CLuc and NLuc-β-arrestin2 fusion genes into U2OS cells. As a negative interaction control, NLuc was fused to herpes simplex virus thymidine kinase (HSV-tk) in place of β-arrestin2, and cotransfected with S1P 1 -CLuc. In transfected cells treated with vehicle, minimal luciferase activity was detected ( Supplementary  Fig. 1a, b). However, addition of S1P or RP-001, a synthetic S1P 1selective agonist 25 , caused a significant increase in luciferase activity in the S1P 1 -CLuc/NLuc-β-arrestin2-transfected cells compared with the S1P 1 -CLuc/NLuc-HSV-tk-transfected cells. These results demonstrate that the split luciferase complementation system, utilizing S1P 1 and β-arrestin fusions, can specifically detect S1P 1 activation.
Gene targeting in embryonic stem cells was employed to produce two individual mouse lines, each carrying one of the fusion constructs ( Supplementary Fig. 2a, b). One mouse line was generated with the S1P 1 -CLuc fusion knocked in to the S1P 1 coding region. In this configuration, the native S1P 1 promoter elements were maintained to allow for expression of the S1P 1 fusion gene in the endogenous context. The other mouse line was established with the NLuc-β-arrestin2 fusion knocked in to the safe harbor Rosa26 locus under control of the Rosa26 promoter elements to provide a ubiquitous expression pattern, ensuring that NLuc-β-arrestin2 would be co-expressed along with S1P 1 -CLuc. The S1P 1 -CLuc knock-in mice were crossed with the NLuc-β-arrestin2 knock-in mice to derive mice carrying both alleles (Fig. 1a), which are termed S1P 1 luciferase signaling mice.
To determine whether the synthetic S1P 1 signaling pathway genetically encoded within the S1P 1 luciferase signaling mice was responsive to ligands with specificity similar to native S1P 1 , primary mouse embryonic fibroblasts (MEFs) derived from S1P 1 luciferase signaling mice were utilized. Addition of S1P but not vehicle rapidly induced bioluminescence in the MEFs (Fig. 1b). Bioluminescence activity peaked within 3 min after addition of S1P to the MEFs and was rapidly lost thereafter (Fig. 1b). This loss of activity indicates that luciferase fragment complementation is transient and may reflect the rapid kinetics of S1P 1 desensitization after activation and the subsequent dissociation of the receptor-β-arrestin2 complex 26 . S1P and RP-001, a potent synthetic S1P 1 -selective agonist 25 , induced luciferase activity in the S1P 1 luciferase signaling MEFs, but not in MEFs carrying only the S1P 1 -CLuc or the NLuc-β-arrestin2 alleles (Fig. 1b, c), showing that both components of the genetically encoded split luciferase complementation system were necessary to report the ligand-activated S1P 1 -CLuc interaction with NLuc-β-arrestin2. Two structural analogs of S1P, lysophosphatidic acid (LPA) and sphingosine, that are not ligands for S1P 1 did not induce bioluminescence when added to S1P 1 luciferase signaling MEFs at a concentration of 10 −6 M (Fig. 1d, e).

ARTICLE
To determine if the genetically encoded split luciferase complementation system could detect basal S1P 1 activation in vivo, S1P 1 luciferase signaling mice, as well as mice carrying only the S1P 1 -CLuc or NLuc-β-arrestin2 alleles, were subjected to BLI while under anesthesia. Significantly higher levels of bioluminescence activity were detected in the S1P 1 luciferase signaling mice compared with the mice carrying only the S1P 1 -CLuc or NLuc-β-arrestin2 alleles (Fig. 1g). We quantified bioluminescence activity in the cephalic, thoracic, and epigastric regions, as the activity varied between different body regions. The highest activity was observed in the epigastric region, followed by the thoracic and cephalic regions (Fig. 1h). To identify the anatomical sites of basal S1P 1 activation in S1P 1 luciferase signaling mice, S1P 1 luciferase signaling mice were subjected to BLI after surgically exposing the thoracic region and abdominal cavity. The highest bioluminescence signals were detected in lungs, in lymph nodes, and in the splenic region. Notably, no signal was detected in the liver (Fig. 1i).
