The combination of zalfermin and semaglutide has additive therapeutic effects in a diet-induced obese and biopsy-confirmed mouse model of MASH

The Danish Animal Experiments Inspectorate approved all experiments (license #2013-15-2934-00784). All animal experiments conducted were approved by the internal Gubra Animal Welfare Body, were in full compliance with internationally accepted principles for the care and use of laboratory animals, and conformed to the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines.

Male C57BL/6J mice (5–6 weeks old) were from Janvier Labs (Le Genest Saint Isle, France) and housed in a controlled environment (12 h light/dark cycle, lights on at 3 a.m., 21 ± 2°C, humidity 50 ± 10%). Each animal was identified by an implantable subcutaneous microchip (PetID Microchip, E-vet, Haderslev, Denmark). Mice had ad libitum access to tap water and chow (3.22 kcal/g, Altromin 1324; Brogaarden, Hoersholm, Denmark) or AMLN diet (4.5 kcal/g; 40 kcal-% fat, of these 22% trans-fat and 26% saturated fatty acids by weight; 22% fructose, 10% sucrose, 2% cholesterol; D09100301, Research Diets, New Brunswick, NJ). Mice were fed chow or AMLN diet for 36 weeks before initiation of treatment (Fig 1A). Animals underwent liver biopsy for assessment of baseline histopathology, as described in detail previously [21]. Animals were single-housed after the operation and allowed to recover for 3 weeks before the start of treatment. Only AMLN DIO-MASH mice with moderate-to-severe steatosis (score ≥2) and fibrosis (stage ≥F1) were included and evaluated using standard clinical biopsy histopathological scoring criteria [2].

AMLN DIO-MASH mice were randomized and stratified to treatment (n = 11–12 per group) based on baseline quantitative alpha-1 type I collagen (Col1a1) staining (primary factor; proportionate [%] area, see below) and body weight (secondary factor). Chow-fed mice served as normal controls (n = 10). AMLN DIO-MASH mice were dosed (subcutaneously [SC], daily [QD]) for 8 weeks with vehicle, zalfermin 0.05 mg/kg (low-dose zalfermin), zalfermin 0.2 mg/kg (high-dose zalfermin), semaglutide 3 µg/kg (low-dose semaglutide), semaglutide 120 µg/kg (high-dose semaglutide), or zalfermin 0.05 mg/kg + semaglutide 3 µg/kg (two separate injections) (low-dose zalfermin-semaglutide). Low-dose zalfermin-semaglutide combination treatment aimed to promote a weight loss equivalent to that attained by individual high-dose monotherapies. For semaglutide, the selected high dose aimed to reflect a clinically relevant dose in patients with MASH but adjusted for shorter circulating half-life in mice compared with humans (7.5 vs. 168 h) [22,23]. Body weight was measured daily over the entire dosing period. Twenty-four-hour food intake was measured daily during the first 14 days of treatment and once weekly thereafter (on treatment days 19, 26, 33, 40, 47, and 53). Animals were terminated by cardiac puncture under isoflurane anesthesia.

Four-hour-fasted terminal blood was sampled from the tail vein, kept on ice, and centrifuged (5 min, 4°C; 6000 g) to generate ethylenediaminetetraacetic acid-stabilized plasma. Plasma alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides (TG), total cholesterol (TC), and liver TG and TC levels were determined as described previously [21].

Baseline and terminal liver biopsy samples (both from the left lateral lobe) were fixed overnight in 4% paraformaldehyde. Liver tissue was paraffin-embedded and sectioned (3 μm thickness). Sections were stained with hematoxylin-eosin (HE), picro-Sirius Red (PSR; Sigma-Aldrich, Brøndby, Denmark), anti-cluster of differentiation molecule 11B (anti-CD11b [Integrin αM]) (cat. ab1333357, AbCam, Cambridge, UK), anti-galectin-3 (cat. 125402; Biolegend, San Diego, CA), anti-alpha-smooth muscle actin (α-SMA, cat. ab124964; Abcam, Cambridge, UK), or anti-Col1a1 (cat. 1310−01; Southern Biotech, Birmingham, AL) using standard procedures [21]. Histopathological scoring of steatosis, lobular inflammation, hepatocyte ballooning, and fibrosis stage was performed using the NASH Clinical Research Network scoring system as outlined by Kleiner et al. [2]. Histopathological scores were assessed using an automated deep learning–based image analysis pipeline (Gubra Histopathological Objective Scoring Technique, GHOST), validated previously [24]. Digital image analysis, using the Visiomorph software (Visiopharm, Hørsholm, Denmark), was applied for quantification of markers of inflammation (CD11b, Gal-3), fibrosis (Col1a1), and fibrogenesis (α-SMA) on immunohistochemically stained slides. Area fractions of immunopositive staining were expressed as percentages relative to total parenchymal (fat-free) area by subtracting the corresponding fat area determined on adjacent HE sections.

