Role of Adenosine Kinase in Sphingosine-1-Phosphate Receptor 1-Induced Mechano-Hypersensitivities

Emerging evidence implicates the sphingosine-1-phosphate receptor subtype 1 (S1PR1) in the development of neuropathic pain. Continued investigation of the signaling pathways downstream of S1PR1 are needed to support development of S1PR1 antagonists. In rodents, intrathecal (i.th.) injection of SEW2871, a selective S1PR1 agonist, activates the nod-like receptor family, pyrin domain containing 3 inflammasome, increases interleukin-1β (IL-1β) and causes behavioral hypersensitivity. I.th. injection of a IL-1β receptor antagonist blocks SEW2871-induced hypersensitivity, suggesting that IL-1β contributes to S1PR1’s actions. Interestingly, previous studies have suggested that IL-1β increases the expression/activity of adenosine kinase (ADK), a key regulator of adenosine signaling at its receptors (ARs). Increased ADK expression reduces adenosine signaling whereas inhibiting ADK restores the action of adenosine. Here, we show that SEW287-induced behavioral hypersensitivity is associated with increased expression of ADK in astrocytes of the dorsal horn of the spinal cord. Moreover, the ADK inhibitor, ABT702, blocks SEW2871-induced hypersensitivity. These findings link ADK activation to S1PR1. If SEW2871-induced pain is mediated by IL-1β, which in turn activates ADK and leads to mechano-allodynia, then blocking ADK should attenuate IL-1β effects. In support of this idea, recombinant rat (rrIL-1β)-induced allodynia was blocked by at least 90% with ABT702, functionally linking ADK to IL-1β. Moreover, the selective A3AR antagonist, MRS1523, prevents the ability of ABT702 to block SEW2871 and IL-1β-induced allodynia, implicating A3AR signaling in the beneficial effects exerted by ABT702. Our findings provide novel mechanistic insight into how S1PR1 signaling in the spinal cord produces hypersensitivity through IL1-β and ADK activation.


Introduction
Recent epidemiological studies on chronic pain reveal that 50.0 million of U.S. adults (20.4%) have experienced chronic pain and 19.6 million (8.0%) have experienced high-impact chronic pain (Dahlhamer et al. 2016). Since pain drugs often fail to achieve adequate relief in patients, and opioid treatments are associated with abuse and addiction, developing new pain medicines is crucial. Over the last several years, the sphingosine-1-phosphate (S1P) receptor subtype 1 (S1PR1) has emerged as an attractive molecular target for therapeutic intervention (Squillace et al. 2020). Several studies found that alterations in sphingolipid metabolism in the spinal dorsal horn increase S1P production and contribute to hypersensitivity associated with cancer pain (Grenald et al. 2017), opioid-induced hyperalgesia (Muscoli et al. 2010;Doyle et al. 2020a), neuropathic pain caused by chemotherapy (Stockstill et al. 2018;Janes et al. 2014), multiple sclerosis (Doolen et al. 2018) or traumatic peripheral nerve injury . Blocking S1PR1 with functional or competitive S1PR1 antagonists prevents and reverses neuropathic pain states of various etiologies Janes et al. 2014;Stockstill et al. 2018Stockstill et al. , 2020. Moreover, studies in breast cancer patients showed that S1PR1 is a potential molecular driver of sensory peripheral neuropathy induced by chemotherapy (Chua et al. 2020). In the spinal cord, S1PR1 activation leads to the development of mechanical and thermal hypersensitivity through nucleotide-binding oligomerization domain-like receptor protein 3 (NLRP3) inflammasome activation and interleukin-1β (IL-1β) signaling . IL-1β, a potent neuroexcitatory, inflammatory and pronociceptive cytokine (Rossi et al. 2014) has been linked to the development of S1PR1-induced hypersensitivity ), but the mechanisms whereby IL-1β contributes to S1PR1's induced hypersensitivity are not known. Here, we provide evidence that IL-1β drives S1PR1's effects through alterations in the adenosine pathway in the spinal cord that are dependent upon adenosine kinase. Our findings uncover the existence of a functional connection between S1PR1 and adenosine signaling.