Treatment of S1P 1 signaling mice with antagonist. To determine if the bioluminescence activity detected in S1P 1 luciferase signaling mice was the result of endogenous S1P 1 activation, the mice were injected with W146 28 , a potent S1P 1 -selective antagonist (Fig. 2a). W146 treatment administered 30 min prior to BLI significantly lowered the bioluminescence in the S1P 1 luciferase signaling mice by 70-80% in a dose-dependent manner in the cephalic, epigastric, and thoracic regions compared with the same mice treated with vehicle ( Fig. 2b-d). The basal bioluminescence activity was completely restored in these S1P 1 luciferase signaling mice 1 day post-W146 injection ( Fig. 2b-d), which is consistent with the rapid clearance reported for this compound 29 . The inhibition of bioluminescence activity by an S1P 1 antagonist indicates that the basal activity detected in the live mice is primarily the result of endogenous S1P 1 activation.
Treatment of S1P 1 signaling mice with agonists. FTY720 (fingolimod/Gilenya TM ), a sphingosine analog, is phosphorylated by sphingosine kinases in vivo, and converted to a form that, at nanomolar concentrations, activates S1P 1 , as well as S1P 3 , S1P 4 , and S1P 5 10 . Its half-life in blood after oral administration in rats is~24 h 30 . To determine the time course of S1P 1 activation by FTY720 in the S1P 1 luciferase signaling mice, the mice were imaged prior to the administration of compounds and then imaged serially at 1.5, 6, 24, and 48 h after intraperitoneal injection of each compound (Fig. 3). FTY720 concurrently induced peak levels of bioluminescence activity in the cephalic, thoracic, and epigastric regions at 6 h after injection. Significant bioluminescence activity continued to be detected at 24 and 48 h after injection of FTY720, commensurate with the relatively long halflife of the drug. BLI performed on FTY720-treated S1P 1 luciferase signaling mice (5 h after FTY720 treatment) with their internal organs surgically exposed showed highly elevated bioluminescence signals over lung, lymph nodes, spleen, and liver (Supplementary Fig. 3a). Compared with FTY720, RP-001 is a very shortacting S1P 1 agonist that decreases to undetectable levels in blood by 8 h after administration in mice 25   induced a peak of bioluminescence activity 1.5 h after administration to S1P 1 luciferase signaling mice in the cephalic, thoracic, and epigastric regions. The bioluminescence signal was most intense in the epigastric region. By 24 and 48 h after dosing, the bioluminescence activity in RP-001-treated S1P 1 luciferase signaling mice declined to baseline levels ( Fig. 3b-e).
LPS-induced systemic inflammation in S1P 1 signaling mice. S1P 1 signaling is activated during systemic inflammation induced by bacterial lipopolysaccharide (LPS); however, the timing and anatomical distribution of receptor activation is not well established 22 . In order to define these parameters, S1P 1 luciferase signaling mice were intraperitoneally injected with a sublethal dose of LPS and then subjected to serial BLI at 2, 6, 24, 48, 72, and 96 h (Fig. 4). At 2 h after injection, a weak increase in bioluminescence was observed primarily in the epigastric region, which returned to baseline levels at 6 h. At 24 h after administration, the bioluminescent signal was significantly increased in the cephalic and thoracic regions (Fig. 4b, c and Supplementary Fig. 5). At 48 h after administration, bioluminescent signal was elevated in the thoracic region. At 72 h after administration, the strongest signal was observed in the epigastric region. At 24 h after LPS administration, BLI performed in S1P 1 luciferase signaling mice with surgically exposed internal organs demonstrated bioluminescent signals induced over lungs, lymph nodes, spleen, and liver ( Supplementary Fig. 3b). These results show that LPS induces heightened S1P 1 activation in a sustained manner systemically. In addition, the timing of S1P 1 activation was distinctive for specific anatomical locations.
