Altered hepatic metabolism in Down syndrome

SUMMARY

Trisomy 21 (T21) gives rise to Down syndrome (DS), the most commonly occurring chromosomal abnormality in humans. T21 affects nearly every organ and tissue system in the body, predisposing individuals with DS to congenital heart defects, autoimmunity, and Alzheimer’s disease, among other co-occurring conditions. Here, using multi-omic analysis of plasma from more than 400 people, we report broad metabolic changes in the population with DS typified by increased bile acid levels and protein signatures of liver dysfunction. In a mouse model of DS, we demonstrate conservation of perturbed bile acid metabolism accompanied by liver pathology. Bulk RNA sequencing revealed widespread impacts of the Dp16 model on hepatic metabolism and inflammation, while single-cell transcriptomics highlighted cell types associated with these observations. Modulation of dietary fat profoundly impacted gene expression, bile acids, and liver pathology. Overall, these data represent evidence for altered hepatic metabolism in DS that could be modulated by diet.

In brief

Dunn et al. show that individuals with Down syndrome display widespread evidence of altered hepatic function in the plasma proteome and metabolome, typified by elevated plasma bile acids. Using a mouse model of Down syndrome, they show that these alterations are driven, in part, by diet.

Graphical Abstract

INTRODUCTION

Triplication of chromosome 21 (chr21), known as trisomy 21 (T21), results in the genetic disorder commonly known as Down syndrome (DS).¹ T21 is the most common chromosomal abnormality in the human population, affecting approximately 300,000 individuals in the United States and around 6 million worldwide.² T21 causes widespread perturbations across cell types and organ systems.³ Although individuals with DS are predisposed to obesity, Alzheimer’s disease, and autoimmune disorders, the mechanisms driving these increased rates remain poorly understood.^3,4 High-throughput “omics” approaches have provided new insights into the system-wide impacts of T21 on transcriptome,^5–7 proteome,^5,6,8–10 metabolome,^6,9 and peripheral immune compartment,^6,7,11 including widespread metabolic dysregulation in DS.^12–17

Bile acids (BAs) have diverse roles in metabolism, homeostasis, and disease.¹⁸ Primary BAs are synthesized from cholesterol in hepatocytes and secreted into bile ducts, then trafficked to the gut where they play roles in the metabolism of lipids, fats, and cholesterol.¹⁹ In the gut, the microbiome can modify BAs into secondary BAs. In the liver and intestine, BAs serve as ligands for nuclear receptors, such as farnesoid X Receptor (FXR),²⁰ where they stimulate gene expression programs that regulate further production of BAs and maintain lipid homeostasis. Elevated plasma BAs are a hallmark of altered liver metabolism.^18,21 Although around 4.5 million adults in the United States have a diagnosed liver disease, the American Liver Foundation estimates that another 80–100 million adults in the United States have an undiagnosed liver disease, and that number is rising.^22,23

Liver disease diagnoses are commonly made on the results of liver function tests (LFTs), such as elevated alanine aminotransferase (ALT) or aspartate aminotransferase (AST), which are markers of hepatocellular damage commonly seen in patients with metabolic associated steatotic liver disease (MASLD), or elevated plasma BAs, which are commonly associated with cholestasis.^24,25 Imaging-based methods such as ultrasound or FibroScan assist in proper diagnosis of liver disease.^25,26 The rates of liver disease in individuals with DS remains unknown, although a recent large cohort study examined LFTs, including ALT and AST, and concluded that T21 has no biochemical evidence of hepatocellular injury,¹⁶ while another study using electronic medical records concluded that individuals with DS are less likely to suffer from liver disease.²⁷ In contrast, a 2017 study demonstrated that more than 60% of children with DS had ultrasound findings consistent with fatty liver even in the absence of obesity and independent of elevated AST and ALT levels.²⁸ These conflicting reports suggest the presence of liver dysfunction in the population with DS that is not easily detectable through conventional LFTs.

Here, we report that the population with DS has increased plasma BAs and that within the plasma proteome, there exists evidence of liver dysfunction. Furthermore, hepatocytes differentiated from T21-induced pluripotent stem cells (iPSCs) demonstrated concordance with the features dysregulated in the plasma of individuals with DS, including protein abundance, lipid production, and BA production. Dp(16)1Yey/+ mice (referred to hereinafter as Dp16), which model DS, display a liver phenotype characterized primarily by increased inflammation, arterial thickening, and ductular reaction.²⁹ Metabolomic analysis of plasma from Dp16 mice shows elevated levels of BAs. Transcriptome analysis of the Dp16 liver identified signatures of extracellular matrix remodeling, immune system dysregulation, and metabolic dysfunction. Single-cell RNA sequencing (scRNA-seq) of Dp16 liver samples highlighted cell-type-specific alterations in metabolic and inflammatory pathways.

Finally, we challenged Dp16 mice with a high-fat diet (HFD) and found that the liver phenotype is influenced by diet. Transcriptome analysis of the liver after HFD showed perturbations in pathways associated with the inflammation and metabolism, and plasma BAs are perturbed after this diet, suggesting an association between diet and BA dysregulation. Overall, the evidence presented here indicates that individuals with DS, as well as the Dp16 model, experience widespread liver dysfunction, defined primarily by dysregulated hepatic metabolism, which may be modulated by dietary fat.

RESULTS

Altered bile acid metabolism in Down syndrome

To expand upon our observations that DS globally alters metabolism, we reanalyzed datasets generated by the Human Trisome Project (HTP).^{6,10,17,30,31} These datasets represent molecular profiling of thousands of features across multiple omics platforms from 316 individuals with DS (T21) and 103 euploid controls (D21). The control group was 56% male, and the T21 group was 47% male (Table S1). The average age for controls was 28 years with an average body mass index (BMI) of 24, whereas the T21 group was slightly younger at 23 years, with an average BMI of 27. The T21 cohort was representative of the broader T21 population with respect to co-occurring conditions; for example, 56% of this cohort had a congenital heart defect and 40% had sleep apnea (Table S1). We first examined the impact of T21 on the plasma metabolome by analyzing the relative levels of plasma metabolites. We identified 102 differentially abundant metabolites (Figure 1A; Table S2A). Next, we grouped metabolites based on pathway, which highlighted global metabolic changes in DS, typified by elevation of fatty acids/eicosanoids, carnitines and fatty acid metabolism, and BAs, as well as depletion of amino acids and nucleotides (Figure 1B). Closer examination of BAs revealed a total of 11 measured, with nine elevated in DS, including taurolithocholic acid (TLCA) (Figures 1C, 1D, and S1A). Primary BAs trended higher, and secondary BAs were significantly increased in T21 plasma (Figure S1B). Notably, these elevated levels were conserved across the lifespan of individuals with DS, with no significant impact of age on the levels of TLCA(Figures 1E; Table S2B), as is the case for other metabolites including kynurenine.³² Some metabolites, however, are perturbed by T21 in an age-specific manner, making the consistent elevation of BAs noteworthy.³² In the typical population, elevated BMI is frequently associated with liver disease; however, no significant association between BMI and TLCA was observed in our T21 cohort (Figure S1C). Finally, we asked whether several co-occurring conditions associated with liver dysfunction impacted levels of BAs and concluded that they did not (Figure S1D; Table S2C).

Figure 1. Altered bile acid metabolism in Down syndrome.

(A) Volcano plot summarizing results of plasma metabolomic analysis of individuals with T21 (n = 316) versus controls (n = 103). Significantly differentially abundant (q < 0.1) metabolites are red.

(B) Waterfall plot summarizing plasma metabolites with significant differences in individuals with T21 versus controls (q < 0.1).

(C) Heatmap displaying log₂(foldchange) of plasma bile acids in those with T21 relative to controls. Asterisks denote significance (q < 0.1).

(D) Sina plot displaying log₂(relative abundance) of taurolithocholic acid in the plasma of D21 individuals versus T21. Boxes represent interquartile ranges and medians, with notches approximating 95% confidence intervals.

(E) Scatterplot displaying the relationship between relative abundance of taurolithocholic acid and age in years in D21 and T21. Spearman rho values and p values are shown.

(F) Volcano plot summarizing results of plasma metabolomic analysis of Dp16 mice (n = 13) versus wild-type (n = 9). Significantly differentially abundant (q < 0.1) metabolites are red.

(G) Waterfall plot summarizing plasma metabolites with significant differences between Dp16 and wild-type mice.

(H) Heatmap displaying log₂(foldchange) of plasma bile acids in Dp16 mice relative to wild-type. Asterisks denote significance (q < 0.1).

(I) Sina plots displaying log₂(relative abundance) of indicated plasma metabolites in wild-type and Dp16 mice.

For (D and I), the Benjamini-Hochberg adjusted p values (q values) are indicated.

To investigate potential mechanisms leading to BA dysregulation in individuals with T21, we used the Dp16 mouse model, which harbors a segmental duplication of a portion of mouse chr16, conferring trisomy of around 120 protein coding genes orthologous to human chr21.³³ We first characterized the plasma metabolome of a cohort of Dp16 animals and matched wild-type (WT) littermate controls by measuring the relative abundance of ~150 metabolites and identified 80 as differentially abundant (Figure 1F; Table S2D). Strikingly, BAs were the most enriched pathway in Dp16 animals, with 12 of 17 significantly elevated (Figures 1G and 1H). We observed differential abundance of both primary and secondary BAs (Figures 1G, 1H, S1E, and S1F). Thus, the Dp16 mouse model of DS exhibits defects in BA metabolism like those seen in individuals with DS.

Evidence for widespread liver dysfunction in Down syndrome

Given that high levels of plasma BAs can indicate the presence of liver disease, we reanalyzed our HTP SOMAscan plasma proteomic data for markers of liver function.³⁴ We measured more than 4,500 epitopes and found hundreds of proteins with altered abundance in individuals with T21 (Figure S2A; Table S3A). Ingenuity pathway analysis (IPA) of these differentially abundant proteins uncovered several signatures of liver dysfunction in individuals with DS (Figure 2A). Two of the top six enriched canonical pathways were related to liver dysfunction: liver X receptor/retinoid X receptor (LXR/RXR) activation and hepatic fibrosis/stellate cell activation (Figure 2A). Gene set enrichment analysis (GSEA) revealed similar findings: positively enriched hallmarks included multiple immune-related pathways, and negatively enriched pathways included the coagulation pathway (Figure S2B; Table S3B).