Results and Discussion
Intrathecal (i.th.) injection of the highly selective S1PR1 agonist, SEW2871 (2 nmol (Sanna et al. 2004) activates NLRP3 in the dorsal horn of the spinal cord (DH-SC) and evokes significant mechano-allodynia that is maximal within 2 h post-injection in both female and male rats (Fig. 1a). NLRP3 is one of seven known nod-like receptors that form multimolecular inflammasome complexes (Kim et al. 2016;Davis et al. 2011) in the central nervous system (CNS) that function as intracellular sensors for infectious agents, as well as for host-derived danger signals (Voet et al. 2019;Tsuchiya and Hara 2014;Walsh et al. 2014). On activation, NLRP3 forms the inflammasome complex by binding and stimulating the autoactivation of the cysteine protease caspase-1 that catalyzes the cleavage of proIL-1β to the active form, IL-1β (Lopez-Castejon and Brough 2011;Franchi et al. 2009). At time of peak allodynia (Fig. 1a) and when compared with animals that received its vehicle, SEW2871 causes a significant increase in the expression of the NLRP3 subunit (Fig. 1b), cleaves caspase-1 (Fig. 1c) and increases IL-1β expression (Fig. 1d). IL-1β, a potent neuroexcitatory cytokine (Viviani et al. 2003) released from neurons, microglia and astrocytes (Ren and Torres 2009) contributes directly to S1PR1's induced hypersensitivity . Although the molecular mechanisms remain unexplored, we implicate roles for adenosine kinase (ADK), the key determinant of adenosine signaling (Boison 2013;Aronica et al. 2013). In the adult CNS, ADK is highly expressed in astrocytes and is upregulated in response to increased astrocyte reactivity (Boison et al. 2010). Noteworthy, S1PR1 is highly expressed on astrocytes (Healy and Antel 2016;Van Doorn et al. 2010). Astrocytes are a key cellular hub for the pharmacological actions of S1PR1 activity ) and S1PR1 antagonism (Doyle et al. 2020a).
In vitro studies in astrocytes suggested that IL-1β increases the expression/activity of ADK (Aronica et al. 2011). It is well-documented that adenosine signaling is significantly reduced when ADK expression is increased (Boison 2013). We recently reported that activation of the A 3 AR subtype, one of adenosine's four known receptors, inhibits behavioral hypersensitivities that accompany many pain states . Moreover, increased expression of ADK in the spinal cord has been linked to decreased adenosine-A 3 AR signaling and inhibiting ADK activity blocks the development of hypersensitivity in an A 3 AR-dependent manner (Little et al. 2015;Wahlman et al. 2018). Whether S1PR1-induced hypersensitivity is driven by IL-1β and effects downstream of ADK is not known.
