PET Imaging of Sphingosine-1-Phosphate Receptor 1 with [18F]TZ4877 in Nonhuman Primates

Purpose The sphingosine-1-phosphate receptor-1 (S1PR1) is involved in regulating responses to neuroimmune stimuli. There is a need for S1PR1-specific radioligands with clinically suitable brain pharmcokinetic properties to complement existing radiotracers. This work evaluated a promising S1PR1 radiotracer, [18F]TZ4877, in nonhuman primates. Procedures: [18F]TZ4877 was produced via nucleophilic substitution of tosylate precursor with K[18F]/F− followed by deprotection. Brain PET imaging data were acquired with a Focus220 scanner in two Macaca mulatta (6, 13 years old) for 120–180 min following bolus injection of 118–163 MBq [18F]TZ4877, with arterial blood sampling and metabolite analysis to measure the parent input function and plasma free fraction (fP). Each animal was scanned at baseline, 15–18 min after 0.047–0.063 mg/kg of the S1PR1 inhibitor ponesimod, 33 min after 0.4–0.8 mg/kg of the S1PR1-specific compound TZ82112, and 167–195 min after 1 ng/kg of the immune stimulus endotoxin. Kinetic analysis with metabolite-corrected input function was performed to estimate the free fraction corrected total distribution volume (VT/fP). Whole-body dosimetry scans were acquired in 2 animals (1M, 1F) with a Biograph Vision PET/CT System, and absorbed radiation dose estimates were calculated with OLINDA. Results [18F]TZ4877 exhibited fast kinetics that were described by the reversible 2-tissue compartment model. Baseline [18F]TZ4877 fP was low (< 1%), and [18F]TZ4877 VT/fP values were 233–866 mL/cm3. TZ82112 dose-dependently reduced [18F]TZ4877 VT/fP, while ponesimod and endotoxin exhibited negligible effects on VT/fP, possibly due to scan timing relative to dosing. Dosimetry studies identified the critical organs of gallbladder (0.42 (M) and 0.31 (F) mSv/MBq) for anesthetized nonhuman primate. Conclusions [18F]TZ4877 exhibits reversible kinetic properties, but the low fP value limits quantification with this radiotracer. S1PR1 is a compelling PET imaging target, and these data support pursuing alternative F-18 labeled radiotracers for potential future human studies.