RNA was extracted from snap-frozen terminal liver samples (15–20 mg fresh tissue, n = 8 per group) and liver transcriptome analysis was performed by RNA sequencing analysis, as previously described in detail [25]. RNA sequence libraries were prepared using NeoPrep (Illumina, San Diego, CA) and the Illumina TruSeq stranded mRNA Library kit for NeoPrep (Illumina) and sequenced using the NextSeq 500 (Illumina) with NSQ 500 hi-Output KT v2 (75 CYS, Illumina). RNA sequencing data were aligned with STAR v2.5.2a using decoy-aware indexes (GRCm38_89). Quality control and alignment statistics were obtained with MultiQC (v1.11). All subsequent data processing and analyses were performed in R (v4.2.0) using the tidyverse packages (v2.0.0). After loading of RNA sequencing data, DESeq2 v1.36.0 (default parameters) was used for the quality control and differential expression analyses. Only genes that were expressed in at least three samples with five reads or more were kept for testing of differential expression. Log-fold changes were adjusted using DESeq2 and Benjamini–Hochberg’s method (5% false discovery rate [FDR < 0.05]) was used for multiple testing correction of p-values. Overlap of differentially expressed genes was visualized using the UpSetR package (v1.4.0). Heatmaps depicting changes in MASH and fibrosis-associated candidate gene expression were prepared using the pheatmap package (v1.0.12). Venn diagrams were prepared using the eulerr package (v7.0.0). One outlier sample from the DIO-MASH low-dose semaglutide treatment group was excluded after inspection of the clustering results of the variance-stabilized transformed transcriptome data (DESeq2’s vst) function. After this exclusion, the RNA sequencing dataset had an average 17M mapped reads (standard deviation, 4.5M). Body weight was used as a covariate in the DESeq2 analysis when performing body weight–independent analyses as previously described [26].

Data are presented as mean ± standard error of the mean. A one-sided Fisher’s exact test with Bonferroni correction was used for within-subject comparison of histopathological scores before and after treatment intervention. All other statistical analyses were performed using a one-way analysis of variance (ANOVA) with Dunnett’s correction for multiple comparisons. In cases of significantly different standard deviations using a Brown-Forsythe test, Welsh correction ANOVA with Dunnett’s T3 correction was applied. The vehicle-dosed chow-fed mouse group was excluded from statistical tests. Superiority to high-dose semaglutide and individual monotherapies was tested using the above statistical methods. A p-value less than 0.05 was considered statistically significant.

Maximal weight loss (relative to baseline) attained by zalfermin and semaglutide mono- and combination therapy regimens was observed within the first 2 weeks of dosing and was sustained thereafter (Fig 1B). AMLN DIO-MASH mice showed a modest but significant sustained body weight loss after 8 weeks of monotherapy with low-dose zalfermin (−5.9 ± 1.2%, p < 0.001 vs. DIO-MASH vehicle) and low-dose semaglutide (−4.0 ± 1.1%, p < 0.001 vs. DIO-MASH vehicle) relative to baseline. Compared with low-dose monotherapies, combined low-dose zalfermin-semaglutide treatment resulted in more-than-additive weight loss (−18.6 ± 1.4%, p < 0.001 vs. DIO-MASH vehicle, Figs 1B and 1C, Table 1). The magnitude of weight loss attained by low-dose zalfermin-semaglutide combination treatment was comparable to monotherapy with high-dose semaglutide (−15.3 ± 1.3%, p < 0.001 vs. DIO-MASH vehicle; p > 0.05 vs. combination treatment) and high-dose zalfermin (−16.3 ± 0.8%, p < 0.001 vs. DIO-MASH vehicle; p > 0.001 semaglutide low), respectively. High-dose, but not low-dose, semaglutide monotherapy reduced cumulative food intake vs. DIO-MASH vehicle (p < 0.001, S1 Fig). In contrast, both low- and high-dose zalfermin monotherapy significantly increased cumulative food intake (both p < 0.001 vs. DIO-MASH vehicle, S1 Fig). Cumulative food intake was unaffected following low-dose zalfermin-semaglutide combination treatment (p > 0.05 vs. DIO-MASH vehicle, S1 Fig).