A major fraction of circulating S1P is produced by hematopoietically derived cells 31 . To determine if hematopoietic cell-derived S1P is critical for systemic S1P 1 activation during LPS-induced inflammation, irradiated S1P 1 luciferase signaling mice were transplanted with sphingosine kinase-deficient bone marrow from plasmaS1Pless mice, which lack the ability to produce S1P in the hematopoietic system 31 , and then subjected to BLI (Fig. 5a). Plasma S1P and dhS1P levels were significantly decreased in the plasmaS1Pless bone marrow-transplanted mice to~20 and 10%, respectively, of the levels observed in control bone marrow-transplanted mice (Fig. 5b). The residual S1P and dhS1P in the plasmaS1Pless bone marrow-transplanted mice are likely produced by the sphingosine kinase-replete endothelial cells of the recipient mice 32   and in the epigastric region at 24 h (Fig. 5c-f). The residual bioluminescence activity in plasmaS1Pless bone marrowtransplanted mice at 24 h after LPS treatment was significantly inhibited by the S1P 1 antagonist W146 (Supplementary Fig. 4). These results suggest that the low levels of dhS1P and S1P remaining in these mice may provide some signaling activity. A high degree of mortality occurred 24 h after LPS treatment in S1P 1 luciferase signaling mice transplanted with plasmaS1Pless bone marrow, precluding further measurements. These results indicate that hematopoietically derived S1P is responsible for a significant portion of the S1P 1 activation that occurs in vivo during LPS-induced inflammation. S1P 1 activation in the brain by inflammation. During LPSinduced inflammation, bioluminescence activity was significantly increased in the cephalic region at 24 h (Fig. 4b, c and Supplementary Fig. 5), raising the possibility that S1P 1 may be activated in the central nervous system (CNS). To determine if S1P 1 signaling was activated within the CNS and, if so, in which cell types, we utilized the previously described S1P 1 GFP signaling mice 22 . In these mice, S1P 1 activation is also based on interaction of the GPCR with β-arrestin2; however, in this reporter system the interaction leads to the proteolytic release of a transcription factor tethered to S1P 1 that enters the nucleus and activates a histone-GFP reporter gene. The cells in which S1P 1 activation occurs are then stably marked with GFP-labeled nuclei (Fig. 6a). LPS was administered to S1P 1 GFP signaling mice and, 7 days later, brain tissue was collected and sectioned, then the cortex, cerebellum, and brainstem were examined for GFP-labeled cells (Fig. 6b, c).
Compared with vehicle-treated mice, LPS-treated mice exhibited increased numbers of GFP-positive cells in each of these regions, indicating that S1P 1 activation had taken place in the CNS. The GFP-positive cells appeared to be associated with the vasculature and were CD31 positive, establishing their identity as endothelial cells (Fig. 6c). Next, brain endothelial cells were isolated from S1P 1 luciferase signaling mice, treated with S1P or RP-001, and examined by BLI. Bioluminescent activity was induced in the brain endothelial cells in a dose-dependent manner, with RP-001 exhibiting a higher potency than S1P (Fig. 6d-f). Collectively, the results indicate that S1P 1 is activated in endothelial cells of the neurovascular unit of the brain during systemic inflammation induced by LPS and illustrate that the two different S1P 1 signaling mouse models provide complementary information for defining S1P 1 activation in vivo.

Discussion
The luciferase signaling model for GPCR activation described here affords a way to study GPCRs that was not previously possible, yielding spatial and temporal information on the status of endogenous receptor activation in real time in a living animal. The nearly universal interaction between GPCRs with βarrestin 13 was exploited to induce assembly of inactive fragments of luciferase, producing an active enzyme complex and enabling BLI upon activation of a GPCR, S1P 1 , in living mice. The in vivo S1P 1 signaling detection system was shown to be activated by both synthetic and natural ligands of S1P 1 . BLI of the luciferase signaling mice revealed the anatomical locations of homeostatic S1P 1 signaling, as well as the dynamics of S1P 1 signaling under  inflammatory conditions. Although the resolution of the luciferase signaling model was limited to the identification of signaling at an organ or tissue level, the associated use of the previously described GFP signaling model for S1P 1 activation 22 , which relies on transcription factor reporter gene activation, provided a means to ultimately define S1P 1 signaling sites at a cellular level. While visualizing the source of the bioluminescent signal with single cell resolution is not possible at present, detection of signaling in realtime in vivo at a single cell level would be an important goal to achieve.