Introduction
Sphingosine-1-phosphate (S1P) is a sphingolipid that potently regulates brain immune responses, classically promoting cellular survival 1 .S1P affects neuroimmune function through activation with S1P receptors.The S1P system plays important roles in lymphocyte tra cking and vascular integrity 2 , and S1P modulators are approved for treating patients with multiple sclerosis and of interest for other conditions with altered brain immune function, such as amyotrophic lateral sclerosis, glioblastoma, and schizophrenia 3 .Of the ve S1P receptor subtypes, S1PR 1 is the most abundant in brain and is a G-protein coupled receptor.S1PR 1 activation shifts brain immune state towards anti-in ammatory phenotypes, promoting cell survival, motility, and proliferation 4 .S1PR 1 signaling also regulates key neuroimmune transcription factors such as nuclear factor-κB 5,6 , triggering cytokine and chemokine release and inhibiting histone deacetylases (HDAC1 and HDAC2) 7 .Indeed, inhibiting S1PR 1 attenuates proin ammatory chemokine release 8 .Interestingly, pro-in ammatory conditions functionally upregulate S1PR 1 9,10 , likely as a compensatory mechanism.Consequently, PET radiotracers that quantify S1PR 1 in vivo would provide important data regarding S1PR 1 target engagement, and potentially provide a biomarker for brain immune status.
Several PET radiotracers speci c for S1PR 1 are reported.Importantly, S1PR 1 is highly expressed in human brain, with estimated B max values of 135 ± 7 nM in post mortem frontal cortex 11 , which should be su ciently high for quantitative PET imaging given a suitably speci c radiotracer.The radiotracer [ 11 C]CS1P1 (initially [ 11 C]TZ3321; 1 in Fig. 1) exhibits elevated uptake in animal models of neuroin ammation [12][13][14] .These successes culminated in approval of [ 11 C]CS1P1 for human imaging 15 , supporting its utility as a suitable radiotracer for imaging S1PR 1 in human subjects.
Despite the current promise for [ 11 C]CS1P1, there is important rationale for complementary PET radiotracers.[ 11 C]CS1P1 does not exhibit fast kinetic properties 16 , while labeling with uorine-18, which has a 109.8min half-life, would allow for longer scan sessions that could improve quanti cation and facilitate human use at centers with no onsite cyclotron.To address this important need, several 18 Flabeled compounds that target S1PR 1 have been developed, including some based on [ 11 C]CS1P1 16,17 (see 2, Fig. 1).Of these radiotracers, [ 18 F]TZ4877 (3, Fig. 1) has an IC 50 value of 14.01±0.05nM for S1PR 1 and selectivity over other S1PR subtypes (> 100-fold over S1PR 2 − 5 18 ), and in rat is well-described by the 2-tissue compartment model (2TCM) with reproducible estimates of volume of distribution (V T ), with no evidence of radiolabeled metabolites penetrating the brain 17 .[ 18 F]TZ4877 also exhibited elevated uptake in a mouse model of bacterial infection 19 and rat model of experimental autoimmune encephalomyelitis (EAE) 17 .These data indicate that [ 18 F]TZ4877 has potential for imaging S1PR 1 in human subjects.To further examine this potential, we conducted studies with [ 18 F]TZ4877 in nonhuman primates with the goals of characterizing its pharmacological and kinetic properties and estimating organ dosimetry.

Animals
The animals for this study were four Maccaca mulatta (1 female, 11 years old, 12.0 kg; 3 males, 6-13 years old, 10.3-18.0kg).A total of 10 PET scans were conducted, including 8 brain-only acquisitions and 2 whole body acquisitions for dosimetry estimates.On the day of scans, subjects were initially sedated with combination of alphaxalone (1-2 mg/kg), midazolam (0.3 mg/kg) and dexmedetomidine (0.01 mg/kg) at least 2 h prior to radiotracer administration.Subjects were then maintained on oxygen and 1.5-3% iso urane for the duration of scans.Vital signs, SpO2, and end tidal CO WAT023501).The C18 cartridge was washed with 0.001 N HCl (10 mL) and dried with argon ow.The product was eluted off the C18 cartridge sequentially with 1 mL of USP ethanol and 3 mL of USP saline, and passed through a 0.22 µm membrane lter (GV, Millipore, Sigma) into a 10 mL dose vial pre-charged with 7 mL of USP saline and 8.4% NaHCO 3 solution (20 µL).An aliquot of the nal production solution was injected onto an analytic HPLC to determine radiochemical purity and molar activity.Tracer authentication was performed by co-injection with non-radiolabeled standard TZ4877 sample solution.
The analytic HPLC conditions were: Phenomenex Luna C18(2) column, UV absorbance at 260 nm, with mobile phase of 55% CH 3 CN in 0.1 M ammonium formate with 5% AcOH (pH = 4.2) at a ow rate of 2.0 mL/min.The retention time of [ 18 F]TZ4877 was 7.4 min.

<<< INSERT FIGURE 2 HERE >>> Brain PET Scanning Procedures
Brain-dedicated scans consisting of a baseline scan, preblocking scans with ponesimod and TZ82112, and an endotoxin challenge (see below) were acquired in each of two animals for a total of 10 brain studies.These data were acquired with a Focus 220 PET scanner (Simens/CTI, Knoxville, TN).A transmission scan was rst acquired using a continuously rotating 57 Co source for 9 min.[ 18 F]TZ4877 was administered as a 118-163 MBq slow bolus injection over 3 min with a Harvard syringe pump (PHD 22/2000, Harvard Apparatus, Holliston, MA).PET data were then continuously acquired in list-mode for 120-180 min.An arterial line for blood sampling was inserted in a radial or femoral artery on the limb opposite the tracer administration line.Discrete arterial blood samples were acquired throughout PET scanning, with rapid (45 s) sampling immediately post-injection and gradually slowing to 30 min sampling at the end of scans.