The low-dose zalfermin-semaglutide combination treatment had additive benefits on hepatomegaly in AMLN DIO-MASH mice. Accordingly, low-dose zalfermin-semaglutide combination treatment resulted in more than a 50% reduction in liver weight compared with vehicle controls (1.7 ± 0.1 g vs. 3.5 ± 0.2 g, p < 0.001, Fig 1D). Monotherapies showed dose-dependent effects on liver weight (low-dose zalfermin, 2.6 ± 0.2 g, p < 0.01; high-dose zalfermin, 1.7 ± 0.1 g, p < 0.001 vs. DIO-MASH vehicle; low-dose semaglutide, 2.8 ± 0.1 g, p < 0.05; high-dose semaglutide 2.1 ± 0.1 g, p < 0.001 vs. DIO-MASH vehicle, Fig 1D). Liver index was also additively lowered by low-dose zalfermin-semaglutide combination treatment (5.2 ± 0.2%) compared with DIO-MASH vehicle controls (8.9 ± 0.4%, p < 0.001, Fig 1E, Table 1). In comparison, improvements in the liver index were less pronounced following low-dose zalfermin (7.0 ± 0.3%, p < 0.001 vs. DIO-MASH vehicle) and low-dose semaglutide monotherapy (7.3 ± 0.3%, p < 0.001 vs. DIO-MASH vehicle). Further reductions were achieved with high-dose zalfermin (4.9 ± 0.1%, p < 0.001 vs. DIO-MASH vehicle) and semaglutide monotherapy (6.2 ± 0.3%, p < 0.001 vs. DIO-MASH vehicle).

Plasma transaminases were significantly elevated in AMLN DIO-MASH mice (Figs 2A and 2B). Notably, AMLN DIO-MASH mice achieved near-normal plasma ALT levels upon low-dose combination treatment with zalfermin and semaglutide (Fig 2A), reflecting near-additive effects of the treatment regimen (Table 1). Both high-dose monotherapies also promoted substantial reductions in plasma ALT levels. Low-dose zalfermin-semaglutide combination treatment, but not low-dose monotherapies, also significantly reduced plasma AST levels (Fig 2B). The effect of combined low-dose combination treatment was more than additive (Table 1). The efficacy of low-dose zalfermin-semaglutide combination treatment on plasma AST levels was comparable to high-dose zalfermin and semaglutide (Fig 2B).

Plasma TG concentrations were marginally reduced by the high-fat and cholesterol-rich AMLN diet. Plasma TG levels did not change further due to treatments (Fig 2C). In contrast, hypercholesterolemia and elevated liver TG levels in AMLN DIO-MASH mice were markedly lowered by low-dose zalfermin-semaglutide combination treatment (p < 0.001). The combination regimen demonstrated significantly larger reductions in these two lipid markers compared with high-dose semaglutide monotherapy and was more than additive for liver TG reductions of the two low-dose monotherapies (p < 0.01 for both endpoints, Figs 2D and 2E, Table 1). High-dose zalfermin was comparable to combination treatment on liver TG, while high-dose semaglutide and zalfermin monotherapy demonstrated comparable therapeutic efficacy on plasma TC. Less-pronounced effects on plasma TC were noted for low-dose monotherapies compared with DIO-MASH controls (zalfermin low-dose, p < 0.05; semaglutide low-dose, p < 0.01, Fig 2D). Also, low-dose monotherapies did not improve liver TG levels (Fig 2E). Elevated liver TC levels in AMLN DIO-MASH mice were unaffected by treatments (Fig 2F).

