Animal use was approved by the institutional care and animal use committee (IACUC) of the Mayo Clinic and conducted in accordance with the public health policy on the humane use and care of laboratory animals. C57BL/6J male mice were purchased from Jackson laboratory (Bar Harbor, Maine) at 6 weeks of age. When 12 weeks old, mice were randomized to receive either a diet high in saturated fat, fructose, and cholesterol (FFC, AIN-76A Western Diet, Test Diets 5342) or standard rodent chow diet (CD) for 24 weeks. As previously described by us, at 24 weeks of feeding, this diet recapitulates cardinal features of human NASH including histologic parameters of steatosis, inflammation, ballooning, fibrosis and metabolic parameters of obesity and insulin resistance (14). Mice had unrestricted access to food and water, and were housed in standard pathogen-free facilities, with 12-hour day-night circadian cycles. Twenty weeks into the feeding study mice started receiving drug or vehicle by daily oral gavage for 1 month. The drugs employed were Etrasimod (Cayman Chemical, 21661) or Amiselimod hydrochloride (Cayman Chemical, 20970). Etrasimod was dissolved at 20 mg/mL in DMSO and then diluted to 1.44 mg/mL in 0.5% sodium methylcellulose and administered to mice at a dose of 3 mg/Kg. Amiselimod was dissolved in DMSO at a concentration of 20 mg/mL and then diluted to 1 mg/mL in 0.5% sodium methylcellulose and administered to mice at a dose of 2 mg/Kg. A matched volume of DMSO, similarly diluted in 0.5% sodium methylcellulose, served as the vehicle. Each treatment group (Etrasimod or Amiselimod) had its own corresponding vehicle cohort. Upon completion of the study, liver and plasma were collected after euthanasia. Prior to liver excision, 10 mL PBS was flushed through the liver via inferior vena cava (IVC) cannulation in a retrograde fashion. Then the liver was excised, weighed, and divided for downstream analyses. Approximately 1g of liver tissue per mouse was processed for flow cytometry, described below, and the rest divided for several assays including: cryopreservation (snap frozen in liquid nitrogen) for RNA extraction and protein extraction, fixation in 10% neutral buffered formalin for paraffin embedding, and cryo-sectioning for histological analysis. Platelet poor plasma was isolated from citrated blood (described below) and stored at -20°CC till further analyses. Archived flash frozen, in liquid nitrogen, liver tissue of C57BL/6J wild type mice fed chow or FFC diet for 24 weeks were employed for RNA isolation and quantification of whole liver S1P_1-5 expression (15).

Bone marrow-derived macrophages (BMDM) were isolated from the hind legs of untreated wild type C57BL/6 mice as previously described by us (10). BMDM media was changed every 2 days and on day 7 differentiated macrophages were dissociated with Accutase (Invitrogen, 00-4555-56) for use in experiments. After dissociation, BMDMs were plated in two 100 mm petri dishes and were allowed to attach for 16 hours. Next, BMDMs were serum starved for 2 hours after which they were treated with palmitate (400 µM) complexed to 1% bovine serum albumin and volume-matched vehicle (isopropanol) for 8 hours as described by us (16). Cells were lysed in TRIzol reagent (Invitrogen, 15596026) and lysates were stored at -80°C for RNA extraction. Primary mouse hepatocytes (PMH) were isolated based on standardized techniques as described previously (17). PMHs were treated with palmitate (400 µM) and volume-matched vehicle (isopropanol) for 8 hours and collected in TRIzol reagent and stored at -80°C for RNA extraction.

Intraperitoneal glucose tolerance test (GTT) was performed on mice in both CD and FFC-fed cohorts at 22 weeks on diet. Mice were fasted overnight for 16 hours and injected with 20% glucose intraperitoneally at a dose of 2 mg/Kg body mass. Blood glucose values were acquired before injection and 15, 30, 45, 60, 90, 120, and 150 minutes after injection. Tail vein blood sampling was done using AssurePlatinum blood glucose meter and test strips (Arkray, 67841996). The area under the glucose tolerance curve was calculated in Microsoft Excel and glucose disposal capacity was compared between groups.