We have demonstrated that the S1P 1 luciferase signaling mice can be used for the characterization of the in vivo signaling activity of the receptor-active pharmacological compounds FTY720 and RP-001 (Fig. 3). FTY720 10 , an approved drug (fingolimod/Gilenya TM ) used for the treatment of multiple sclerosis, targets S1P receptors 1, 3, 4, and 5. It has a relatively long half-life in blood 30 and has been demonstrated to have diverse physiologic effects in the immune, cardiovascular, and nervous systems 7 . RP-001 is an S1P 1 -selective agonist that is more potent than FTY720, but with a much shorter half-life in blood. Similar to FTY720, RP-001 has been shown to trigger lymphopenia 25 . Administration of these compounds to the S1P 1 luciferase signaling mice enabled determination of the timing of signaling and anatomical location of signaling events, which indicates that this model will be generally useful for characterization of S1P 1 modulators. The signaling induced by RP-001 was relatively intense but decayed rapidly, commensurate with its high potency and short half-life (Fig. 3b-e). The very intense epigastic signal induced by RP-001 may reflect its rapid clearance in the liver. In contrast, the signaling induced by FTY720 in vivo persisted over a long duration, lasting up to 48 h (Fig. 3b-e). While consistent with the relatively long half-life of the compound, prolonged signaling, and the apparent absence of rapid desensitization would be compatible with the existence of a receptor reserve. Under normal conditions, a receptor reserve may underpin the long-term tonic S1P 1 signaling in the vascular endothelium. Compartmentalization of S1P 1 receptors may be a mechanism underlying such a receptor reserve 33,34 . However, a different mechanism must operate on   a Bone marrow cells from control or plasma(p)S1Pless mice were transplanted (BMT) into irradiated S1P 1 luciferase signaling (S1P 1 LS) mice to produce control transplanted (Control-T) or pS1Pless transplanted (S1Pless-T) mice, respectively. LPS was injected intraperitoneally into the stably transplanted mice 20 weeks later. Mice were subjected to BLI prior to the LPS injection (baseline, B/L) and 2, 6, and 24 h after LPS injection. b Plasma S1P and dhS1P levels in S1P 1 luciferase signaling mice, transplanted with control (Control-T) or pS1Pless (S1Pless-T) bone marrow. Data represent the mean ± SEM. n = 6 for Control-T mice, n = 8 for S1Pless-T mice. c Representative bioluminescence images of the same S1P 1 luciferase signaling mice, transplanted with control (Cont-T) or pS1Pless (S1Pless-T) bone marrow, at the specified time points. Mice were imaged in the supine position. Red open rectangles representing ROI were positioned around cephalic, thoracic, and epigastric regions. d-f The bioluminescence activity was quantified by determining the total flux (photons/sec; p/sec) in the cephalic (d), thoracic (e), and epigastric (f) ROI. Data represent the mean ± SEM. n = 5 for control BM-transplanted S1P 1 luciferase signaling (Control-T) mice and n = 7 for pS1Pless BM-transplanted S1P 1 luciferase signaling (S1Pless-T) mice at the baseline, 2 h, and 6 h time points. n = 4 for control BMtransplanted S1P 1 luciferase signaling (Control-T) mice and n = 6 for pS1Pless BM-transplanted S1P 1 luciferase signaling (S1Pless-T) mice at 24 h. P values were determined two-tailed Student' t-test; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01340-7 ARTICLE other cell types, such as lymphocytes, that rapidly downmodulate S1P 1 upon exposure to agonist 25 .
The S1P 1 luciferase signaling mice provide the unique capability to detect the signaling activity of the endogenously produced natural ligand under basal and pathophysiological conditions in real time. Under basal conditions, lung and lymphoid tissues showed the highest levels of signaling (Fig. 1i). This is compatible with ample evidence for a functional S1P 1 signaling requirement in these tissues. In lung, blocking S1P 1 signalingeither by removal of S1P from the blood 35 or by S1P 1 antagonism 28,29 -induced vascular leakage, indicating that S1P 1 signaling is critical to maintain vascular integrity in lung. In lymphoid tissues, S1P 1 signaling has been shown to be necessary for lymphocyte egress 36 and for the positioning of cells, such as splenic marginal zone B cells 37 . S1P 1 signaling is also required in endothelial cells in lymph nodes, notably on high endothelial venules, which serve as the portals for lymphocyte entry to maintain their vascular integrity 38 .