Figure 2. Evidence for widespread liver dysfunction in Down syndrome.

(A) Bar plot summarizing Ingenuity Pathway Analysis of plasma SOMAscan proteomics from individuals with DS (n = 316) versus controls (n = 103). Dashed line indicates the significance threshold of p < 0.05.

(B) Sina plots displaying the relative abundance of various plasma proteins in D21 versus T21.

(C) Enrichment plots showing enrichment scores for gene set enrichment analysis signatures of fibrosis and steatosis in plasma proteomics of individuals with DS. NES, normalized enrichment score.

(D) Sina plots displaying the relative abundance of plasma proteins in D21 and T21.

(E) Sina plots showing the concentration of biomarkers of liver function in wild-type (n = 8–11, 3–6 females) and Dp16 (n = 9–11, 3–6 females) plasma.

(F) Representative images of H&E-stained liver sections from wild-type and Dp16 mice. Arrows indicate the following features: red, enlarged periportal sinusoids; blue, bile duct; yellow, hepatic artery branches; green, periportal inflammatory cells. PV indicates portal vein, and CV indicates central vein. Scale bars, 100 μm.

(G) Sina plots displaying metrics of liver pathology scoring from adult wild-type (n = 12, 6 females) and Dp16 (n = 11, 5 females) mice.

For (B and D), boxes represent interquartile ranges and medians, with notches approximating 95% confidence intervals. Benjamini-Hochberg adjusted p values (q values) are indicated.

For (E and G), individual data are presented with a bar at the median, and p values, as determined by Mann-Whitney U test, are shown.

Consistent with prior reports,^10,35 we found no significant difference in the plasma levels of ALT or AST in the T21 group (Figure S2C). However, we observed significant differences in several proteins known to be dysregulated in liver disease, some of which have the potential to predict the severity of various disorders.^36,37 For example, we observed increased levels of serum amyloid A1 (SAA1) and collagen 6A1 (COL6A1) and decreased levels of albumin (ALB) and insulin-like growth factor 1 (IGF1) (Figure 2B).³⁸ Hypoalbuminemia is a hallmark of liver dysfunction, and lower levels of IGF1 have been associated with the progression of various liver diseases.^39,40

Next, we utilized a study that clinically staged both steatosis and fibrosis within euploid individuals diagnosed with liver disease and generated plasma proteomic biosignatures of disease using the SOMAscan platform.^6,41 The aptamer-based nature of this assay allowed us to directly compare results across studies. Using the lists of differentially abundant proteins identified by Govaere et al., we performed GSEA on our HTP SOMAscan data and found that both fibrosis and steatosis signatures were significantly enriched in individuals with DS (Figure 2C; Table S3C). Within these signatures, we observed proteins linked to liver dysfunction, including innate immunity/extracellular matrix remodeling (CHI3L1, FCN2, and THBS2), cell adhesion (SELE), and metabolism (AMY2A) (Figures 2D and S2C). Finally, we found that changes to several liver-related features were not significantly associated with co-occurring conditions (Figure S2D; Table S3D).

We also measured plasma markers of liver function in Dp16 animals and found no differences in ALT or AST (Figure 2E; Table S3E).^29,42 ALB levels were depleted in Dp16 animals relative to WTs (Figure 2E; Table S3F), and SAA trended toward increased levels, while IGF1 and CHI3L1 trended toward decreased levels in Dp16 animals (Figure S2E; Tables S3G–S3I).

We previously reported a liver phenotype in Dp16 mice.²⁹ We reproduced these results in new cohorts of adult mice, showing that Dp16 mice have an increase in overall histopathology scoring of the liver, with the most drastic score coming from the reactive changes parameter (Figures 2F and 2G; Table S3J). This score was driven by “ductular reaction” (DR) characterized by proliferation of bile ducts, increased thickness and number of arteries, and hyperplasia of portal triad endothelial cells (Figures 2F and 2G). Adult Dp16 animals also had elevated levels of inflammation in the liver, with periportal inflammation and accumulation of lymphocytes (Figures 2F and 2G). Liver pathology was also present at 1 month of age, with modest contributions to the overall pathology score coming from all three categories: cell injury, inflammation, and reactive changes (Figures S2F and S2G; Table S3K). However, only the elevated reactive change scores were significant in Dp16 livers. We did not observe steatosis in either 1- or 4-month-old animals (Figures S2G and S2H). Finally, plasma markers of liver function, such as alkaline phosphatase (ALP) and creatinine (CREA), trended toward elevation in 1-month-old animals, and SAA and CHI3L1 concentrations were decreased, while no difference in CRP nor ALB was observed between the two genotypes (Figure S2I; Tables S3L–S3Q).

Dp16 livers exhibit stellate cell activation and fibrosis

Motivated by our observations of liver pathology in the Dp16 model, we performed bulk transcriptome analysis of the adult murine liver, which revealed ~1,500 differentially expressed genes (DEGs) in Dp16 compared to WT (Figure 3A; Table S4A). GSEA demonstrated positive enrichment scores for the epithelial-mesenchymal transition (EMT), inflammatory response, interferon (IFN) gamma response, and coagulation gene sets, among others (Figure 3B; Table S4B). Negatively enriched pathways included cholesterol homeostasis, bile acid metabolism, fatty acid metabolism, and adipogenesis (Figures 3A and 3B). Within the EMT pathway, we found significant upregulation of extracellular matrix remodeling, including Col1a1, a gene that encodes for alpha chains of type I collagen (Figures 3C and 3D), which is deposited in excess during injury or disease, causing fibrosis.⁴³

Figure 3. Stellate cell dysfunction drives liver fibrosis in the Dp16 model.

(A) Volcano plot summarizing results of whole liver transcriptome analysis of wild-type (n = 6, 3 females) versus Dp16 (n = 6, 3 females) mice. Significantly differentially expressed (q < 0.1) genes are colored in red.

(B) Bar plot summarizing the results of gene set enrichment analysis of gene expression changes data in (A).

(C) Heatmap showing the median Z score of the top 10 leading-edge genes from the epithelial-mesenchymal transition gene set.

(D) Sina plots displaying the normalized relative expression (RPKM) of Col1a1 and Eln in wild-type and Dp16 mice.

(E) Uniform manifold approximation and projection (UMAP) plot of single-cell RNA sequencing (scRNA-seq) analysis of mouse liver, color coded by cell clusters identified using Seurat.

(F) UMAP plot displaying differential cellular abundance of clusters from scRNA-seq analysis of mouse liver. Significant (false discovery rate [FDR] <0.1) clusters are colored by mean fold-change.

(G) Sina plots showing the Dp16 versus wild-type abundance of hepatocyte cluster 1 and endothelial cell cluster 2 from scRNA-seq analysis of mouse liver.

(H) UMAP plot displaying the epithelial-mesenchymal transition gene set normalized enrichment score (NES) for each cluster.

(I) Bubble plot summarizing Col1a1 single-cell expression across clusters and animals, with color representing mean expression and size representing percent of cells with expression.

(J) Sina plot displaying the normalized pseudobulk counts of Col1a1 expression within the stellate cell cluster in wild-type and Dp16 animals.

(K) Representative images of multiplexed stained liver sections from wild-type and Dp16 mice where DAPI is stained in blue, desmin in red, and smooth muscle actin (SMA) in yellow. Sina plot showing number of desmin+ and SMA+ cells as a percentage of total cells in wild-type mice (n = 6, 3 females) and Dp16 mice (n = 6, 3 females).

(L) Representative images of picrosirius-red-stained (PSR) liver sections from wild-type (n = 8, 4 females) and Dp16 (n = 7, 4 females) mice, imaged under polarized light. Sina plot shows the total percent positive PSR staining in wild-type and Dp16 liver.

For (K and L), individual data are presented with a bar at the median, and the p value, as determined by a Mann-Whitney U test, is shown. Scale bars, 100 μm.

Next, we performed scRNA-seq of WT and Dp16 livers and identified 16 distinct cell clusters representing all major cell types of the liver (Figures 3E and S3A; Tables S5A–S5R). We observed an increase in abundance in the Dp16 liver of hepatocyte cluster 1, as well as increases in Kupffer cell cluster 2 and B cells (Figure 3F). Conversely, we observed slight decreases in abundance in endothelial cell cluster 2 and cholangiocytes (Figures 3F and 3G). DE-Seq2 analysis of cluster by gene-level pseudobulk count data followed by GSEA showed widespread effects of Dp16 on gene expression across clusters, with some gene sets consistently activated (e.g., allograft rejection), some activated in a cell-type-specific fashion (e.g., hedgehog signaling), and others discordant across cell types (e.g., oxidative phosphorylation) (Figure S3B). The EMT pathway was dysregulated in numerous cell types: activated in some and inactivated in others (Figure 3H). Dp16 hepatic stellate cells (HSCs) displayed the highest degree of EMT activation (Figure 3H), and, when looking at the expression of the leading-edge genes for this pathway, for example Col1a1, we observed that expression and upregulation were restricted to HSCs (Figures 3I and 3J). Expression of Gas6, whose axis with Axl is required for complete activation of HSCs, was also elevated (Figure S3C).⁴⁴

HSCs can be activated by numerous factors, including Kupffer cells, which release cytokines and chemokines such as TGFβ, which was upregulated in Kupffer cell cluster 1 (Figure S3D).⁴⁵ The purinergic receptor P2rx7 has been implicated in many liver diseases, specifically for its role in NLRP3 inflammasome activation and interleukin (IL)-1B release and was upregulated in Dp16 Kupffer cells (Figure S3E).⁴⁶ Another cell type known to frequently activate HSCs is hepatocytes. Dp16 hepatocyte cluster 2 displayed increased expression of Egfr (epidermal growth factor receptor) (Figure S3F). At baseline, Egfr supports liver homeostasis, but upon injury, it promotes the recruitment of immune cells, hepatocyte proliferation to repair damaged tissue, and activation of HSCs.⁴⁷ We also observed that there were multiple inflammatory signatures present in the Dp16 HSCs themselves, including elevated expression of Cd74, which could lead to HSC activation through migration inhibitory factor (MIF) signaling (Figure S3G).