Five μm sections of formalin-fixed, paraffin-embedded (FFPE) liver tissues were stained with hematoxylin and eosin (H&E) using standard techniques at the Mayo Clinic Histology Core. Histology was observed and H&E slides were graded for steatosis and inflammation using the NAFLD activity score. Immunohistochemistry for Mac-2/Galectin-3 (Invitrogen, 14-5301-85) and Ki67 (Abcam, ab16667) was performed to assess macrophage accumulation and hepatocyte proliferation respectively. Five μm FFPE liver sections were dewaxed with xylene and rehydrated through graded 100%, 95%, 70%, and 50% ethanol sequentially. Slides were then immersed in 10 mM sodium citrate buffer with 0.05% Tween-20 (pH 6.0) for 20 minutes at 95°C for epitope retrieval. The slides were allowed to cool on the benchtop to room temperature. After cooling, endogenous peroxidase activity was blocked by incubating the slides in 3% hydrogen peroxide for 10 minutes. For Mac-2 staining, non-specific immunoglobulin binding was blocked using the Rat-IgG ABC Kit (Vector Laboratories, PK 4004) time for 30 minutes, followed by the Rodent Block M solution (Biocare Medical, RBM 961). For Ki67 blocking was performed using Rabbit-IgG ABC Kit (Vector Laboratories, PK 4001) for 30 minutes. All slides were further blocked with avidin and biotin blocking solution for 15 minutes each using the Avidin/Biotin Blocking Kit (Vector Laboratories, SP 2001). Primary antibodies for Mac-2 (1:200 dilution) and Ki67 (1:200 dilution) were applied and incubated overnight at 4°C in humidified chambers. Slides were washed and incubated for 30 minutes each in biotinylated secondary antibody (1:500 dilution) and ABC reagent from the ABC Kit per manufacturer’s instructions. Signals were visualized using the ImmPACT DAB Peroxidase Substrate Kit (Vector Laboratories, SK 4105), and counterstained with Nuclear Fast Red Solution for Mac-2 (MilliporeSigma, N 8002) for 20 minutes and with Hematoxylin for Ki67 for 30 seconds. Sections were dehydrated through graded ethanol and mounted using Organo/Limonene Mount (Santa Cruz, SC-45087). The positive areas were quantified using the NIS-Elements imaging software (Nikon) on a Nikon microscope with a DS-U3 camera (Nikon, Japan) on the 20x objective lens. Images with preserved settings for light and exposure were used during image analysis and area quantification for Mac-2. Ten high power fields (20x objective lens) were manually counted for nuclear positivity for quantification of Ki67 signal and expressed as average positive nuclei per high power field. Immunohistochemistry for hepatocyte nuclear factor 4 alpha (HNF4α) (Abcam, ab201460) was performed to assess hepatic differentiation as above except for antigen retrieval, which was performed by immersing slides in Tris-EDTA Buffer with 0.05% Tween-20 (pH 9.0) for 30 minutes at 95°C. Non-specific immunoglobulin binding was blocked using the Rabbit-IgG ABC Kit for 1 hour. Primary antibody for HNF4α (1:250 dilution) was applied and incubated overnight at 4°C in humidified chambers. Representative images were captures with the 10x objective lens.