Our immunofluorescence data show that the S1PR1 agonist, SEW2871, increases ADK expression in the dorsal horn lamina I and II of the DH-SC ( Fig. 1g-i). This increase was associated with astrocytes identified by colocalization staining with the specific astrocyte marker, GFAP (Fig. 1h, i). We measured levels of adenosine and upstream metabolites (adenosine triphosphate, ATP; adenosine diphosphate, ADP; adenosine monophosphate, AMP) or breakdown products (inosine, AMP) using a ultra-high-performance liquid chromatograph/high-resolution mass spectrometry (UHPLC-MS/ HRMS) assay in DH-SC in response to S1PR1 activation and found no changes (Supplementary Figs. 1-3, Table 1). The LC-HRMS/MS method developed was linear over the measured concentration range of 50 nM to 100 µM and we detected adenosine and inosine in each sample. It is possible that the enzymatic processes were not fully quenched, thus preventing any observation of changes in the tissue samples since these analytes are involved in rapid breakdown processes. In a separate assay, we also measured nicotinamide (NAM), nicotinamide mononucleotide (NMN), 1-methyl NAM, nicotinamide adenine dinucleotide (NAD), nicotinamide adenine dinucleotide phosphate (NADPH), ATP, guanosine triphosphate (GTP), cytidine triphosphate (CTP), thymidine triphosphate (TTP), uridine triphosphate (UTP), inosine monophosphate (IMP), AMP, cytidine monophosphate (CMP), guanosine monophosphate (GMP), uridine monophosphate (UMP) and thymidine monophosphate (TMP). Again, we found no significant differences (data not shown). However, S1PR1-induced ADK expression must have compromised endogenous adenosine signaling, since i.th. injection of the potent (IC 50 = 1.7 nM) and cord drives S1PR1-induced mechano-allodynia. When compared to its vehicle (3% DMSO in saline; n = 6 males and n = 4 females), i.th. administration of SEW2871 (2 nmol; n = 6 males and n = 4 females) induces mechano-allodynia in both male and female rats with peak effects reached at 2 h after injection (a). The development of mechanoallodynia was prevented by i.th. administration of the ADK inhibitor ABT702 (30 nmol; n = 6 males and n = 4 females). I.th. injection of the A 3 AR antagonist MRS1523 (3 nmol; n = 4 males and n = 4 females) blocked the beneficial effect induced by ABT702 (30 nmol) (a). Levels of NLRP3 (b, n = 5), cleaved caspase-1 (c, n = 5) and IL-1β (d, n = 5) in DH-SC harvested from SEW2871treated rats were significantly increased when compared to DH-SC harvested from rats 2 h after i.th. vehicle (b-d, n = 5). Results are expressed as mean ± SD for n rats and analyzed by one-way ANOVA with Tukey's comparisons (a, f) or student's t test (b-e). *p < 0.05 vs vehicle, † p < 0.05 vs SEW2871 (a-f). ADK protein expression (e, n = 6) in DH-SC harvested from rrIL-1β-treated rats were significantly increased when compared to DH-SC harvested from rats 1 h after i.th. vehicle (e, n = 6). Results are expressed as mean ± SD for n rats and analyzed by one-way ANOVA with student's t test (e). When compared with its vehicle (3% DMSO in saline; n = 7), i.th. administration of rrIL-1β (100 ng; n = 7) induced mechano-allodynia with peak effects reached at 1 h after injection (f). The development of mechano-allodynia was prevented by i.th. administration of the adenosine kinase inhibitor ABT702 (30 nmol; n = 7). I.th. injection of A 3 AR antagonist MRS1523 (3 nmol; n = 4) blocked the beneficial effect induced by ABT702 (30 nmol) (f). Results are expressed as mean ± SD for n rats and analyzed by one-way ANOVA with Tukey's comparisons (f). *p < 0.05 vs vehicle, † p < 0.05 vs IL-1β. Representative images of the dorsal horn from immunofluorescence studies show increased ADK (red) expression in animals receiving intrathecal injection of SEW2871 (2 nmol; n = 4) (h) compared to vehicle (3% DMSO in saline; n = 5)-treated animals (g). This increase is particularly prominent among astrocytes (GFAP, green) (see inset in panels) from lamina I and II (approximately bounded by yellow box). The ADK signal associated with astrocytes is significantly increased in SEW2871-treated animals compared to vehicle-treated animals (i). Results are expressed as mean ± SD for n rats and analyzed by student's t test (i). *p < 0.5 vs vehicle highly selective non-nucleoside ADK inhibitor, ABT-702 (30 nmol) (Kowaluk et al. 2000) blocked SEW2871-induced hypersensitivity in both female and male rats (Fig. 1a). These findings establish a functional link of ADK activation to S1PR1. Moreover, the anti-allodynic effects of ABT-702 were significantly attenuated in both female and male rats in the presence of MRS1523 (3 nmol), a potent and selective A 3 AR antagonist (Little et al. 2015) (Fig. 1a, f), consistent with adenosine signaling at the A 3 AR. If SEW2871-induced pain is mediated by IL-1β, which in turn increases ADK expression/activation and leads to mechano-allodynia, then IL-1β should recapitulate the effects seen with SEW2817. In support of this idea, i.th. injection of recombinant rat IL-1β (rrIL-1β) increases ADK expression in the DH-SC (Fig. 1e) and mechano-allodynia (Fig. 1f). These effects are blocked with ABT702 by at least 90% (Fig. 1f) and MRS1523 attenuates ABT702's ability to block IL-1β-induced allodynia (Fig. 1f), implicating A 3 AR signaling in the beneficial effects exerted by ABT702.