To examine spatial and cell-type-specific alterations in Dp16 liver architecture, we performed multiplexed immunofluorescence analysis. This allowed us to visualize the location of major non-immune cell types in the liver, including hepatocytes, cholangiocytes, liver endothelial cells (LECs), liver sinusoidal endothelial cells (LSECs), as well as both inactive and activated HSCs. Consistent with our scRNA-seq data, several cell types, including activated HSCs, were more abundant, particularly around the portal vein (Figures 3K and S3H; Table S5S). As activated HSCs are fibrotic, we next tested for collagen buildup with picrosirius red and observed a significant increase in collagen in the adult Dp16 liver compared to WT but not younger animals (Figures 3L and S3I). We also saw a trend toward increased abundance of non-activated HSCs in the Dp16 liver and a significant increase in both Lyve1+ (lymphatic vessel endothelial hyaluronan receptor-1) and PDPN+ (Podoplanin) cells (Figures S3J–S3M). Taken together, these data suggest that the activated HSCs are driving the observed EMT signature in the bulk Dp16 liver analysis and likely contribute to the liver fibrosis and DR in these animals.

Increased copy number of the interferon receptor locus is not a major contributor to Dp16 liver dysfunction

We and others have shown that T21 consistently activates the IFN response, driven at least in part by triplication of four of the six IFN receptors (IFNRs) that are encoded on chr21.^5,6 The Dp16 mouse model harbors a triplication of the four MMU16-encoded Ifnrs, and dysregulated IFN signaling in this model contributes to various pathophysiological features.⁴⁸ GSEA signatures of IFN gamma response, along with additional inflammatory signatures, were enriched in the Dp16 liver bulk transcriptome analysis (Figures 4A and S4A), typified by Cxcl9, which encodes an IFN-stimulated chemokine (Figure 4B). Furthermore, IFN gamma response was dysregulated in most Dp16 cell types in the scRNA-seq analysis, with the exceptions of endothelial cell cluster 2 and cholangiocytes (Figure 4C). The greatest upregulation was in neutrophils, monocytes, and in the two Kupffer populations (Figure 4C). Cxcl9 expression was most significantly dysregulated in endothelial cell cluster 1 (Figures 4D and 4E).

Figure 4. Increased copy number of the interferon receptor locus is not a major contributor to Dp16 liver dysfunction.

(A) Heatmap showing the median Z score of the leading-edge genes from the IFN gamma response gene set from wild-type and Dp16 liver bulk transcriptome analysis.

(B) Sina plot displaying relative expression (RPKM) of Cxcl9 in wild-type and Dp16 liver bulk transcriptome analysis.

(C) UMAP plot displaying the IFN gamma response gene set normalized enrichment score (NES) for each cluster.

(D) Bubble plot displaying the mean expression and the percent of cells per cluster expressing Cxcl9 in wild-type and Dp16 mice.

(E) Sina plot displaying the log₂ of the normalized counts of Cxcl9 expression within endothelial cell cluster 1 in wild-type and Dp16 animals.

(F) Representative images of H&E-stained liver sections from adult wild-type, Dp16, and Dp16^2xIFNRs mice. Scale bars, 100 μm.

(G) Sina plots showing liver pathology scoring from adult wild-type (n = 13, 6 females), Dp16 (n = 12, 5 females), and Dp16^2xIFNRs (n = 13, 7 females) mice. p values, determined by a Mann-Whitney U test, are shown.

(H) Sina plot displaying the grams/deciliter of plasma albumin from adult wild-type (n = 6, 4 females), Dp16 (n = 6, 2 females), and Dp16^2xIFNRs (n = 8, 1 female) mice. p values, as determined by a Mann-Whitney U test, are shown.

(I) Heatmap displaying the median Z score of plasma bile acids in adult wild-type (n = 9, 4 females), Dp16 (n = 13, 6 females), and Dp16^2xIFNRs (n = 10, 5 females) mice.

For (B and E), Benjamini-Hochberg adjusted p value (q value) are indicated.

For (B, E, and G), individual data are presented with a bar at the median.

To examine the contribution of Ifnr triplication in the liver pathology of DS, we employed the Dp16^2xIfnrs mouse model.⁴⁸ This modified Dp16 mouse model has a normalized copy number of the four Ifnrs—from three copies back to two—while maintaining the rest of the segmental duplication. Histopathological analysis of these mice revealed that normalization of the Ifnrs did not rescue the liver pathology observed in adult Dp16 mice (Figures 4F and 4G; Table S6A). The adult Dp16^2xIfnrs animals surprisingly had significantly higher inflammation scores compared to both WT and Dp16 mice and trended toward increased reactive changes and overall scores (Figure 4G). At 1 month of age, Dp16^2xIfnrs scores trended lower than Dp16 scores (Figures S4B and S4C; Table S6B). Dp16^2xIfnrs animals did not have increased plasma albumin abundance compared to Dp16 animals (Figure 4H; Table S6C) nor did they have decreased liver fibrosis compared to Dp16 animals (Figure S4D). One-month-old Dp16^2xIfnrs also did not have decreased levels of fibrosis compared to Dp16 animals (Figure S4E; Table S6D). Importantly though, Dp16^2xIfnrs animals retained elevated levels of plasma BAs (Figure 4I; Tables S6E and S6F). Overall, these data suggest that the increased Ifnr gene dosage in the Dp16 mouse model is not a major driver of the observed liver pathology and BA metabolism dysregulation.

Metabolic dysfunction in the Dp16 liver is associated with specific cell populations

Next, we examined the metabolic signatures negatively enriched in Dp16 compared to WT, including cholesterol homeostasis, adipogenesis, bile acid metabolism, and fatty acid metabolism (Figures 3A and 5A). Within the cholesterol homeostasis pathway, significantly downregulated genes include farnesyl pyrophosphate (Fdps) and NAD(P)-dependent steroid dehydrogenase-like (Nsdhl) (Figure 5B). Fdps is involved in the later stages of the cholesterol biosynthesis pathway, producing a key cholesterol precursor, farnesyl pyrophosphate. Nsdhl encodes an enzyme responsible for converting lanosterol into cholesterol.⁴⁹ Within the bile acid metabolism signature, two significantly downregulated leading-edge genes include Solute Carrier Family 23 Member 1 (Slc23a1) and Cytochrome P450 46A1 (Cyp46a1) (Figure 5B). Slc23a1 encodes a protein important for maintaining vitamin C levels in the body. These data are consistent with a strong impact of Dp16 gene triplication on hepatic metabolism.

Figure 5. Liver dysfunction in Down syndrome is driven by hepatocyte dysfunction.

(A) Heatmaps showing the median Z score of the leading-edge genes from the cholesterol homeostasis, adipogenesis, bile acid metabolism, and fatty acid metabolism GSEA signatures in whole liver tissue.

(B) Sina plots displaying the log₂(RPKM) of various genes. Benjamini-Hochberg adjusted p values (q values) are indicated.

(C) UMAP plot displaying the cholesterol homeostasis gene set NES in each cluster.

(D) Bubble plot displaying the mean expression and the percent of cells per cluster expressing Fasn in both wild-type and Dp16 mice.

(E) Sina plot displaying the log₂ of the normalized counts of Fasn expression within hepatocyte cluster 2 in wild-type and Dp16 animals.

(F) UMAP plot displaying the bile acid metabolism gene set NES in each cluster.

(G) Bubble plot displaying the mean expression and the percent of cells per cluster expressing Agxt in both wild-type and Dp16 mice.

(H) Sina plot displaying the log₂ of the normalized counts of Agxt expression within hepatocyte cluster 2 in both wild-type and Dp16 animals.

(I) Schematic illustrating iPSC-derived hepatocyte differentiation protocol, with microscopy images of cells in each stage. Scale bars, 50 μm.

(J) Sina plots displaying the gene expression of Alb and Cyp3a4 relative to disomic iPSCs. Values from control cells are shown in gray and from T21 cells are shown in blue. p values, as determined by a Mann-Whitney U test, are shown.

(K) Representative images of albumin (green) and DAPI (blue) immunofluorescent staining of D21 and T21 iPSC-derived hepatocytes. Sina plot displaying the concentration of albumin in the supernatant of iPSC-derived hepatocytes.

(L) Sina plot displaying the concentration of albumin in the supernatant of iPSC-derived hepatocytes.

(M) Sina plot displaying the concentration of urea in the supernatant of iPSC-derived hepatocytes.

(N) Heatmap displaying the enrichment of various GSEA gene sets in whole mouse liver, hepatocytes, and two single-cell hepatocyte clusters. Asterisks indicate significance after Benjamini-Hochberg correction for multiple testing (q value< 0.1).

(O) Heatmap showing the median Z score of the leading-edge genes from the bile acid metabolism GSEA pathway in the iHepatocytes.

(P) Sina plot displaying the log₂(RPKM) of CYP27A1.

(Q) Heatmap displaying the log₂(fold change) of bile acids detected in the supernatant of trisomic versus disomic iPSC derived hepatocytes. Color coding indicates a positive (red) or negative (blue) log₂(foldchange).

(R) Sina plot displaying the log₂(relative abundance) of taurolithocholic acid detected in the supernatant of disomic and trisomic iPSC-derived hepatocytes. Nominal and Benjamini-Hochberg adjusted p values (q values) are indicated.

For (E, H, and P), q values determined by DEseq2 are shown.

For (J–M), p values determined by a Mann-Whitney U test are shown.

For (J–R), four separate cell lines were used per genotype, with 2–4 replicates per line.