Arterial Input Function Measurement
Radioactivity assay of arterial blood samples was performed with a cross calibrated well-type gamma counter (Wizard 1480, Perkin Elmer, Waltham, MA).Whole blood samples were assayed and then centrifuged (2,930 g for 5 min).Plasma samples were then separated and assayed for radioactivity.Select plasma samples (drawn at 3, 8, 15, 30, 60, 90, 120, 180 min post-injection) were analyzed with HPLC to measure radioligand metabolism.
For metabolite analysis, plasma samples were mixed with urea to a nal concentration of 8 M and ltered through 1.0 µm Whatman 13 mm CD/X lters (GE, Florham Park, NJ).Samples were then analyzed on an adapted column-switching HPLC system 20 .Upon injection, samples were rst trapped on a C18 sorbent capture column (Strata-X, Phenomenex, Torrance CA) with a mobile phase of 1:99 (v:v) MeCN:H 2 O at 2 mL/min for 4 min.The capture column was then back ushed with a mobile phase of 45% acetonitrile and 55% 20 mM ammonium bicarbonate (v/v), and the eluent passed through a Phenomenex Gemini-NX analytical column (5 µm, 4.6 × 250 mm) at a ow rate of 1.65 mL/min.The eluent was collected with a fraction collector (CF-1 Fraction Collector, Spectrum Chromatography, Houston, TX) in discrete 2 min bins and counted in a gamma counter (Wizard 1480, Perkin Elmer, Waltham, MA).The fraction of unmetabolized parent was measured as the ratio of the eluted parent (retention time of ~ 12 min) to the total radioactivity collected.The time course of this parent fraction was tted to an inverted gamma function and corrected for ltration e ciency.Finally, the input function was calculated as the product of the assayed radioactivity concentration in plasma and the unmetabolized parent fraction.
In addition, the free fraction (f P ) was determined from plasma samples with ultra ltration techniques.An arterial blood sample drawn prior to radiotracer injection (3.0 mL) was vigorously mixed with ~ 3 kBq of [ 18 F]TZ4877.After partitioning the plasma from red blood cells via centrifugation, the plasma sample was extracted, loaded onto an ultra ltration cartridge (Millipore Centrifree UF devices), and centrifuged at 1,228 g for 20 min.The free fraction was calculated as the ratio of radioactivity in the ultra ltrate to the total radioactivity in the plasma sample.Measurements of f P were performed in triplicate for each scan.

Challenge Studies with [ 18 F]TZ4877
To evaluate [ 18 F]TZ4877 speci c binding, blocking studies were acquired with each of ponesimod, and TZ8112, a compound with high a nity (IC 50 value of 9 nM) and selectivity for S1PR 1 21 .Ponesimod (0.047 or 0.063 mg/kg) was infused 15-18 min before [ 18 F]TZ4877 injection, with PET image data acquired for 120 min.TZ8112 (0.4 or 0.8 mg/kg) was infused 33 min before [ 18 F]TZ4877 injection, with PET image data acquired for 120 min.
To evaluate the effects of an acute immune stimulus on [ 18 F]TZ4877 uptake, the classic acute immune stimulus lipopolysaccharide (LPS; NIH Clinical Center Reference E. coli serotype O:113) was used.A dose of 1 ng/kg, previously shown to trigger acute immune stimulus 22,23 , was administered 167-195 min before [ 18 F]TZ4877 injection, with PET image data acquired for 180 min.Cytokine levels for TNF-α, IL-1β, IL-6, IL-8, IL-18, IL-22, IL-28A, and MCP-1 were measured in plasma samples taken 10 minutes prior to LPS administration and 90, 180, and 300 min after LPS injection.Cytokine levels were measured in triplicate using MILLIPLEX panel assay (MillporeSigma, Burlington, MA, USA).