Under conditions of systemic inflammation induced by LPS exposure, the S1P 1 luciferase signaling mice reported regionspecific dynamics in S1P 1 activation (Fig. 4). Liver showed little or no signaling activity under basal conditions (Fig. 1i), but exhibited dramatically increased S1P 1 signaling during systemic inflammation (Supplementary Fig. 3b). Hepatocytes have been shown to express abundant S1P 1 that is accessible for signaling in response to FTY720 administration 22 . Based on the findings reported here, they are apparently not exposed to sufficient S1P under basal conditions to stimulate signaling. However, during conditions that induce vascular leakage, such as inflammation, S1P derived from blood sources can stimulate S1P 1 signaling in hepatocytes and endothelial cells of liver, and endothelial cells in lung, as has been demonstrated in an S1P 1 GFP signaling mouse model 22 . We have extended these finding here by showing increased S1P 1 signaling in brain endothelial cells during systemic inflammation (Fig. 6).
The strategy used to produce the S1P 1 luciferase signaling mice provides a template for the development of a library of other GPCR-signaling models. Creation of models for other GPCRs that similarly interact with β-arrestin2 should only require the derivation of a mouse with a receptor knock-in of the C-terminal Luc fragment (amino acids #394-550). Subsequent cross-breeding with the NLuc-β-arrestin2 mouse described here, which carries the complementary luciferase N-terminal fragment (amino acids #1-416), would yield unique GPCR-signaling models. These a Schematic of the design to monitor S1P 1 activation in S1P 1 GFP signaling mice. S1P 1 is fused to a transcription factor (TF) via a protease recognition sequence. Ligand binding to the receptor stimulates the recruitment of a β-arrestin2-TEV protease fusion protein, triggering the release of the transcription factor from the C terminus of modified S1P 1 . The transcription factor stimulates expression of a histone-GFP reporter gene, fluorescently labeling cell nuclei. b LPS or vehicle was injected intraperitoneally into S1P 1 GFP signaling (S1P 1 GS) mice, and the brains were harvested 7 days after injection and examined by immunofluorescence. The tissues of three mice injected with LPS and three mice injected with vehicle were examined. c Representative histological sections were immunostained with antibodies to CD31, and the images of cerebral cortex, cerebellum, and brainstem were captured using an inverted laser scanning confocal microscope. Scale bar, 50 μm. d Brain endothelial cells (EC) were isolated from S1P 1 luciferase signaling (S1P 1 LS) mice, grown in triplicate in 24-well plates, treated with S1P or RP-001 and D-luciferin, and immediately subjected to BLI. e Bioluminescence images of the cells exposed to the indicated concentrations of S1P or RP-001 represent a 15-min acquisition. The experiment was repeated twice. f The bioluminescence activity was quantified by determining the total flux (photons/sec; p/sec). Data represent the mean ± SEM, n = 6 models can be used to gain new understandings of GPCR signaling in normal and disease-specific biological contexts, and enable in vivo analysis of GPCR-signaling perturbation by receptor-active compounds, thereby facilitating drug development for this important target class.