When we explored which cell types were associated with metabolic dysfunction, we found that cholesterol homeostasis was negatively enriched in several single-cell clusters, including hepatocyte clusters, cholangiocytes, and Kupffer cells (Figure 5C). Fasn, which encodes fatty acid synthase, is most significantly dysregulated in the same clusters that exhibit the greatest perturbations in the cholesterol homeostasis signature: cholangiocytes, the hepatocyte clusters, and Kupffer cells (Figures 5D and 5E). Alterations in the bile acid metabolism signature were highly discordant across cell types, despite being strongly downregulated in the bulk transcriptome data (Figures 5A and 5F). Whereas this signature was negatively enriched in endothelial cells, for example, it is positively enriched in both hepatocyte clusters (Figures S3B and 5F). Hepatocyte specific perturbations are apparent in the top-ranked genes, including Agxt and Apoa1 (Figures 5G, 5H, S5A, and S5B). These data suggest that the Dp16 model exhibits intrinsic deficits in maintaining cholesterol, fatty acid, and BA homeostasis and that these deficits are driven by dysfunctional hepatocytes.

Given the impact of hepatocyte function on the liver, we next examined cell-intrinsic features of hepatocyte dysfunction by differentiating a panel of iPSCs from individuals with T21 and controls into hepatocyte-like cells (iHepatocytes) (Figure S5C).⁵⁰ Successful differentiation was tracked using RT-qPCR to measure the expression of stage-specific markers (Figure 5I; Tables S7A and S7B), and expression of mature hepatocyte genes ALB and CYP3A4 were reduced in T21 iHepatocytes compared to controls (Figure 5J). Next, we performed a series of assays to confirm functionality and maturity of the iHepatocytes (Tables S7C–S7E). Both D21 and T21 iHepatocytes stained positively for carbohydrate storage as measured by periodic acid Schiff (PAS) staining, suggesting both groups of iHepatocytes were mature (Figure S5D). Immunofluorescence staining for ALB revealed production of this protein in both D21 and T21 iHepatocytes, with a trend toward lower levels in T21 cells (Figure 5K), whereas secreted ALB was significantly lower in T21 cells compared to controls (Figure 5L). Oil Red O staining for lipid accumulation revealed elevated lipid storage in T21 iHepatocytes compared to controls (Figure S5E). Finally, we measured secreted urea and found that while both karyotypes had urea present in the supernatant, T21 iHepatocytes had less than their D21 counterparts (Figure 5M). Taken together, these data are consistent with T21 iHepatocytes that are mature, albeit less functional than D21 iHepatocytes.

Transcriptome analysis of T21 iHepatocytes revealed hundreds of DEGs relative to controls (Figure S5F; Tables S7F and S7G). GSEA was dominated by down-regulation of cell-cycle-related gene sets such as E2F Targets and G2M Checkpoint, which reflect diminished proliferative capacity in the T21 iHepatocytes (Figures S5G and S5H). Next, we compared normalized enrichment scores (NESs) for the metabolic “module” from the bulk liver transcriptome and single-cell hepatocyte clusters 1 and 2 to the iHepatocytes and observed widespread discordance between the bulk and hepatocyte-specific analyses. For example, BA metabolism and adipogenesis were downregulated in the bulk and upregulated in all three hepatocyte-specific analyses, whereas cholesterol metabolism was downregulated in all but the iHepatocytes (Figure 5N). Closer examination of top-ranked genes for BA metabolism in the iHepatocytes revealed strong upregulation of genes shown to be central to this process, including CYP27A1, which encodes for the enzyme that controls the rate-limiting step in BA biosynthesis (Figures 5O and 5P).

Finally, given the observed elevation of plasma BAs in individuals with DS and that hepatocytes are the cells that produce these BAs in vivo, we measured BA abundance in both iHepatocytes and supernatants (Tables S7H and S7I). We observed a trend toward an increase in numerous BAs in the iHepatocytes, as well as the supernatant (Figures 5Q, 5R, and S5I–S5K). Together, these data suggest that T21 is associated with cell-intrinsic dysregulation of hepatocyte function, where expression of key mature hepatocyte genes appears to be diminished, while other hepatocyte features, such as lipid storage, are elevated.

Liver pathology in Dp16 mice is dependent on diet

Given that metabolic homeostasis is sensitive to diet, we investigated the impact of diet on BA metabolism and liver function in Dp16 mice. Animals were fed either a high-fat diet (HFD, 60 kcal% fat) or a low-fat diet (LFD, 10 kcal% fat) for 2 months, beginning at 1 month of age (Figure S6A). As expected, both WT and Dp16 animals on HFD gained significantly more body weight over the course of treatment than the animals on LFD, with a trend toward increased liver mass (Figures S6B and S6C; Tables S8A and S8B). Dp16 animals were heavier than WTs after 2 months on LFD but not HFD (Figure S6B). HFD-fed Dp16 animals exhibited significant elevations in inflammation and reactive changes, as was the case for animals fed standard chow (Figures 6A and 6B; Table S8C). Notably, while Dp16 animals do not develop steatosis or obesity on their standard diet, which contains 16 kcal% fat (Figure S2G), HFD-fed Dp16 animals showed a significant elevation in steatosis (Figure 6B). Strikingly, Dp16 animals on LFD exhibited no elevation in liver pathology relative to WT (Figures 6B and S6D). Finally, none of the treatment groups had increased fibrosis (Figure S6E), and overall pathology was elevated in both sexes (Figure S6F).

Figure 6. Liver pathology in Dp16 mice is dependent on diet.

(A) Representative images of H&E-stained liver sections from wild-type and Dp16 mice fed either an LFD or an HFD. Scale bars, 100 μm.

(B) Sina plots showing the metrics of liver pathology scoring from wild-type and Dp16 mice fed either an LFD or an HFD (wild-type LFD n = 10, 4 females, Dp16 LFD n = 8, 4 females, wild-type HFD n = 14, 4 females, Dp16 HFD n = 13, 4 females). Individual data are presented with a bar at the median, and p values, as determined by a Mann-Whitney U test, are shown.

(C) Volcano plot summarizing results of plasma metabolomic analysis of Dp16 LFD mice versus wild-type LFD mice (wild-type LFD n = 10, 4 females, Dp16 LFD n = 8, 4 females).

(D) Volcano plot summarizing results of plasma metabolomic analysis of Dp16 HFD mice versus wild-type HFD mice (wild-type HFD n = 13, 4 females, Dp16 HFD n = 13, 4 females). Significantly differentially expressed (q < 0.1) genes are red.

(E) Scatterplot comparing fold changes of plasma metabolites from Dp16 LFD mice versus wild-type LFD mice (x axis) versus Dp16 HFD mice versus wild-type HFD mice (y axis). Points are colored according to significance in both groups (red), LFD group only (green), HFD group only (yellow), or not significant (gray).

(F) Sina plots displaying the relative abundance of indicated plasma bile acids. Individual data are presented with a bar at the median.

For (C–F), Benjamini-Hochberg adjusted p values (q values) are indicated, with q > 0.1 significant.

Next, we analyzed the plasma metabolome of this cohort. Notably, the global metabolic dysregulation of Dp16 animals fed the standard chow was moderated in the LFD group (Figures 1F, 6C, and 6D; Table S9A). Furthermore, HFD-fed Dp16 animals exhibited strong dysregulation in BA metabolism relative to HFD-fed WT animals (Figure 6D; Table S9B). Direct comparison of dietary impact on the plasma metabolome highlighted this profound impact on BA metabolism in Dp16 animals (Figures 6E, S6F, S6G, and S6H). Closer inspection of individual BAs revealed various modes of dysregulation. For example, tauroursodeoxycholic acid, tauromuricholic acid, and taurocholic acid, all BAs that have been conjugated to taurine to facilitate biliary secretion, showed similar patterns: not differentially abundant in LFD-fed animals and elevated only in HFD-fed Dp16 animals (Figure 6F; Table S9A–S9D). Muricholic acid, in contrast, showed a trend toward lower levels in LFD-fed Dp16 animals compared to LFD-fed WT animals and was decreased in HFD-fed WT animals but not HFD-fed Dp16s (Figure 6F). Glycochenodeoxycholic acid showed similar trends to muricholic acid (Figure 6F). Taken together, these data indicate that alterations in BA metabolism and liver pathology are mediated by diet in the Dp16 mouse model of DS.

Diet shapes the Dp16 hepatic transcriptome

We next analyzed the liver transcriptome of the WT and Dp16 animals under LFD and HFD conditions (Tables S10A–S10D). When comparing Dp16 to WT under either LFD or HFD conditions, we detected thousands of DEGs (Figures 7A and S7A). Comparison of Dp16 chr16-triplicated DEGs demonstrated a strong concordance between the two treatment conditions with 71 of 82 genes (86.6%) significant in both analyses (Figure 7B). This was not the case, however, for DEGs in the non-triplicated region, where only 748 of 3,531 DEGs (21.2%) were significant in both analyses (Figure 7B). GSEA demonstrated genome-wide impacts of diet on the transcriptome in both genotypes (Figures S7B and S7C). We next compared NES from the Dp16 versus WT LFD analysis and the Dp16 versus WT HFD analysis to visualize transcriptomic differences in response to diet and found that many gene sets were perturbed similarly in both diet groups. Interferon and inflammation-related pathways, however, were activated in LFD conditions and repressed or less activated in HFD conditions, clustering in the bottom right quadrant (Figure 7C). Examination of key top-ranked genes from the IFN alpha pathway demonstrated that this differential response is due to IFN activation in WT mice that is blunted in Dp16 (Dp16 versus WT LFD analysis). This observation could explain the lack of rescue by Ifnr copy-number normalization (Figure 4).

Figure 7. A high-fat diet contributes to global transcriptomic changes in the Dp16 liver.

(A) Volcano plots summarizing the results of transcriptome analysis of livers from Dp16 versus wild-type mice fed a low-fat diet (LFD) and from Dp16 mice versus wild-type fed a high-fat diet (HFD) (wild-type LFD n = 10, 4 females, Dp16 LFD n = 8, 4 females, wild-type HFD n = 14, 4 females, Dp16 HFD n = 13, 4 females). Points in blue are in the MMU16 region triplicated in Dp16 mice.

(B) Venn diagrams displaying the overlap in the significant differentially expressed genes in the Dp16 triplicated region between HFD fed mice and LFD mice (top) or in the differentially expressed genes that are not triplicated in Dp16 mice (bottom).

(C) Scatterplot comparing GSEA NES values for signatures in the livers of Dp16 versus wild-type mice on LFD (x axis) and Dp16 versus wild-type mice on HFD diet (y axis).