Collectively, these results suggest a model whereby S1PR1 activation engages NLRP3, leading to IL-1β formation, increased ADK expression/activation and reduced A 3 AR signaling. Blocking ADK engages A 3 AR signalling and prevents the development of S1PR1-evoked hypersensitivity. Our findings, which provide novel mechanistic insights into molecular signaling pathways engaged by S1PR1 in the spinal cord, in turn suggest that the beneficial effects of S1PR1 antagonists observed in neuropathic pain Janes et al. 2014;Stockstill et al. 2018Stockstill et al. , 2020 may include their ability to restore A 3 AR signaling.
We tested this possibility in a well-characterized rat model of traumatic nerve injury neuropathic pain caused by chronic constriction of the sciatic nerve (CCI) (Bennett and Xie 1988). Two S1PR1 antagonists were used: the competitive S1PR1 antagonist, TASP0277308 (Fujii et al. 2012b) and FTY720, a functional S1PR1 antagonist (Brinkmann et al. 2010;Bigaud et al. 2014). We previously reported that in this model, ADK expression/activation increases in response to nerve injury in the DH-SC at the time of peak allodynia, and that systemic administration of an ADK inhibitor (ABT-70) reverses mechano-allodynia (Little et al. 2015). Moreover, i.th. injection of MRS1523 blocks the effects of ABT-702, consistent with adenosine signaling at the A 3 AR. Chronic constriction of the sciatic nerve leads to the development of mechano-allodynia, which peaks around 7 days . At time of peak mechanoallodynia (7 days after nerve injury), oral administration of TASP0277308 or FTY720 reverses mechano-allodynia in a time-dependent manner (Fig. 2a, b). These effects are abrogated by an i.p. injection of MRS1523 (Fig. 2b). In contrast, the anti-allodynic effects of morphine are not altered by MRS1523 (Fig. 2c), supporting previous findings (Doyle et al. 2020b). TASP0277308 or FTY720 were used at doses (3 mg/kg) used in previous studies . Moreover, we show that the beneficial effects observed with systemic administration of FTY720 are blocked by i.th. injection of MRS1523, implicating the DH-SC as a prominent site of action (Fig. 2d). In order to exclude any potential effects of MRS1523 as an S1PR1 antagonist, we performed a GPCR cell-based antagonist β-arrestin assay at Eurofins Discovery. MRS1523 did not inhibit S1PR1 agonist activity, confirming that it is not a S1PR1 antagonist (Fig. 2e).
In summary, we present the first evidence that S1PR1induced behavioral hypersensitivity is mediated by an IL-1β-ADK-A 3 AR signaling pathway, bridging the gap in our understanding of the mode of action of S1PR1 antagonists.

Animals
Female or male Sprague-Dawley rats (200-220 g) were purchased from Envigo Laboratories (Indianapolis, IN USA). Animals of the same sex were housed 3 per cage with 12-h light/dark cycles and food/water ad libitum. All experiments were conducted in accordance with the National Institutes of Health, the International Association for the Study of Pain guidelines for laboratory animal welfare, and following the Animal Utilization Protocol #2241 approved by the Saint Louis University Institutional Animal Care and Use Committee.

Drug Delivery
Drugs were co-administrated in a single i.th. injections performed using the Wilcox method (Hylden and Wilcox 1980): rats were lightly anesthetized with isoflurane and a 50 μL Hamilton syringe (Hamilton, Reno, NV USA) with a 25-ga needle was inserted between the L5/L6 vertebrae puncturing the dura (confirmed by presence of reflexive tail flick and hind limbs) and vehicle or test substance(s) (10 µL) was injected. The vehicle used for all i.th. injections of test agents was 3% (for SEW2871, ABT702 and rrIL-1β), or 15% (for A-438079) dimethyl sulfoxide (DMSO) in saline. The vehicle used for all oral administration of test agents was 5% DMSO/0.5% methylcellulose in water (0.2-mL dosing volume). The vehicle used for i.p. injections was 50% DMSO/saline.