The effect of zalfermin and semaglutide was evaluated on liver semiquantitative and quantitative liver histological outcomes. Chow controls had normal liver histology. For AMLN DIO-MASH mice, treatment groups were randomized and stratified to obtain comparable severity of MASH (NAFLD activity score [NAS] 5–6) and fibrosis (stage F2–F3) at baseline (). Treatment outcomes on these clinically derived semiquantitative endpoints were determined by evaluating within-subject pre- to post-liver biopsy histopathological scores (). S2 Fig S2 Fig

Differential histological efficacy profiles were observed for zalfermin and semaglutide monotherapy. Accordingly, of the two, only high-dose zalfermin treatment led to significant improvement in NAS (1 point, 12/12 mice, p < 0.001 vs. DIO-MASH vehicle), being largely driven by robust and consistent reductions in steatosis scores (Figs 3A and 3C). Zalfermin and semaglutide alone had no significant effect on lobular inflammation (Fig 3D). Low-dose zalfermin-semaglutide combination treatment resulted in further robust histological benefits, as indicated by improvements in NAS (1 point, 4/11 mice, p < 0.01; ≥ 2 points, 6/11 mice, p < 0.05, Fig 3A), which were explained by combined more-than-additive benefits on both steatosis scores (1 point, 9/11 mice, p < 0.001; 2 points, 1/11 mice, p > 0.05, Fig 3C, Table 1) and lobular inflammation scores (1 point, 4/11 mice, p < 0.05; 2 points, 2/11 mice, p > 0.05, Fig 3D, Table 1). Hepatocyte ballooning was marginal in AMLN DIO-MASH mice and unaffected by treatments (Fig 3E). Fibrosis stage worsened in all AMLN DIO-MASH mouse groups over the 8-week study period, with an incidence rate ranging from 27% to 42% (Fig 3B). Treatments for 8 weeks did not improve fibrosis scores (Fig 3B).

A series of histomorphometric analyses were performed to confirm histological benefits of low-dose zalfermin-semaglutide combination treatment (Fig 4, S3 Fig). Reduced steatosis severity was corroborated by quantitative digital image analysis of HE-stained sections used for pre- to post-steatosis scoring. Using this approach, we detected significantly reduced fractional (%) area of lipids following monotherapy with low-dose zalfermin or semaglutide (both p < 0.001 vs. DIO-MASH vehicle), emphasizing the greater sensitivity of histomorphometry compared with semiquantitative scoring modules. Consistent with further improved steatosis scores, low-dose zalfermin-semaglutide combination treatment reduced liver fat content more than low-dose monotherapies, indicating more-than-additive effects of the low-dose combination regimen (Fig 4A, Table 1). The magnitude of quantitative liver fat depletion afforded by low-dose zalfermin-semaglutide combination treatment was comparable to high-dose monotherapies (Fig 4A). Immunohistochemical analyses were performed to evaluate quantitative effects on standard molecular markers of inflammation (CD11b, Gal-3), fibrosis (Col1a1), and stellate cell activation/fibrogenesis (α-SMA). Overall, low-dose combination treatment reduced parenchymal (fat-free) expression of CD11b, Gal-3, and α-SMA (all p < 0.001 vs. DIO-MASH vehicle) to a similar degree to individual monotherapies (Figs 4B, 4C, and 4E). As for fibrosis scores, treatments had no significant effect of quantitative Col1a1 histology (Fig 4D).

Taken together, our data demonstrate that low-dose zalfermin and semaglutide have added therapeutic benefits on body weight, plasma liver enzymes, lipid biochemistry, hepatic steatosis, and liver inflammation in DIO-MASH mice. To further investigate the underlying molecular effects, we evaluated liver transcriptome changes for gene-expression programs associated with these treatment outcomes.

High-quality transcriptome profiles were obtained from liver samples across the experimental groups (Fig 5, S4 Fig). A principal component analysis (PCA) of the 500 most variable genes across all samples indicated clear group separation of differentially expressed genes (DEGs). Most of the variance in the hepatic transcriptome data set was contributed by the disease phenotype (DIO-MASH vehicle vs. chow vehicle, n = 7,758 DEGs, Figs 5A and 5B). All treatments promoted a rightward shift in the PCA plot (i.e., toward chow-fed vehicle controls), indicating gradual normalization of the hepatic global gene expression profile. The rightward shift was most pronounced for high-dose monotherapies and low-dose zalfermin-semaglutide combination therapy (Fig 5A). The total number of DEGs in the treatment groups vs. vehicle-dosed DIO-MASH mice were as follows: low-dose zalfermin (n = 221), low-dose semaglutide (n = 0), low-dose zalfermin-semaglutide combination (n = 3,323), high-dose zalfermin (n = 2,766), and high-dose semaglutide (n = 649) (Fig 5B, S4 Fig). It is noteworthy that the number of DEGs in the low-dose zalfermin-semaglutide group was 15-fold higher compared with summation of the corresponding low-dose monotherapy groups, indicating more-than-additive effects at the hepatic transcriptome level.