(D) Scatterplots comparing the fold changes of leading-edge genes from the cholesterol homeostasis, fatty acid metabolism, and adipogenesis hallmark gene sets for Dp16 versus wild-type mice on LFD (x axis) and Dp16 versus wild-type on HFD (y axis).

(E) Sina plots displaying the log₂(RPKM) of various genes. Horizontal bars indicate median values. Benjamini-Hochberg adjusted p-values (q values) are indicated. For (C–E), points are colored according to significant in both groups (red), significant in the LFD group only (green), significant in the HFD group only (yellow), or not significant in any group (gray). Significance is defined as q < 0.1 using GSEA or DESeq2.

Other prominent gene sets that showed diet-specific effects include peroxisome, adipogenesis, fatty acid metabolism, and cholesterol homeostasis. The peroxisome and adipogenesis gene sets were repressed under LFD conditions and significantly activated under HFD conditions. Cholesterol homeostasis was modestly, but not significantly, repressed by LFD but strongly repressed by HFD (Figure 7C). We also found that very few of the top-ranked genes within the cholesterol homeostasis gene set were differentially expressed in both HFD and LFD Dp16 animals relative to WTs. Despite the gene sets not rising to the level of significance at the NES level, several leading-edge genes were significantly downregulated in Dp16 animals under LFD conditions, consistent with a defect in cholesterol homeostasis even in the absence of HFD (Figure 7D). Conversely, numerous unique top-ranked genes were downregulated in Dp16 animals upon HFD, indicating a differential response to diet (Figure 7D), including Idi1 and Sqle, both of which are required for cholesterol biosynthesis, with Sqle controlling the rate-limiting step in the process (Figure 7E).

Fatty acid metabolism genes showed similar patterns to those in cholesterol homeostasis, with genes downregulated in LFD-fed Dp16 that varied from those downregulated under HFD conditions (Figure 7D). For example, Retsat and Ephx1, two genes involved in lipid metabolism, were both downregulated only under LFD-fed conditions, which could contribute to an altered response to HFD (Figures 7D and S7D). Nsdhl, which was downregulated in our 4-month-old cohort fed standard chow, was only downregulated in Dp16 by HFD (Figure 7D). Adipogenesis-related genes were variably impacted by diet. For example, Fabp4, which binds to fatty acids and assists with uptake and metabolism was upregulated in Dp16 liver by either diet (Figure 7D). Notably, among the genes only affected by HFD in this group was Dhcr7, which encodes the enzyme that catalyzes the final step in cholesterol biosynthesis (Figure 7D). Finally, BA metabolism was upregulated by HFD regardless of genotype, with a slightly greater elevation in Dp16 animals (Figure S7B). Taken together with our standard chow fed Dp16 data, these observations are consistent with a broad dysregulation of metabolism in Dp16 animals that is greatly exacerbated by an HFD and rescued by an LFD.

DISCUSSION

Extensive research shows that widespread consequences for altered liver function include renal toxicity,⁵¹ neurological impairment,⁵² pulmonary inflammation,⁵³ and even reproductive deficits,⁵⁴ but it remains unclear how liver dysfunction might contribute to these conditions in the context of T21. The potential links between altered BA metabolism, obesity, and metabolic syndromes in T21 are readily apparent though.^16,55–57 The impact of liver dysfunction on IGF1 levels—which are decreased in individuals with T21 across the lifespan, inversely correlated with neurodegeneration markers, and associated with short stature—represents one way in which liver dysfunction could affect multiple systems within DS.¹⁰ Furthermore, liver disease in the general population is linked to conditions more common in DS, including congenital heart defects,⁵⁸ celiac disease,⁵⁹ hypothyroidism,⁶⁰ alopecia areata,^61,62 and Alzheimer’s disease.⁶³ Finally, altered BA metabolism has been associated with the pathogenesis of various disorders, such as cardiometabolic and inflammatory diseases.¹⁸

Liver disease in the typical population is often accompanied by metabolic abnormalities, such as dyslipidemia, hypoalbuminemia, hypoglycemia, and dysregulated BA metabolism. Dyslipidemia is more common in the population with DS and even more so in those that are obese.⁶⁴ Although there is evidence that obesity is a driver of dyslipidemia in DS,^65,66 other reports show that dyslipidemia occurs regardless of weight.^64,67 However, the multitude of potential contributing features (diet, obesity, and genetic factors, among others) to metabolic alterations in people with DS makes it hard to disentangle the root cause. Here, we provide evidence that, independent of BMI, individuals with DS and Dp16 mice show clear elevations of plasma BAs and that these changes can be modified by diet.

Dysregulated BA metabolism can also impact lipid metabolism. While baseline expression of BA receptors is crucial for maintaining proper lipid metabolism, increased activation through increased levels of BAs disrupts lipid homeostasis and leads to increased lipid storage in the liver.¹⁸ A HFD can significantly disrupt the BA pool and increase the secretion of BAs, showing the influence that diet plays on the delicate balance of this pool.⁶⁸ Our data show a strong relationship among HFD (60 kcal%), increased levels of plasma BAs, and liver steatosis in Dp16 mice.

Dp16 mice fed their standard diet exhibit liver pathology characterized by an increase in inflammation and DR. This phenotype arises when hepatic progenitor stem cells become activated upon injury, and research has shown that DR specifically occurs to reconstruct the biliary tree network by creating new bile-secreting ducts.⁶⁹ The DR in our mouse model is likely a response to diet-induced injury, as this phenotype is not seen in LFD-fed animals. The DR phenotype may also link to dysregulated BA metabolism in our model, as damaged bile ducts lead to elevated plasma BAs, and an LFD also ameliorates the observed elevated plasma BAs.²¹

Although GWAS studies have investigated genetic associations with NAFLD, none have identified a chr21 gene with a significant association to NAFLD.⁷⁰ The gene most significantly associated with NAFLD though is PNPLA3, which regulates fat metabolism. In DS, PNPLA3 is positively associated with liver steatosis, although the study only had 84 participants.⁷¹ Additionally, two chr21 genes that have been shown to be differentially methylated in those with NAFLD include TIAM1 and LSS.⁷² This suggests it is unlikely one chr21 gene driving these phenotypes but rather is driven by the impact of multiple dysregulated genes.

This study lays the foundation for future investigations into how diet impacts liver function in individuals with DS. Despite the fact that genetics contribute to MASLD, an unhealthy diet, a sedentary lifestyle, and excessive calorie intact are the strongest contributors to MASLD.⁷³ Approximately 50% of those with DS are overweight or obese, and a recent study showed that a Mediterranean diet and exercise for 6 months reduced the severity of MASLD in children with DS.^74,75 These details and the data shown here in this study support the notion that individuals with DS should strongly consider following an LFD.

Limitations of the study

One limitation of this study is the use of mice to understand human metabolism. While mice and humans are similar in many regards, they also vary in many ways.⁷⁶ These include differences in mitochondrial density, reactive oxygen species production, brown fat deposition, and the relative mass of metabolically active tissues, such as the liver.⁷⁷ However, obtaining a liver biopsy is extremely invasive, and individuals with DS do not present with the clinical phenotypes required for a liver biopsy.⁷⁸ Additionally, while the Dp16 mouse model is one of the most used models of DS, it does not contain a freely segregating extra chromosome, present in ~95% of cases with DS.⁷⁹ The Dp16 model also does not confer full trisomy of the human chr21 mouse orthologs, but rather around 120 of the protein coding genes on mouse chr16. Another limitation is the examination of the impact of a single HFD at a single time point. While plasma analyses are informative, future studies should expand beyond conventional blood-based methods and include imaging tools such as FibroScan, which are critical to establish liver pathology.^26,80 In addition, future studies should consider the role of the diet and the gut microbiome in BA metabolism.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kelly Sullivan (kelly.d.sullivan@cuanschutz.edu).

Materials availability

This study did not generate any unique materials.

Data and code availability

• All data are available in the manuscript, supplementary tables, or GEO Superseries GSE296748. Previously published plasma proteomics data using the SOMAscan platform are available through Synapse (https://doi.org/10.7303/syn31488781), and plasma metabolomics data can be accessed through Synapse (https://doi.org/10.7303/syn31488782) or Metabolomics Workbench: ST002200. Mouse metabolomics data are available at Metabolomics Workbench: ST004412, ST004413, and ST004414.

• No unique code was generated for this manuscript.

• Additional raw data available upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human Trisome Project participants

All human data in this manuscript were previously reported and reanalyzed here.^6,10 All human study participants were enrolled in the Crnic Institute’s Human Trisome Project (HTP) under a study protocol approved by the Colorado Multiple Institutional Review Board (COMIRB 15–2170 and NCT02864108; see also www.trisome.org). Written informed consent was obtained from all study participants or their legal guardians.

Blood sample collection and processing

The human datasets analyzed herein were derived from peripheral blood samples collected into BD Vacutainer K2 EDTA tubes (BD, catalog no. 366643). Tubes were processed by centrifugation at 700g for 15 min to separate plasma from other blood fractions. Plasma was aliquoted, flash-frozen, and stored at −80°C. Subsequent processing was carried out as described below.

iPSC generation

Four pairs of iPSC lines were used for these studies. These were previously generated age- and sex-matched individuals with and with Down syndrome and have been described in previous publications.^84–86 Lines used here include iLC62–6, derived from a 27.1 year old euploid male; iLC42–3, derived from a 6.1 year old euploid male; iLC67–3, derived from a 28.8 year old euploid female, iLD7–6, derived from a 30.3 year old male with T21, iLD19–4, derived from a 6.2 year old male with T21; iLD7–6, derived from a 25.1 year old female with T21, and the isogenic pair iLD11 #3(D21)/iLD11(2)-1(T21) derived from a 46.7 year old female with T21. Briefly, the iPSCs were generated from urine-derived renal epithelial cells using a cocktail of mRNAs encoding six reprogramming factors (a modified version of Oct4 fused with the MyoD transactivation domain (called M₃O), Sox2, Klf4, cMyc, Lin28A, and Nanog) along with the reprogramming mimic miRNAs −367 and −302 to reprogram cells.⁸⁷

iPSCs were maintained on matrigel-coated plates (Corning, Cat. # 354277) in mTeSR Plus media (StemCell Technologies, Cat. # 100–0276), manually removing any spontaneously differentiating cells. iPSCs were routinely passaged as aggregates using ReLeSR (StemCell Technologies, Cat. # 100–0484). Mycoplasma testing performed through the Functional Genomics Shared Resource at the University of Colorado indicated no mycoplasma contamination. Genomic integrity and retention of the triplicated chromosome 21 in the iPSCs were validated by g-banded karyotype analysis through the University of Colorado Cancer Center Pathology Shared Resource – Cytogenetics Section.