Estrus Smears
Rat vaginal smears were taken daily for 7 days before and after the experiment to verify normal cyclicity. Cells were placed on a glass slide and analyzed with a light microscope to determine their stage of estrus cycle as described (Byers et al. 2012). All animals displayed a normal 4-to 5-day estrus cycle.

Chronic Constriction Injury (CCI) Model of Neuropathic Pain
CCI to the sciatic nerve of the left hind leg in rats was performed under general anesthesia using the method described by Bennett et al. (Bennett and Xie 1988). Briefly, rats (weighing 200-250 g at the time of surgery) were anesthetized with 3% isoflurane/O 2 inhalation and maintained on 2% isoflurane/O 2 for the duration of surgery. The left thigh was shaved and scrubbed with dermachlor solution (0.2% chlorhexidine gluconate), and a small incision (1-1.5 cm in length) was made in the middle of the lateral aspect of the thigh to expose the sciatic nerve. The nerve was loosely ligated at 3 sites spaced 1 mm apart using 4.0 silk sutures. The surgical site was closed with a skin clip.

Fig. 2
Deficits in adenosine signaling at the A 3 AR contribute to the pronociceptive actions of S1PR1 activation. In rats with established neuropathic pain, mechano-allodynia was reversed by an acute oral administration on day 7 post-CCI of selective S1PR1 antagonists TASP0277308 (a, 3 mg/kg; n = 3) and FTY720 (b, 3 mg/kg; n = 6). An i.p. injection of the A 3 AR antagonist MSR1523 (5 µmol/kg) blocked the anti-allodynic effects of TASP0277308 (a, n = 3) and FTY720 (b, n = 6). The anti-allodynic effects of morphine were not affected by MRS1523 (c, 3 mg/kg; n = 4). Furthermore, i.th. injec-tion of MRS1523 (3 nmol; n = 5) blocked the anti-allodynic effects of FTY720 given systemically (orally) (d, n = 5). Results are expressed as mean ± SD for n animals and analyzed by two-way repeated measures ANOVA with Dunnett's comparisons. *p < 0.05 vs day 0 and † p < 0.05 versus day 7 = (a-d). MRS1523 tested in antagonist mode with The PathHunter® β-Arrestin assay vs control dose response curve (S1P) for the selected GPCR Biosensor Assay (e). Data represent two point determinations at each concentration tested (e)

Behavioral Testing
Mechanical allodynia as a readout was assessed as the hind paw withdrawal response to von Frey hair stimulation using the up-and-down method (Dixon 1980). Briefly, the animals were first acclimatized (30 min) in individual clear Plexiglas boxes on an elevated wire mesh platform to facilitate access to the plantar surface of the hind paws. Subsequently, a series of von Frey hairs (1.4,2,4,6,8,10,15,26 g;Stoelting,Wood Dale,IL USA) were applied perpendicular to the plantar surface of the hind paw, until the filament buckles, for 2-5 s. A test began with the application of the 4 g for rats. A positive response was defined as clear paw withdrawal or shaking. In the event of a positive response, the next lighter hair was applied, whereas the next higher hair was applied in the event of a negative response. At least three readings are obtained after the first positive response and the pattern of response was converted to a 50% paw withdrawal threshold (PWT), using the method described by Chaplan et al. (1994). Mechano-allodynia was defined as a significant (2 SD) decrease in the measured behavior compared to the individual animal's baseline behavior prior to mechanical or pharmacological pain induction.