Brain PET Data Processing
Raw list-mode PET data was histogrammed (frames of 6 × 0.5 min; 3 × 1 min; 2 × 2 min; and N×5 min to scan termination) and reconstructed with Fourier rebinning followed by 2D ltered back projection, using a Shepp lter and including corrections for scanner normalization, detector deadtime, randoms, scatter, and attenuation.This resulted in a reconstructed image resolution of ~ 3.2 mm.The PET images were then registered to MR image space with a 6-parameter rigid body registration 24 .The MR native space was then normalized using nonlinear a ne registration to a high-resolution rhesus monkey atlas 25 using BioImage Suite 3.01 (http://www.bioimagesuite.org/index.html).Time-activity curves were extracted by mapping atlas-de ned regions to PET native space using the optimal transformation matrices calculated in the registration and normalization steps.ROIs extracted included caudate, cerebellum, frontal cortex, hippocampus, pons, putamen, temporal cortex, and thalamus.
[ 18 F]TZ4877 Kinetic Analysis For all PET scans, the primary outcome measure was the total volume of distribution 26 , both uncorrected (V T ) and corrected (V T /f P ) by the plasma free fraction.Regional V T values were estimated using both one-tissue (1TCM) and two-tissue (2TCM) compartment models (see 27 for review).Model suitability was compared with the corrected Akaike Information Criterion (cAIC; 28 ).Additionally, the multilinear analysis method (MA1; 29 ) was assessed as a more stable, data-driven analysis method.To visualize [ 18 F]TZ4877 V T images, MA1 was also used to calculate V T on the voxel level.Receptor occupancy (Occ) and nondisplaceable volume of distribution (V ND ) was estimated with occupancy plots using the following Eq. 3 0 : V T /f P (baseline) -V T /f P (block) = Occ(V T /f P (baseline) -V ND /f P )

Radiation Dosimetry Study
The 2 whole body biodistribution scans were acquired in two animals (18.0 kg Male, 12.0 kg F).These data were acquired with a Biograph Vision PET/CT system (Siemens Medical Systems, Knoxville, TN) after i.v.injection of 94 MBq and 86 MBq [ 18 F]TZ4877.Animals were imaged for approximately 3 hours in a sequence of 22 passes from top of the head to the mid-thigh.Images were reconstructed and visually inspected for organ activity concentrations exceeding background level.The organs included were heart, bladder, testes, kidney, liver, and gallbladder.Regions of interest were hand-delineated on these organs to compute mean time activity curves.
Within-pass decay correction was removed to re ect the actual activity in each organ, and the tail portions of each curve beyond the end of the scan were extrapolated assuming only physical decay of the radiotracer.The cumulative activity (Bq⋅h/cm 3 ) was computed by integrating these data.These values were multiplied by the organ volumes of standard 60 kg adult female and 73 kg adult male reference mathematical phantoms 31 and normalized to injected activity to obtain organ residence times.
Residence times were then entered into OLINDA/EXM 2.0 software to compute absorbed doses in all organs 32 , which were computed without voiding.
[ 18 F]TZ4877 in Arterial Plasma Analysis of radiolabeled [ 18 F]TZ4877 metabolites in arterial plasma was performed with HPLC.

<<< INSERT FIGURE 3 HERE >>> [ 18 F]TZ4877 Brain Tissue Kinetics
Uptake of [ 18 F]TZ4877 in the brain occurred rapidly, peaking at SUVs of 0.6-1.3within 20 min postinjection and followed by clearance from the brain (see Fig. 4).Higher uptake was observed in thalamus and putamen, moderate uptake was observed in neocortical brain regions, and low uptake was observed in hippocampus and cerebellum (see Fig. 5).