Hepatocyte differentiation

Hepatocyte-like cells (iHepatocytes) were differentiated from a panel of 8 iPSC lines, four control lines and four lines trisomic for chromosome 21. The iPSCs were maintained on matrigel-coated plates (Corning, Cat. # 354277) in mTeSR Plus media (StemCell Technologies, Cat. # 100–0276), manually removing any spontaneously differentiating cells. Genomic integrity and retention of the triplicated chromosome 21 in the iPSCs were validated by g-banded karyotype analysis through the University of Colorado Cancer Center Pathology Shared Resource – Cytogenetics Section.

iPSCs were differentiated into iHepatocytes using previously described protocols with minor variations.⁵⁰ Briefly, iPSCs were singularized with Accutase (Fisher Scientific, Cat. # NC9464543) and seeded into matrigel-coated six-well plates at 2 × 10⁶ cells per well in mTeSR Plus media supplemented with 10uM Y-27632 (StemCell Technologies, Cat. # 72304). After 24 h, when cells were 90–100% confluent, Y-27632 was removed, and the iPSCs were induced to form the definitive endoderm using the STEMDiff Definitive Endoderm Kit (StemCell, Cat. # 05110) according to the manufacturer’s instructions. Day 0 of iHepatocyte induction was considered when the definitive endoderm media was placed on the cells. On day 4 of differentiation, cells were singularized and plated 7.9 × 10⁴ cells/cm² in hepatoblast media supplemented with 10 μM Y-27632. Hepatoblast media consisted of DMEM/F12 (Cytiva, Cat. # SH30023.01) supplemented with 10% KnockOut Serum Replacement (KOSR, Gibco, Cat. # 10828028), 1% GlutaMAX (Gibco, Cat. # 35050061), 1% Non-Essential Amino Acids (NEAA, Gibco, Cat. # 11140050), 1% Dimethyl-sulfoxide (DMSO, Sigma-Aldrich, Cat. #D2650), and 100 ng/mL of Hepatocyte Growth Factor (HGF, Peprotech, Cat. # 100–398H). After 24 h Y-27632 was removed, and the media was changed daily until day 12. From day 12 onward, cells were cultured in hepatocyte maturation media consisting of DMEM/F12, 10% KOSR, 1% NEAA, 1% GlutaMAX, and 100 nM dexamethasone (Sigma-Aldrich, Cat. #D4902). Media was changed daily until day 15.

Animal husbandry and genotyping

Experiments were approved by the Institutional Animal Care and Use Committee at the University of Colorado Anschutz Medical Campus under protocol #00111 and performed in accordance with National Institutes of Health (NIH) guidelines. Dp16 (B6.129S7-Dp(16Lipi-Zbtb21)1Yey/J) mice were originally purchased from the Jackson Laboratory and maintained on the C57BL/6J background in specific pathogen–free conditions. Mice were housed separately by sex in groups of one to five mice per cage under a 14-h light:10-h dark cycle with controlled temperature and 35% humidity and had ad libitum access to food (16 kcal% fat diet) and water. For genotyping, genomic DNA was prepared from 1 to 2 mm of toe, tail, or ear tissue for automated genotyping by reverse transcription polymerase chain reaction with specific probes designed for Lipi and Zfp295 (Transnetyx). Upon termination of the studies, all animals were euthanized by CO₂ asphyxiation and cardiac puncture, then immediately perfused with 1x PBS using a Masterflex L/S Variable-Speed Economy Modular Drive (VWR, cat. #MFLX07559–00). Dp16^2xIfnrs mice were generated as described previously.⁴⁸

High-fat diet experiments

Mice had ad libitum access to either a low-fat diet (6 kcal% fat, D12450J, Research Diets, New Brunswick, NJ) or high-fat diet (60 kcal% fat, D12492, Research Diets) for 8 weeks, beginning at 28 days of age. At the conclusion of the study, mice were euthanized as described above, tissues collected, and either placed into 10% neutral buffered formalin or added to RLT/BME.

METHOD DETAILS

Mass spectrometry–based plasma metabolomics and lipidomics

Plasma samples were thawed on ice and extracted via a modified Folch method (chloroform/methanol/water 8:4:3). Briefly, 20 μL of sample was diluted in 130 μL of liquid chromatography–mass spectrometry grade water, 600 μL of ice-cold chloroform/methanol (2:1) was added, and the samples were vortexed for 10 s. Samples were then incubated at 4°C for 5 min, quickly vortexed, and centrifuged at 14,000g for 10 min at 4°C. The top (i.e., aqueous) phase was transferred to a new tube for metabolomics analysis and flash-frozen. The bottom (i.e., organic) phase was transferred to a new tube for lipidomics analysis and then dried under N₂ flow. Analyses were performed using a Vanquish UHPLC coupled online to a Q Exactive high-resolution mass spectrometer (Thermo Fisher Scientific). Samples (10 μL per injection) were randomized and analyzed in positive and negative electrospray ionization modes (separate runs) using a 5-min C18 gradient on a Kinetex C18 column (Phenomenex) as described.⁸⁸ Data were analyzed using Maven in conjunction with the Kyoto Encyclopedia of Genes and Genomes database and an in-house standard library.

SOMAscan proteomics of human plasma

A total of 125 μL of EDTA plasma was analyzed by SOMAscan using established protocols.⁶ Each of the 4500+ SOMAmer reagents binds a target peptide and is quantified on a custom Agilent hybridization chip. Normalization and calibration were performed according to SOMAscan Data Standardization and File Specification Technical Note (SSM-020). The output of the SOMAscan assay is reported in relative fluorescence units (RFU).

Association of co-occurring conditions with molecular features

The most frequent co-occurring conditions (present in ≥10% of individuals with Down syndrome) were identified from clinical metadata. Associations between each selected condition and the specific proteins or bile acids were evaluated using two-sided Wilcoxon rank-sum tests, with p-values adjusted for multiple testing within each condition by the Benjamini-Hochberg method.

Fibrosis/steatosis GSEA signature

Human plasma SOMAscan proteomic data was obtained from Govaere et al., 2020.⁴¹ Statistical significance was defined as a Benjamini–Hochberg corrected p-value less than 0.05 and an absolute fold change of more than 1.25-fold. Proteins significantly more abundant in those with advanced fibrosis versus mild fibrosis, as determined by a liver biopsy and their respective NAFLD Activity Score, were used for the Fibrosis signature. The same method was used for the steatosis GSEA signature. GSEA then performed using the HTP SOMAscan proteomic data as described below.

Liver histopathology

Formalin-fixed paraffin-embedded (FFPE) pieces of liver were sectioned at 5 μm and stained with hematoxylin and eosin (H&E). Scoring of liver sections used procedures adapted for mice from the validated histological scoring system established by Kleiner et al. 2005.^29,81,89 The pathologist was blind to animal genotype, sex, and treatment group. Histological images were captured on an Olympus BX51 microscope equipped with a 17mp Olympus DP73 high-definition color digital camera using the Olympus CellSens software (Olympus, Waltham, MA).

Multispectral fluorescence immunohistochemistry (MFI)

FFPE liver was sectioned at 5 μm, mounted on glass slides, and stained with antibodies against albumin (hepatocytes, Invitrogen cat#MA5–32531), Lyve1 (liver endothelial cells, Abcam cat#ab218535), Desmin (activated hepatic stellate cells, Abcam cat#ab32362), PDPN (lymphatic endothelial cells, Abcam cat#ab256559), Smooth Muscle Actin (hepatic stellate cells, CST cat#19245S), and DAPI (cell nuclei, Cat# FP1490 Akoya Biosciences). A whole slide scan was performed at 20x on a Vectra Polaris instrument (Akoya Biosciences), then 6 regions of interest (ROIs) per slide were selected using the Phenochart viewer version1.0.12 (Akoya Bioscience). Individual image and batch analysis was performed using inForm software version 2.6 (Akoya Biosciences). The number of cells positive for a specific cell marker as a percent of total cells per ROI was calculated and the mean per animal was used for further analysis.

Picrosirius red staining

FFPE liver was sectioned at 5 μm, mounted on glass slides, pre-treated with Bouin’s Fixative, and stained with Picrosirius Red. Images were obtained under polarized light in a tiling fashion and positive Cy3–42-255 pixels were quantified utilizing the SlideBook software (Intelligent Imaging Innovations, Denver, CO) and then expressed as a percentage of the total pixels examined.

Murine liver RNA extraction and bulk sequencing

Upon euthanasia, livers were immediately extracted and placed in 594 μL of lysis buffer RLT Plus (QIAGEN) and 6 μL of 2-mercaptoethanol (Sigma-Aldrich) for RNA extraction, then stored at −80 C. Total RNA was isolated using the AllPrep Kit (QIAGEN) following the manufacturer’s instructions. Library preparation was carried out using a Universal Plus mRNA Kit Poly(A) (Tecan). Paired-end, 150-bp sequencing was carried out on an Illumina NovaSeq 6000.