Chemiluminescence antibody signals were documented using Chemidoc XRS + image system and ImageLab software (BioRad) using exposure settings and quantified for band densitometry. Each membrane was then probed for β-actin (1:5000, Sigma-Aldrich, St. Louis, MO, #A5441) as endogenous loading controls. The antibody used to probe for NLRP3, IL-1β, caspase 1 (p20), adenosine kinase detect the same bands as reported in a previous publications Wahlman et al. 2018). Relative protein expression was quantified by measuring band densitometry. Images for the blots of a protein of interest and their corresponding β-actin were selected for analyses and presentation based on a predetermined upper grayscale value (45,000-65,000 units) to assure linear densitometric values. Images were analyzed using the lane and band functions of the ImageLab™ software. For presentation purposes, the grayscale range was set between 0 units and the predetermined upper grayscale value (45,000-65,000 units) for the protein of interest and corresponding β-actin prior to exporting as an image file. Post-export modifications of the images were limited to cropping to the regions of interest.

Immunofluorescence
Immunofluorescence was performed using modifications of previously reported methods (Wahlman et al. 2018). After behavioral measurements, rats were perfused with 4%PFA and the lower lumbar enlargement of the spinal cord (L4-L6) was harvested. Tissue was fixed for 30 min with 4% PFA in 1X PBS, washed with PBS and cryoprotected in 30% sucrose before embedding in OCT, and frozen in 2-methylbutane and liquid nitrogen bath. Transverse sections (20 µm) were cut using a cryostat and collected on glass microscope slides. Spinal cord sections were fixed in 4% PFA (10 min), blocked (5% donkey serum, 0,5% bovine serum albumin in 1X PBS, for 1 h), and then immunolabeled using an 18-h incubation (4 °C) with rabbit polyclonal anti-ADK (1:100; Thermo Fisher Scientific, Waltham, Massachusetts, #PA5-27399) and antiglial fibrillary acidic protein (GFAP) (1:1000; Sigma-Aldrich, St. Louis, MO, #G3893). After a series of PBS rinses, sections were incubated for 2 h with donkey antirabbit Alexa Fluor 594 conjugated (1:250; Jackson ImmunoResearch, #711-585-152) and antimouse Alexa Fluor 488 conjugated antibody (1:250; Jackson ImmunoResearch, #715-545-151). The coverslips were mounted with Prolong Gold antifade reagent containing DAPI [4,6-Diamidino-2-phenylindole, dihydrochloride] to label nuclei (Electron Microscopy Sciences, Hatfield, PA). Immunofluorescence images were captured on a Leica TCS SP8 Confocal microscope (Leica Microsystems, Exton, PA) using a 20 × lens (0.75 NA) using the same laser power and detector gain for all animals. 8 bit images were collected at 1024 × 768 resolution and analyzed using the SVI Huygens Object Analysis module (SVI, Netherlands). GFAP and adenosine antibodies were used in a previous publication (Wahlman et al. 2018). GFAP signal in hand selected areas representing lamina 1 and 2 was used to create a ROI for measuring intensity and counting numbers of voxels corresponding to a volume of ADK encompassing the GFAP signal + 1 µm of surrounding area to approximate astrocytes. One animal from the SEW group was excluded as a clear outlier based on ADK signal out of observed ranges seen for either treatment group. Example images produced using Surface Visualization module of SVI Huygens with the same thresholding applied to both images.

Ultra-High-Performance Liquid Chromatograph/ High-Resolution Mass Spectrometry (UHPLC-MS/ HRMS) Analysis for Adenosine and Inosine
All solvents were of liquid chromatography-mass spectrometry (LC-MS) or better quality. Water with 0.1% formic acid (FA) and methanol (MeOH) with 0.1% FA were purchased from Honeywell Chromasolv™ (Charlotte, NC, USA). Adenosine standard was purchased from ACROS Organics (New Jersey, USA). Inosine standard was purchased from Millipore Sigma (Darmstadt, Germany). Adenosine-13 C 5 was purchased from Toronto Research Chemicals (Toronto, ON, Canada). Inosine-15 N 4 was purchased from Cambridge Isotope Laboratories (Tewkesbury, MA, USA).