<<< INSERT FIGURE 4 HERE >>>
Compartment modeling of [ 18 F]TZ4877 TACs with the 1TCM produced visually poor curve ts, while use of the 2TCM (reversible uptake) produced visually good curve ts but did exhibited high (> 100%) standard error in estimates of [ 18 F]TZ4877 V T .Therefore, estimation of [ 18 F]TZ4877 V T was performed with the MA1 method using t * =30 min, which yielded regional baseline V T values of 2.0-4.8mL/cm 3 (see Table 1).For regions with stable 2TCM estimates, K 1 values ranged from 0.07 to 0.15 mL•cm − 3 •min − 1 .
LPS Challenge Studies with [ 18 F]TZ4877 Administration of the classic immune stimulus LPS increased [ 18 F]TZ4877 V T in both animals by 26±11% and 7±3%.However, f P appeared relatively unaffected, and changes in [ 18 F]TZ4877 V T /f P were variable (32±10% and − 7±4%).MCP-1 concentrations increased in both animals (8.6-9.7 fold) at the time of [ 18 F]TZ4877 scan start (see Supplementary Table 1). in addition, both TNF-α and IL-6 increased relative to baseline for M1, while M2 results were di cult to interpret as baseline cytokine levels were near or below detection thresholds.

Radiation Dosimetry Biodistribution Estimates
Whole-body distribution studies were performed in two rhesus macaques and revealed highest uptake of [ 18 F]TZ4877 in the gallbladder and liver.Organ residence times are presented in Supplementary Table 2, with the liver yielding the highest mean integrated time-activity curve (0.91-1.0 h; 35-38% of injected dose).Absorbed doses estimated for the male and female phantom are presented in Supplementary Table The organs receiving the largest doses were the gallbladder (0.31 mSv/MBq and 0.17 mSv/MBq for F and M, respectively) and the liver (0.16 mSv/MBq and 0.12 mSv/MBq for F and M, respectively).
Based on the gallbladder as the critical organ, the maximum permissible single-study dosage of