Preparation of fixed single cells from mouse liver

Upon euthanasia and cardiac puncture, mouse livers were perfused with cold 1x Dulbecco’s PBS (DPBS), then removed. Livers were minced manually with a scalpel, suspended in DMEM (Corning) with 2 mg/mL Collagenase P (Roche) and 0.2 mg/mL Deoxyribonuclease I (Worthington Biochemical), and incubated with shaking at 37°C for 15 min. Next, the suspension was filtered through a 70 μm cell strainer to remove debris and centrifuged at 100g for 8 min at 4°C to separate primary mouse hepatocytes (pelleted) from the nonparenchymal cells of the liver (supernatant). Nonparenchymal cells were then pelleted by centrifugation at 400g for 7 min at 4°C, resuspended in red blood cell lysis buffer (BD Biosciences) and incubated for 1 min at room temperature to remove red blood cells, filtered through a 40 μm filter, pelleted again by centrifugation at 400g for 7 min at 4°C, and resuspended in 50 μL DPBS with 1% (w/v) BSA (Research Products International). The hepatocyte pellet was resuspended in red blood cell lysis buffer and incubated 1 min at room temperature, filtered through a 40 μm filter, pelleted again by centrifugation at 100g for 8 min at 4°C, and resuspended in 50 μL DPBS with 1% (w/v) BSA. Viability and cell counts for both hepatocyte and nonparenchymal cell fractions were determined using trypan blue and a Countess 2 cell counter (Thermo Fisher Scientific). Per animal, each fraction was then combined at a 1:1 ratio, pelleted by centrifugation, and fixed by resuspending in a 4% (w/v) formaldehyde fixative solution from a Chromium Next GEM Single Cell Fixed RNA Sample Preparation Kit (10X Genomics).

Mouse liver fixed single cell library construction

Fixed liver single cell suspensions were hybridized with the Next GEM Single Cell Fixed RNA Mouse Transcriptome Probe Kit (10X Genomics), according to manufacturer instructions, and loaded into a Chromium X (10X Genomics) microfluidics instrument to generate an oil emulsion with partitioned nanoliter-scale droplets, each ideally containing a barcoded gel bead and a single cell, along with enzyme Master Mix (10X Genomics) for probe pair ligation and gel bead primer barcode extension. The resulting Illumina-compatible sequencing library was subjected to paired-end 150 bp sequencing on a Novaseq 6000 instrument (Illumina) by the Genomics Shared Resource at the University of Colorado Anschutz Medical Campus.

Murine blood plasma analyte analysis

Blood was collected from mice via a cardiac puncture post euthanasia into lithium heparin tubes (Sarstedt). Plasma was isolated from whole blood via centrifugation for 15 min at 700g, then the supernatant was isolated and spun again for 15 min at 2100g.15 μL of plasma per analyte was processed for ALT, AST, ALP, or albumin detection using the DRI-CHEM 7000 Veterinary Chemistry Analyzer (Heska, Loveland, Colorado) in collaboration with the Comparative Pathology Shared Resource at CU-AMC. Significance was determined using a Mann-Whitney test.

Q-RT-PCR

RNA was collected from samples at the iPSC (day 0), iDefinitive Endoderm (day 4), iHepatoblast (day 12), and iHepatocyte (day 15) stages using the RNeasy kit (Qiagen, Cat. #74106) according to the manufacturer’s instructions. cDNA was made using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Cat. # 4368813) according to the manufacturer’s instructions. Quantitative PCR was then performed using Syber Select (Applied Biosystems, Cat. # 4472942) on the Viia7 (Applied Biosystems). Relative gene expression was determined using the delta-delta CT method normalizing to samples from the iPSC stage for albumin or the definitive endoderm stage for CYP3A4, which was undetected in iPSCs. The following oligonucleotide sequences were used in this study. Albumin forward: 5′-GAGACCAGAGGTTGATGTGATG-3′, albumin reverse: 5′-AGTTCCGGGGCATAAAAGTAAG-3′, CYP3A4 forward: 5′-AAGTCGCCTCGAAGATACACA −3′, CYP3A4 reverse: 5′-AAGGAGAGAACACTGCTCGTG-3’.

Immunofluorescence

On day 15 of differentiation, iHepatocytes were fixed in 4% paraformaldehyde (PFA) for 10 min at room temperature, then washed 3x for 5 min in PBS. Samples were permeabilized in 0.1% Triton X-100 for 10 min at room temperature, followed by another 3 × 5 min series of PBS washes, and then blocked for 1 h in a 5% bovine serum albumin solution (BSA, Research Products International, Cat. # A30075). Cells were treated with 5.4 μg/ml of either rabbit anti-human albumin antibody (Proteintech, Cat. # 16475–1-AP) or isotype control (Invitrogen, Cat. # 31235) overnight at 4°C in 0.1% BSA and then washed 3x for 5 min in PBS. Samples were then incubated for 1 h at room temperature in Goat anti-Rabbit IgG, Alexa Fluor 488 (Invitrogen, Cat. # A27034). Samples were washed once in PBS for 5 min and incubated for 10 min at room temperature in 5 μg/ml of Hoescht 33342 (Abcam, Cat. # ab228551). After nuclear staining, samples were washed a final time in DPBS before imaging on a Zeiss LSM 780 at the University of Colorado Anschutz Medical Campus Advanced Light Microscopy Core.

Oil Red O staining

Oil Red O (ORO, Alfa Aesar, Cat. # A12989) was prepared at 3 mg/mL in 100% isopropanol. Immediately before staining, three parts of ORO were mixed with two parts of DI water, filtered through Whatman paper, and used within two hours. Day 15 iHepatocytes were fixed in 4% PFA for 10 min at room temperature, then washed twice in water. Cells were treated with 60% isopropanol for 5 min at room temperature. Isopropanol was replaced with the filtered ORO solution and incubated for 15 min at room temperature. After ORO staining, samples were washed with DI water until no excess stain remained. Images were captured immediately after staining using bright field microscopy.

Urea assay

On day 15 of differentiation, supernatants were collected from iHepatocytes and urea was analyzed using the Urea Assay Kit according to manufacturer’s directions (Abcam cat. #ab83362). Using the standards provided in the kit, a standard curve was created and urea concentration from D21 and T21 Day 15 iHepatocyte supernatant was calculated.

Albumin enzyme-linked immunosorbent assay (ELISA)

On day 15 of differentiation, supernatants were collected from iHepatocytes and albumin was analyzed using the Human Albumin ELISA Kit according to manufacturer’s directions (Invitrogen cat. #EHALB). Using the standards provided in the kit, a standard curve was created and albumin concentration from D21 and T21 Day 15 iHepatocyte supernatant was calculated.

Periodic acid-Schiff staining

Glycogen storage was determined in iHepatocytes using the Periodic Acid-Schiff Staining System (Sigma-Aldrich, Cat. # 395B-1KT) following the manufacturer’s protocol. Briefly, cells were fixed in 4% paraformaldehyde solution for 10 min at room temperature then rinsed in water. The fixed cells were treated with the Periodic Acid Solution for 5 min, followed by several rinses in distilled water. Next, the cells were immersed in Shiff’s reagent for 15 min followed by a 5-min rinse in running tap water. Cells were counterstained for 2 min using Hematoxylin Solution Gill No. 3 then rinsed in running tap water for 30 s before imaging. Bright field images were captured on an Olympus IX71 microscope equipped with an Olympus DP73 camera.

QUANTIFICATION AND STATISTICAL ANALYSES

Data preprocessing, statistical analysis, and plot generation for all datasets were carried out using R as detailed below or Prism GraphPad. All figures were assembled in Adobe Illustrator. Murine liver pathology scoring and blood pathology measurements were assessed for statistical significance by Mann Whitney U test. Multiplexed immunofluorescence and picrosirius red staining of the liver were assessed for statistical significance by Mann Whitney U test. qPCR of the various genes across the iHepatocyte differentiation process were assessed for statistical significance by Mann Whitney U test. The MFI of albumin staining within the iHepatocytes was assessed for statistical significance by Mann Whitney U test. The secreted albumin and Urea in the iHepatocyte supernatant was assessed for statistical significance by Mann Whitney U test.

Analysis of plasma metabolomics and lipidomics data

Metabolite relative abundance values were imported to R. Metabolites with zero values were replaced with a random value sampled from between 0 and 0.5x the minimum nonzero intensity value for that metabolite. For downstream analysis, data were then normalized using a scaling factor derived by dividing the global median intensity value across all metabolites by each sample median intensity. Extreme outliers were classified per karyotype and per analyte as measurements more than three times the interquartile range below or above the first and third quartiles, respectively, and excluded from further analysis. Differential abundance analysis for metabolites was performed using linear regression in R with log₂-transformed relative abundance as the outcome/dependent variable; T21 status as the predictor/independent variable; and age, sex, BMI, and sample source as covariates. Multiple hypothesis correction was performed with the Benjamini-Hochberg method with an FDR threshold of 10% (q < 0.1). Before visualization or correlation analysis, metabolite data were adjusted for age, sex, BMI, and sample source using the removeBatchEffect() function from the limma package.

Analysis of SOMAscan proteomics data

Normalized data (RFU) in the SOMAscan adat file format were imported to R using the SomaDataIO R package (v3.1.0). Extreme outliers were classified per karyotype and per analyte as measurements more than three times the interquartile range (IQR) below or above the first and third quartiles, respectively (below Q1 − 3 × IQR or above Q3 + 3 × IQR), and excluded from further analysis. Differential abundance analysis for SOMAscan proteomics was performed using linear regression in R with log₂-transformed relative abundance as the outcome/dependent variable; T21 status as the predictor/independent variable; and age, sex, BMI, and sample source as covariates. Multiple hypothesis correction was performed with the Benjamini-Hochberg method using an FDR threshold of 10% (q < 0.1). Before visualization or correlation analysis, SOMAscan data were adjusted for age, sex, BMI, and sample source using the removeBatchEffect() function from the limma package (v3.44.3).

Analysis of bulk murine liver RNA-sequencing data

Reads were demultiplexed and converted to FASTQ format using bcl2fastq v2.20.0.422. Data quality was assessed using FASTQC v0.11.5 and FastQ Screen v0.11.0. Filtering of low-quality reads was performed using bbduk from BBTools v37.99 and fastq-mcf from ea-utils v1.05. Alignment to the mouse GRCm38 reference genome index and Gencode M24 annotation GTF was carried out using HISAT2 v2.1.0. Alignments were sorted and filtered for mapping quality (MAPQ >10) using SAMtools v1.5. Gene-level count data were quantified using HTSeq-count v0.6.1. RNA-seq data yield was a minimum of ~30 million raw reads. Differential gene expression was evaluated using DESeq2 v1.28.1 with surrogate variables, determined using the svaseq() function from the sva package (version 3.46.0), as covariates to remove unwanted sources of variation, including sex and batch. Significance was set at q < 0.1 (10% FDR). GSEA was carried out in R using the fgsea package (v 1.14.0), using Hallmark gene sets on a ranked list of log₂-transformed fold changes.