Combined adenosine and inosine stock solution was prepared in 17:3 H 2 O:MeOH + 0.1% FA at 1 mM from 5 mM individual stocks in same. Calibration standards were prepared in same from 50 nM to 100 μM in half-order of magnitude increments via serial dilution from combined stock (e.g. 50 nM, 100 nM, 500 nM, 1 μM, etc.), as were quality control (QC) standards at 2.5 and 25 μM.
All samples were injected as extracted by third party (PCA protein precipitation and KOH neutralization), with the addition of IS 3:17 (IS:sample extract).
All analyses were performed on a Dionex UltiMate™ 3000 UHPLC coupled to a Thermo Q Exactive™ Plus Hybrid Quadrupole-Orbitrap™ MS.
Column used was a Phenomenex Kinetex PFP, 1.7 μm, 2.1 × 150 mm. Mobile phase A was 0.1% FA in H 2 O; mobile phase B was 0.1% FA in MeOH. Isocratic elution at 15% B was accomplished in 2 min with a 500 μL/min flow rate and a column temperature of 45 °C. Injection volume was 2 μL.
An aliquot of biological fluid, cell homogenate, or tissue homogenate was quenched with equal volume of 1 M PCA and spiked with a mixture of heavy isotope-labeled, nucleotide internal standards followed by the addition of 1 M ammonium formate. Samples were vortexed and centrifuged at 18,000×g for 5 min at 10 °C. and then filtered through an AcroPrep Advance 3 K Omega Filter Plate (Pall Corporation) by centrifugation at 3500×g for 60 min prior to LC-MS/MS analysis. Metabolites were separated on a 2.1 × 50 mm, 3 μm Thermo Scientific Hypercarb column (T = 30 °C) using a Dionex Ultimate 3000 UHPLC. The step gradient was 98% A (10 mM ammonium acetate, pH 9.5) and 2% B (ACN) to 64% A and 36% B over 6.3 min. The step gradient began at 2% B (0.6 mL/min flow rate) from 0 to 0.45 min, was increased from 2 to 36% B (0.6 mL/min flow rate) from 0.45 to 6.3 min, was increased from 36 to 95% B (0.8 mL/min flow rate) from 6.3 to 6.4 min, and was held until 8.4 min. Re-equilibration was performed at 2% B from 8.4 to 8.5 min (0.7 mL/min flow rate) and was held until 11.5 min. The flow returned to 0.6 mL/min at 11.6 min and was held until 11.7 min. Quantitation of pyridine nucleotides was achieved using single reaction monitoring (SRM) on a Thermo Scientific Quantiva triple quadrupole mass spectrometer (Thermo Scientific). The mass spectrometer was operated in positive ion mode using electrospray ionization with an ESI capillary voltage of 3500 V. The ion transfer tube temperature was 350 °C and vaporizer temperature was 350 °C. The ESI source sheath gas was set to 40, the auxiliary gas was set to 10, and the sweep gas was set to 1. The mass spectrometer was operated with a mass resolution of 0.7 Da, a cycle time of 0.3 s, and nitrogen collision gas of 1.5 mTorr for the generation and detection of product ions of each nucleotide. Collision energies to produce product ions ranged from 16 to 46 V with RF lens values ranging from 43 to 85 V.

Statistical Analysis
Data are expressed as mean ± standard deviation (SD). No differences were observed between male and female rats behavior (p < 0.05) and therefore, where indicated, data were expressed as a single group. Justification for animal numbers was consistent with NIH policy (NOT-OD-15-102), and experiments were randomized to blinded treatment groups, giving 80% power to detect a treatment effect size of 20% compared to a baseline response of 5% at a significance level of 0.05 (Andrews et al. 2016). Behavioral data were analyzed by student's t test, one-way, two-way, or repeated measures ANOVA with Dunnett's or Tukey's pair-wise comparisons. Protein expression data were determined by two-tailed, one-way ANOVA with Tukey's comparisons. All statistical analyses were performed using GraphPad Prism version 9.0 (Graph Pad Inc., San Diego, CA).