Discussion
radiotracers targeting S1PR 1 could provide important tools to evaluate target engagement with S1PR 1 modulators and evaluate this receptor in conditions featuring brain immune changes.The radiotracer [ 11 C]CS1P1 is approved for human use 15 , and an F-18 labeled version has been characterized in nonhuman primates 16 .These radiotracers exhibit high brain uptake with no radiolabeled metabolites, however, they also exhibit relatively slower tracer kinetics in brain and do not achieve transient equilibrium within two-hours post-injection 16 .These properties motivate the development of complementary radiotracers with faster kinetics to facilitate improved quanti cation.[ 18 F]TZ4877 exhibits rapid brain uptake followed by washout, consistent with fast radiotracer kinetics.Indeed, TZ4877 is roughly an order of magnitude less potent than CS1P1 (IC 50 = 14.0 nM vs 2.0 nM 18 ) and exhibits a much faster reduction in parent fraction, which are both characteristics that likely contribute to the faster kinetic properties.
In addition to its fast kinetics, binding of [ 18 F]TZ4877 was displaced by the S1PR 1 -speci c ligand TZ82112.However, quanti cation of [ 18 F]TZ4877 blockade required incorporation of the plasma free fraction (f P ) term, making [ 18 F]TZ4877 V T /f P the primary outcome measure.[ 18 F]TZ4877 f P is very low (< 1%) at baseline conditions, which makes it challenging to measure [ 18 F]TZ4877 V T /f P with good precision.Alternative radiotracers with higher f P values and/or V T as the outcome measure would therefore be preferable for quanti cation of S1PR 1 in human subjects.Interestingly, the S1PR 1 inhibitor ponesimod did not effectively compete with [ 18 F]TZ4877 for binding to S1PR 1 when infused ~ 15 min prior to injection.This may be due to its poor brain penetration 33 , was ponesimod has an estimated logP of 4-4.5, or insu cient dosing, as administration of the targeted dose (0.1 mg/kg) was halted due to rapid drop in animal heart rate.Ponesimod was nonetheless selected for blocking because it presents a favorable safety pro le compared with other S1PR 1 inhibitors 34 and was previously shown to displace other S1PR 1 radiotracers 35 .Thus, performing PET imaging later after ponesimod dosing, or using a compound with better brain penetration (e.g., siponimod 36 , ngolimod 37 ), could provide improved evaluation of S1PR 1 blocking in future evaluation of S1PR 1 radiotracers, but must be balanced with animal safety concerns.
Administration of the classic immune stimulus endotoxin did not produce noticeable effects on [ 18 F]TZ4877 V T /f P , although [ 18 F]TZ4877 V T did increase (with variable magnitude) in both animals.
Since [ 18 F]TZ4877 V T /f P is required to observe displacement, we conclude that this radiotracer was not sensitive to endotoxin effects at this dose (1 ng/kg) and timing (~ 3 hours pre-injection).This dosing plan elicits robust increases in levels of the 18-kDa translocator protein (TSPO), as measured with [ 11 C]PBR28 V T , in both rhesus monkey 22 and humans 38 .The presented data indicate a robust MCP-1 response to 1 ng/kg endotoxin after 3 h, although other classic cytokines (e.g., IL-8) were more variable than previously reported 23 .Indeed, Previously published in vitro data demonstrate increased S1PR 1 radiotracer binding in brain 39 and liver 40 roughly 24 hours after 15 mg/kg LPS in mice, and S1PR 1 reporter mice exhibit enhanced activation of S1PR 1 after LPS stimulation 41 .Notably, LPS effects of dose and timing vary signi cantly across species and biological target 42 , thus it is possible that the timing of maximal S1PR 1 effect to this LPS dose could be optimized for rhesus monkey.
Dosimetry studies identi ed ED estimates of 0.023 mSv/MBq for female and 0.020 mSv/MBq for male, with the gallbladder as the critical organ.Based on the estimated gallbladder doses, the maximum single-study dose would be 118 MBq to remain compliant with a 50 mSv organ maximum.This dose would allow for a maximum of 3 studies of 118 MBq [ 18 F]TZ4877 each to not exceed an annual organ dose of 150 mSv.However, higher radiotracer residence times in digestion-related organs, including gallbladder, for anesthetized nonhuman primates relative to humans have been reported previously, likely due to decreased gastrointestinal motility from anesthesia 43 .Indeed, human dosimetry studies with [ 11 C]CS1P1 reported the liver as the critical organ 15 , and we speculate the liver may be the critical organ for [ 18 F]TZ4877 in non-anesthetized humans.
To summarize, we present an evaluation of [ 18 F]TZ4877 imaging properties in non-human primates.This radiotracer exhibits fast and reversible kinetic properties, but requires measurement of f P for full quanti cation.Additionally, dosimetry estimates from anesthetized nonhuman primates would limit studies to injections of 118 MBq [ 18 F]TZ4877, although anesthesia and translational differences may result in a different pro le for non-anesthetized humans.Taken together, although [ 18 F]TZ4877 provides a useful imaging tool for quanti cation of S1PR 1 in preclinical models, alternative F-18 radiotracers 16,17 may provide more suitable imaging properties and are of high interest for imaging S1PR 1 in potential future human studies.

[ 18 F
]TZ4877 to remain below the 21 CFR 361.1 dose limit would be 160 MBq (4.31 mCi) for female and 118 MBq (3.20 mCi) for male.The estimated effective dose (ED) is 0.023 mSv/MBq for female and 0.020 mSv/MBq for male.Based on these estimates and assuming direct translation of anesthetized rhesus macaque dosimetry to humans, a single study administering a conservative 115 MBq [ 18 F]TZ4877 would result in an effective dose of 2.6 mSv, with a gallbladder dose of 47.8 mSv.

Figure 4 [
Figure 4 2were continuously monitored and recorded, including respiration rate, blood pressure, heart rate, and temperature.All experiments followed institutional guidelines and were approved by the Yale University Institutional Animal Care and