Mouse intrahepatic leukocytes were isolated according to the published method of Blom et al. (18). Following euthanasia mouse livers were perfused with 10 mL PBS in a retrograde manner via IVC. The perfused livers were dissected out of the abdomen and the gall bladder was removed. Next, the livers were disrupted using a gentleMACS tissue dissociator in 5 mL RPMI-1640 per manufacturer’s recommended preset program for 45 seconds. The dissociate was transferred to a 50 mL conical tube and RPMI-1640 was added to a volume of 40 mL. This suspension was centrifuged at 60 x g for 1 minute at room temperature and the supernatant containing hepatic immune cells was collected and centrifuged again at 480 x g for 8 minutes at room temperature. The pellet was resuspended in 10 mL of 37.5% Percoll (MilliporeSigma, P1644-1L) and centrifuged at 850 x g for 30 minutes at room temperature with no brake. The supernatant was aspirated, and the pellet was resuspended in 5 mL of red blood cell lysis buffer for 5 minutes on ice. Lysis buffer was neutralized by adding 10 mL PBS, and the remaining cells were pelleted down by centrifugation. The cells were resuspended in 200 μL of PBS and counted using the Muse® Cell Analyzer (MilliporeSigma). Approximately 1 million cells were transferred to FACS tubes and stained with 1:200 zombie aqua (Biolegend) to discriminate between viable and dead cells. Cells were washed with cell staining buffer (Biolegend) before blocking non-specific binding with TruStain FcX Plus (Biolegend). Staining for cell surface markers was performed with fluorophore-conjugated primary antibodies; CD45.2 (AF700, clone 104, Biolegend), CD3 (BUV737, clone 145-2C11, BD Bioscience), NK1.1 (APC, clone PK136, Biolegend), CD4 (BUV737, clone RM4-5, BD Bioscience), CD8a (BUV805, clone 53-6.7, BD Bioscience), B220 (BV421, clone RA3-6B2, Biolegend), CD19 (BV785, clone 6D5, Biolegend), CD11c (BV711, clone N418, Biolegend), CD11b (PE/Cy7, clone M1/70, Biolegend), and Ly6G (PE, clone 1A8, Biolegend) for 30 minutes at 4°C. Flow cytometry data was acquired on BD Fortessa X-20 and Fortessa X-30 H0081 (BD Bioscience) cytometers and analyzed using FlowJo™ Software version 10.6.1 (Becton, Dickinson & Company).

Cryopreserved mouse liver tissues of vehicle and treatment groups for both diets were recovered and homogenized in TRIzol reagent. Tissue debris was centrifuged down and the supernatant was transferred into separate Eppendorf tubes. Total RNA was extracted using the Direct-zol RNA Zymo Plus Mini Prep (Zymo Research, R 2070), then the RNA yield and quality was assessed using a NanoDrop ND1000 (ThermoScientific, Waltham, MA). 1 μg of total RNA was reverse transcribed into cDNA with the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, 1708891). Gene expression was analyzed using quantitative real time PCR (qPCR) with gene specific mouse primers ( Table 1 ). qPCR was performed using the LightCycler 480 SYBR Green I Master Mix (Roche Diagnostics, 04707516001)), and run on LightCycler 480 (Roche). Each dataset of the target gene of interest was normalized to the geometric mean of the reference genes Hprt and 18S. All datasets from drug treated mice are expressed as fold change relative to vehicle controls in the CD group. For isolated cells, the data represent fold change compared to vehicle controls.

Platelet poor plasma was isolated by first centrifuging 500-800 μL of citrated blood for 20 minutes at 4,000 x g. The supernatant was transferred to a new Eppendorf tube and spun down at 12,000 rpm for 2 minutes. The clear supernatant was then aliquoted to new tubes for analysis or storage. Approximately 100 μL of plasma was used for analysis of alanine aminotransferase (ALT) levels using VetScan VS2 (Abaxis, Union City, CA).

Graphs and statistical analyses were done using GraphPad Prism version 9 for Windows (GraphPad Software, La Jolla, CA; www.graphpad.com). Data are presented as mean ± S.E.M in each graph. Each dot in the dot plot represents one mouse or biological replicate. The Grubb’s test (extreme studentized deviant method) was used to identify outliers to determine if the most extreme value in a data set is a significant outlier from the rest. For statistical comparisons of multiple experimental groups against the control group, parametric data were analyzed by ANOVA and the Tukey’s Multiple Comparison method was employed to determine which pairs of means are different. A p-value of < 0.05 was considered significant.