Targeting vectors for mouse embryonic stem cells were constructed to insert the S1P 1 -CLuc or NLuc-β-arrestin2 fusions into the S1pr1 or Rosa26 locus, respectively ( Supplementary Fig. 2a, b). In the S1pr1-targeting vector, the neomycin resistance gene (NeoR), flanked by loxP sites, was inserted downstream of the fusion gene. In the Rosa26-targeting vector 39  The scheme to derive S1P 1 luciferase signaling mice is shown in Fig. 1a. Mice heterozygous for S1P 1 -CLuc (-NeoR) knock-in mice were crossed with mice heterozygous for NLuc-β-arrestin2 (-NeoR) knock-in mice to obtain double heterozygotes, which were back crossed to mice heterozygous for NLuc-β-arrestin2 (-NeoR) to obtain mice heterozygous for S1P 1 -CLuc (-NeoR) and homozygous for NLuc-β-arrestin2 (-NeoR). S1P 1 luciferase signaling mice that were heterozygous for the S1P 1 -CLuc (-NeoR) allele and homozygous for the NLuc-β-arrestin2 (-NeoR Stop) allele showed a higher bioluminescence signal after FTY720 treatment than mice heterozygous for both of these alleles (Supplementary Fig. 6). For this reason, mice heterozygous for the S1P 1 -CLuc (-NeoR) allele and homozygous for the NLuc-β-arrestin2 (-NeoR stop) allele were used in S1P1 luciferase signaling mouse studies. Animals were housed in a specific pathogen-free facility and provided food and water ad libitum. Age-matched offspring of both sexes (2-5 months old) were used for experiments. The mice were at the initiation of experiments. The sample size was selected based on expected effect size. No randomization method was employed. Investigators were blinded to the genotype of animals during data analysis. No experimental animals were excluded from the analysis.
For BLI in live mice, the hair was removed with a clipper and hair removal cream. XenoLight Rediject D-luciferin (PerkinElmer) (150 mg/kg BW) was injected intraperitoneally, and the mice were subjected to BLI using an in vivo imaging system (IVIS Lumina II). During the imaging, the mice were anesthetized with isoflurane (Baxter) using an XGI-8 gas anesthesia system (PerkinElmer). Eye lubricant (Puralube Vet Ointment, Dechra) was applied during the anesthesia. Imaging parameters were: binning = medium (8), field of view = 12.5 × 12.5 cm, f/ stop = 1. Each image exposure lasted 5 min, and a total of four sequential images were captured. The second 5-min image was used in all experiments. Images were analyzed using Living Image Software (PerkinElmer). Regions of interest (ROI) were drawn around each body region and the total flux (photons/s) was determined.
For BLI of internal organs, mice were anesthetized by intraperitoneal injection of ketamine (180 mg/kg BW)-xylazine (12 mg/kg BW), and after 15 min received an intraperitoneal XenoLight Rediject D-luciferin (PerkinElmer) injection (500 mg/kg BW). The ventral skin was incised along the midline to expose the internal organs. Bioluminescence activity was immediately detected with an in vivo imaging system (IVIS Lumina II) with four acquisitions lasting 5 min each.
Analysis of S1P 1 activation in brain. S1P 1 GFP signaling mice have been previously described 22 . LPS (16 mg/kg BW; vehicle, PBS) was injected intraperitoneally into S1P 1 GFP signaling mice. Seven days after the injection, the mice were perfused with normal saline followed by chilled 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) and the brains were harvested. The tissues were post fixed in 4% paraformaldehyde for 4 h and incubated in 30% sucrose in 0.1 M phosphate buffer (pH 7.4) at 4°C for 2 days, then embedded in OCT (Sakura) and sectioned at 7-μm thickness. For immunostaining, nonconjugated primary antibody for mouse CD31 (clone SZ31, Dianova) and fluorescently labeled secondary antibody (donkey anti-rat IgG Cyanine Cy3, Jackson ImmunoResearch Laboratories) were used. The images were captured with an inverted laser scanning confocal microscope (LSM 780, Carl Zeiss Microscopy) using Zeiss Zen 2012 software.
Sphingolipid analysis. S1P and dhS1P were measured by HPLC-tandem MS by the Lipidomics Core at the Medical University of South Carolina on a Thermo Finnigan TSQ 7000 triple quadrupole mass spectrometer (Thermo Fisher Scientific), operating in a multiple reaction monitoring-positive ionization mode as described 44 .
Statistical analysis. GraphPad Prism v.6.0 software was used for statistical analyses. To determine the significance between three or more test groups, analysis of variance (ANOVA) was used followed by either Bonferroni's for comparison to the control group or Tukey's to compare all groups. Two-tailed Student's t-test was used for the direct comparison of two groups. In all cases, P ≤ 0.05 was considered statistically significant.
Study approval. All animal procedures were approved by the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases and performed in accordance with the National Institutes of Health guidelines.
Data availability. The authors declare that all data supporting the findings of this study are available within this paper or its Supplementary Information file or are available from the corresponding author on request.