The expression profile of FGF21 and its cognate receptors was also investigated. DIO-MASH mice demonstrated significantly increased Fgf21 expression and reduced Ffgr4 and Klb expression compared with chow-fed controls (Fig 5C). Hepatic Fgfr1, Fgfr2, and Fgfr3 expression was unaltered in DIO-MASH mice. Only low-dose zalfermin-semaglutide combination treatment and high-dose zalfermin monotherapy partially reversed Ffgr4 expression. Low-dose combination treatment also significantly increased hepatic Klb expression relative to DIO-MASH vehicle controls (Fig 5C). A similar effect was observed for high-dose zalfermin and semaglutide, but not for low-dose monotherapies.

To obtain further resolution of hepatic transcriptome regulations, RNA sequencing data were probed for candidate genes linked to MASH and fibrosis. Consistent with histological hallmarks of MASH and fibrosis, DIO-MASH mice showed widespread regulations in candidate gene markers of liver metabolism, immune function, and extracellular matrix (ECM) organization (Fig 6A). Overall, low-dose zalfermin-semaglutide combination treatment reversed changes in all disease-associated gene categories investigated. Changes in liver metabolic markers suggest that low-dose combination treatment improves intrahepatic handling of carbohydrates (upregulation: Irs1, Khk, Pygl, Slc2a4; downregulation: Mtor), bile acids (upregulation: Abcb11, Slc22a7, Slc27a5, Slco1a4), and lipids (upregulation: Apoa1, Apoa5, Apoc3, Hmgcr, Ppard, Scarb1, Thrb, Vldlr; downregulation: Cd36, Pparg) (Fig 6A). Only low-dose combination treatment increased gene expression markers associated with endoplasmic reticulum stress (Atf4, Atf6, Ern1), displaying a similar profile to chow mice. Notably, low-dose combination treatment suppressed expression of nearly all (22/30) gene markers of ECM remodeling. These included all major collagen subtypes (Col1a1, Col3a1, Col4a1, Col5a1/2/3, Col6a1/2/3) as well as several isoforms of matrix metalloproteinases and tissue inhibitors of matrix metalloproteinases, implying reduced pro-fibrogenic activity of the combination treatment (Fig 6A). In comparison, pronounced effects were also observed for gene markers of the immune system (downregulation: CCl2, Ccr1/2/5, Cysltr1; upregulation: Nr1h3, Traf6). Whereas the above changes in candidate genes were comparable overall to high-dose zalfermin treatment, fewer candidate genes were differentially regulated by high-dose semaglutide.

Considering that both zalfermin and semaglutide promoted weight loss in DIO-MASH mice, and body weight loss is limited in patients with MASH treated with FGF21 analogs, the body weight loss–independent benefits of the treatments on transcriptome signatures in DIO-MASH mice were investigated. To this end, the differential gene expression analyses were adjusted for body weight (Fig 6B, S5 Fig). Compared with chow-fed controls, candidate genes involved in liver metabolism, inflammation, and fibrogenesis remained significantly regulated in DIO-MASH control mice independent of body weight, likely signifying re-routing of hepatic metabolism necessary to manage dysmetabolic and proinflammatory effects of the high-fat/fructose diet. While treatment-induced changes in liver metabolic markers, inflammation, and endoplasmic reticulum stress were largely explained by weight loss, a subset of ECM-related genes remained suppressed after adjusting for body weight in DIO-MASH mice. This only applied to low-dose zalfermin-semaglutide combination treatment (downregulation: Col3a1, Col4a1, Col5a1/2, Loxl2) and monotherapy with low-dose zalfermin (downregulation: Col5a2) and high-dose zalfermin (downregulation: Col1a1, Col1a2, Col3a1, Col4a1, Col5a1/2/3, Loxl2, Pdgfa, Timp1). Hence, low-dose zalfermin-semaglutide combination treatment promotes more robust suppression of ECM-associated genes than low-dose zalfermin monotherapy. High-dose zalfermin presented the broadest body weight–independent downregulation of ECM-associated genes. Semaglutide monotherapy did not show any body weight–independent effects on gene expression in this study of 8-week intervention treatment.