A dietary NASH mouse model was generated by feeding mice a FFC diet for 24 weeks. Next, both CD (control) and FFC-fed groups were randomized to receive Etrasimod (3 mg/Kg) or vehicle, or Amiselimod (2 mg/Kg) or vehicle treatment based on effective doses reported in the literature ( Figure 1A ) (12, 13). The liver histology from FFC-fed mice showed a significant increase in steatosis and inflammation compared to those mice that received CD ( Figure 1B ). Steatosis and inflammatory foci were quantified using the NAFLD activity score, which showed that Etrasimod treatment did not reduce steatosis in FFC-fed mice ( Figure 1C ). Similarly, Amiselimod treatment did not impact steatosis ( Supplementary Figures 1A, B ). On the other hand, Etrasimod treatment significantly reduced inflammatory foci on histological examination of the FFC group ( Figure 1D ). Correspondingly, the ALT level was reduced with Etrasimod treatment in the FFC-fed group ( Figure 1E ), demonstrating improvement in liver inflammation and injury in FFC-fed mice. In contrast, Amiselimod treatment did not reduce inflammatory foci though lowered ALT levels ( Supplementary Figures 1C, D ).

CD-fed and FFC-fed mice maintained comparable body mass, liver mass and relative liver mass ( Figures 2A–C , Supplementary Figures 2A–C ), regardless of Etrasimod or Amiselimod treatment, confirming that the improvement in liver injury and inflammation were not due to do a lack of weight gain. Etrasimod administration did not lower the FFC diet-induced increase of cholesterol ( Figure 2D ). There was a trend toward increase in fasting blood sugar in vehicle-treated FFC-fed mice; however, levels were similar to Etrasimod-treated FFC-fed mice ( Figure 2E ). Neither FFC-diet induced hypercholesterolemia nor elevated fasting blood sugars were lowered following Amiselimod administration ( Supplementary Figures 2D, E ). FFC-fed Etrasimod and Amiselimod treated mice demonstrated comparable glucose disposal to vehicle treated mice following intraperitoneal glucose challenge ( Figure 2F , Supplementary Figure 2F ).

We next examined the changes in intrahepatic leukocyte populations in FFC-fed mice treated with Etrasimod or Amiselimod by flow cytometry of the intrahepatic leukocyte populations. After isolation, viable cells were selected for further analysis ( Supplementary Figure 3 ). Our flow cytometry panel was designed to characterize the major immune cell populations within the liver ( Figure 3 ). As expected, the FFC-diet led to an increase in the number of total CD45⁺ cells ( Figure 4A ), T cells, and NKT cells ( Figures 4B, C ) in the liver. Etrasimod treatment led to a reduction in the frequency of T and B cells in the liver of CD- and FFC-fed mice ( Figures 4B, C ). CD3⁺ NK1.1⁺ NKT cells were also reduced in FFC-fed mice ( Figures 4B, C ), though they remained unchanged in CD-fed mice. Further analysis of T-cell subsets demonstrated no change in CD4⁺ T cells in FFC-fed mice compared to CD-fed mice, though CD4⁺ T cells were reduced by Etrasimod in both CD-fed and FFC-fed mice ( Figure 4D ). CD8⁺ T cells were increased in FFC-fed mice in comparison with CD-fed mice and were decreased following Etrasimod administration. We observed an increase in double negative (DN) T cells in FFC-fed mice which were unchanged following Etrasimod administration in FFC-fed mice, though reduced by Etrasimod in CD-fed mice. Etrasimod increased the frequency and percentage of NK cells in the liver in CD-fed mice, but not in FFC-fed mice ( Figures 4E, F ). The frequency and percentage of CD11b⁺ cells, which includes myeloid cells and polymorphonuclear cells (PMNs), were increased in both CD-fed and FFC-fed Etrasimod-treated mice. In contrast, the total number of CD45⁺ cells in the liver were unaffected in either CD-fed or FFC-fed mice treated with Amiselimod ( Supplementary Figure 4A ). Intrahepatic leukocyte populations in Amiselimod treated CD-fed mice had changes similar to Etrasimod with a reduction in T cells and B cells and an increase in CD11b⁺ myeloid cells; however, these effects were more modest in FFC-fed mice suggesting that, at the dose tested, Amiselimod was less effective than Etrasimod in modulating the distribution of immune cells under conditions of nutrient excess ( Supplementary Figure 4 ). Though T cells were not reduced in FFC-fed mice treated with Amiselimod, the T cell subsets were altered like Etrasimod.

NASH is characterized by intrahepatic macrophage accumulation. As our flow cytometry panel did not include markers to differentiate between infiltrating monocyte-derived and tissue-resident macrophages, we assessed their accumulation and activation using immunohistochemistry and gene expression analysis. Immunohistochemistry for Mac-2/Galectin-3 in liver sections showed that activated macrophage accumulation was visibly reduced in livers of both Etrasimod and Amiselimod treated mice ( Figure 5A , Supplementary Figure 5A ). Quantification of Mac-2/Galectin-3 positive areas confirmed significant decrease in macrophage accumulation in Etrasimod ( Figure 5B ) and Amiselimod ( Supplementary Figure 5B ) treated mice. There was also a significant reduction in relative mRNA expression of macrophage markers CD68 and Lgals3 ( Figures 5C, D ). Similar changes were observed in Ly6c, a marker of proinflammatory monocytes, though this did not achieve statistical significance ( Figure 5E ). By extension, we asked whether drug treatment had an effect on levels of proinflammatory chemokines and cytokines in the liver. Relative mRNA expression of monocyte chemotactic protein 1 (Mcp1), tumor necrosis factor alpha (Tnfa) and interleukin 1 beta (Il1b) were significantly reduced with both Etrasimod ( Figures 5F–H ) and Amiselimod ( Supplementary Figures 5F–H ) treatment in FFC mice. Thus, Etrasimod and Amiselimod treatment not only reduced inflammatory macrophage accumulation in the liver, but also reduced levels of proinflammatory cytokines that aggravate the inflammatory response. In Etrasimod treated FFC-fed livers the resident macrophage marker Timd4 ( Figure 5I ) was not upregulated, unlike Amiselimod ( Supplementary Figure 5I ). We also observed an increase in Marco, a marker of non-inflammatory macrophage populations ( Figure 5J , Supplementary Figure 5J ) and the lipid associated macrophage marker Trem2 ( Figure 5K , Supplementary Figure 5K ). These data suggest that the increase in CD11b observed by flow cytometry may be due to an increase in the abundance of resident and non-inflammatory macrophages.

To understand whether the differential effects of Etrasimod and Amiselimod may be mediated by variations in the expression of S1P receptors, the reported expression of S1P receptors in hepatocytes, bone marrow monocytes and macrophages was compared from a publically available single cell RNA sequencing data (19) (https://www.czbiohub.org/tabula-muris/), Table 2 . In this dataset S1P₄ mRNA was most abundant in macrophages and monocytes, and S1P₁ was most abundant in hepatocytes. Next, BMDM and PMH were treated with palmitate to induce lipotoxic stress (10). In BMDMs, the expression of S1P₁ and S1P₂ mRNA was induced following palmitate treatment, whereas the abundance of S1P₄ remained unchanged ( Figure 6A ). S1P₁, S1P₂, S1P₃ and S1P₅ were detected in hepatocytes and remained unchanged with palmitate treatment ( Figure 6B ). In contrast to isolated cells, analysis of the abundance of S1P receptors in whole livers demonstrated no change in S1PR₁ and S1PR₃, upregulation of S1PR₂, and downregulation of S1PR₅ in mice fed the FFC diet in comparison to CD fed controls ( Figure 6C ). As S1P₁ was upregulated in BMDM by lipotoxic stress without any induction in PMH or whole livers, this may be the primary target on macrophages for the beneficial effects of Etrasmiod and Amiselimod. Further examination of hepatocellular proliferation demonstrated low basal Ki67 positivity, consistent with the literature (20). Ki67 positivity was increased in FFC-fed vehicle treated livers, as has been reported in mouse models of NASH (21, 22) and reduced following treatment with Etrasimod and Amiselimod ( Supplementary Figure 6 ). Hepatocellular expression of HNF4α was comparably reduced in vehicle treated and Etrasimod or Amiselimod treated FFC-fed mouse livers ( Supplementary Figure 7 ).