Analysis of mouse liver single cell gene expression data

Multiplexed reads in FASTQ format were processed using Cellranger-8.0.1 in ‘multi’ mode with the mm10–2020-A transcriptome reference and Chromium Mouse Transcriptome Probe Set v1.0.1. From a total of ~2.5 × 10 ^9 reads, ~155,190 single cells were detected, with a mean of 16,141 reads per cell and 90.66% of cell barcodes passing high occupancy GEM filtering.

Cellranger-filtered per-sample barcodes, features, and matrices were read in to R as sparse matrices using the Read10X() function from the Seurat package,⁸² and per-sample Seurat objects created using the CreateSeuratObject() function, filtering to features (genes) detected in at least 3 cells (min_cells = 3) and cells with at least 200 detected features (min_features = 200). The median number cells obtained per sample was 9066 (range 6242–9961), with a median RNA/gene counts per cell of 4363 (range 495–414275), median features per cell of 2309 (range 208–11096), and median percent mitochondrial reads of 0.4 (range 0–88.6). Next, cells with detected features or mitochondrial read percentages greater than the overall median + three standard deviations were removed (N features >6398, 1.5% of total cells, and percent >6.49, 0.8% of total cells, respectively).

Normalization was carried out using the SeuratSCTransform() function, with regression of mitochondrial percentage as a potential confounder. Integration of the SCTransform-normalized data across samples was carried out using the RPCA method of the SeuratIntegrateLayers() function. Clustering of cells was carried out using SeuratFindNeighbors(reduction = “integrated.rpca”, dims = 1:30) followed by SeuratFindClusters() at a range of resolutions, and cluster cell counts and percentages per sample were obtained with the help of the tidyseurat and tidyverse/dplyr packages.⁸³ Cluster stability and optimal resolution were assessed using the clus-tree package.⁹⁰ For visualization, non-linear dimensional reduction was carried out using SeuratRunUMAP(reduction = “integrated.rpca”, dims = 1:30). Qualitative color schemes were generated using the iwanthue() function from the hues package. Clusters were manually annotated with presumptive cell types based on relative expression level and percentage of cells positive for known marker genes. Changes in cluster proportion by genotype (Dp16 vs. WT) were assessed using beta regression with sex as a confounding variable, and per-cluster adjustment for multiple hypothesis correction across genes using the Benjamini-Hochberg false discovery rate (FDR) method. To examine gene expression differences within cell clusters, per sample per cluster ‘pseudobulk’ gene level RNA counts were obtained using the tidyseurat aggregate_cells() function. Analysis of differential gene expression by genotype (Dp16 vs. WT) based on pseudobulk counts was assessed using the DESeq2 package with adjustment for surrogate variables estimated using the svaseq() function from the combat package. GSEA was performed as described below.

Data visualization

For comparison of data distributions between different categories/groups, sina plots showing all points jittered horizontally by local density, modified with boxes representing medians and interquartile ranges, were generated using ggplot2 and the geom_sina() function from the ggforce R package. Heatmaps were generated using the ComplexHeatmap and tidyheatmap R packages.

Gene set enrichment analysis

GSEA was carried out in R using the fgsea package (v 1.14.0), using Hallmark gene sets, and log₂-transformed fold changes (for RNA-seq), log₂(fold change) multiplied by −log₁₀(p-value) (for SOMAscan proteomics) as the ranking metric.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116835.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Lyve1 antibody	Abcam	Cat# ab218535; RRID:AB_2927473
Albumin antibody	Invitrogen	Cat# MA5-32531, RRID:AB_2809808
Desmin antibody	Abcam	Cat# ab32362, RRID:AB_731901
DAPI	Akoya Biosciences	Cat# FP1490
PDPN antibody	Abcam	Cat# ab256559, RRID:AB_2936436
Smooth Muscle Actin antibody	CST	Cat# 19245S; RRID:AB_2734735
Anti-human Albumin antibody	Proteintech	Cat# 16475-1-AP; RRID:AB_2242567
Isotype control	Invitrogen	Cat# 31235; RRID:AB_243593
Goat anti-Rabbit IgG, Alexa Fluor™ 488	Invitrogen	Cat# A27034; RRID:AB_2536097
Hoechst 33342	Abcam	Cat# ab228551
Chemicals, peptides, and recombinant proteins
10x PBS, pH 7.2	Rockland	Cat# MB-008
Bovine Serum Albumin	Research Products International	Cat# 9048-46-8
Ethylenediaminetetraacetic acid, 0.5M aq. soln, pH 8.0	Thermo scientific	Cat# J62786.AK
DMEM	Corning	Cat# 10-017-CM
Collagenase P	Sigma Aldrich	Cat# 11213857001
Deoxyribonuclease I	Worthington-Biochemical	Cat# LS002145
Lysing buffer	BD Biosciences	Cat# 555899
Matrigel coated plates	Corning	Cat# 354277
mTeSR Plus media	StemCell Technologies	Cat# 100-0276
Accutase	Fisher Scientific	Cat# NC9464543
Y-27632	StemCell Technologies	Cat# 72304
DMEM/F12	Cytiva	Cat# SH30023.01
KnockOut Serum Replacement	Gibco	Cat# 11140050
1% Dimethyl-sulfoxide	Sigma Aldrich	Cat# D2650
Hepatocyte Growth Factor	Peprotech	Cat# 100-398H
Dexamethasone	Sigma Aldrich	Cat# D4902
10x DPBS	Gibco	Cat# 14200-075
2-Mercaptoethanol	Sigma Aldrich	Cat #M6250
Syber Select	Applied Biosystems	Cat# 4472942
Oil Red O	Alfa Aesar	Cat# A12989
GlutaMAX	Gibco	Cat# 35050061
Non-Essential Amino Acids	Gibco	Cat# 11140050
Critical commercial assays
Vacutainer EDTA tubes	BD	Cat# 36643
Allprep DNA/RNA/miRNA Universal Kit	Qiagen	Cat# 80224
STEMDiff Definitive Endoderm Kit	StemCell Technologies	Cat# 05110
Chromium Next GEM Single Cell Fixed RNA Sample Preparation Kit	10x Genomics	Cat# 1000414
Rneasy Kit	Qiagen	Cat# 74106
High-Capacity cDNA Reverse Transcription Kit	Applied Biosystems	Cat# 4368813
Human Albumin ELISA Kit	Invitrogen	Cat# EHALB
Urea Assay Kit	Abcam	Cat# ab83362
Periodic Acid-Schiff Stain	Sigma Aldrich	Cat# 395B-1KT
Deposited data
HTP Sample/Participant metadata and co-occurring conditions	Galbraith et al.6	Synapse: https://doi.org/10.7303/syn31488784
HTP plasma proteomics data	Galbraith et al.6	Synapse: https://doi.org/10.7303/syn31488781
Mouse liver RNA-seq data	This study	GEO: GSE296748
Mouse metabolomics data	This study	Metabolomics Workbench: ST004412, ST004413, ST004414
Experimental models: Organisms/strains
B6.129S7-Dp(16Lipi-Zbtb21)1Yey/J	The Jackson Laboratory	Cat# JAX:013530, RRID:IMSR_JAX013530
Oligonucleotides
Albumin forward: 5′ -GAGACCAGAGGTTGATGTGATG-3′	This paper	-
Albumin reverse: 5′-AGTTCCGGGGCATAAAAGTAAG-3′	This paper	-
CYP3A4 forward: 5′-AAGTCGCCTCGAAGATACACA -3′	This paper	-
CYP3A4 reverse: 5′ -AAGGAGAGAACACTGCTCGTG-3′.	This paper	-
Software and algorithms
inForm software version 2.6	Akoya Biosciences	RRID:SCR_019155
Phenochart viewer version1.0.12	Akoya Biosciences	RRID:SCR_019156
Graphpad Prism v 10.3.0	GraphPad	RRID:SCR_002798
R	R Foundation for Statistical Computing	v4.3.1; RRID:SCR_001905
R Studio	R Studio, Inc.	v2023.09.1 + 494; RRID:SCR_000432
Bioconductor	Bioconductor	v3.17; RRID:SCR_006442
Tidyverse collection of packages for R	CRAN	RRID:SCR_019186
limma package for R	Bioconductor	v3.56.2; RRID:SCR_010943
SomaDataIO R package	GitHub, SomaLogic, Inc.	v3.1.0; RRID: SCR_022198 https://github.com/SomaLogic/SomaDataIO
SlideBook software	Intelligent Imaging Innovations	RRID:SCR_014423
SVA R package	Bioconductor	version 3.46.0, RRID:SCR_012836
fgsea R package	Bioconductor	v1.14.0; RRID:SCR_020938
ggforce R package	CRAN	v0.4.1; RRID:SCR_021728
ComplexHeatmap	CRAN	v2.4.2; RRID:SCR_017270
tidyheatmap	CRAN	v1.8.1, https://cran.r-project.org/package=tidyHeatmap
FASTQC	Babraham Institute	v0.11.5; RRID:SCR_014583, https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
FastQ Screen	Babraham Institute	v0.11.0; RRID:SCR_000141, https://www.bioinformatics.babraham.ac.uk/projects/fastq_screen/
bbduk/BBTools	Bushnell et al.81	v37.99; RRID:SCR_016968, https://jgi.doe.gov/data-and-tools/bbtools/
HISAT2	Kim et al.82	v2.1.0; RRID:SCR_015530, https://daehwankimlab.github.io/hisat2/
Samtools	N/A	v1.5; RRID: SCR_002105, http://www.htslib.org/
HTSeq-count	N/A	v0.6.1; RRID:SCR_005514, https://htseq.readthedocs.io/en/master/
DESeq2 R package	Bioconductor	v1.28.1; RRID:SCR_015687
Olympus cellSens software	Olympus	RRID:SCR_014551
Cellranger	10x Genomics	RRID:SCR_017344
Seurat R package	Stuart et at.83	RRID:SCR_016341
Tidyseurat R package	CRAN	https://github.com/stemangiola/tidyseurat
Clustree R package	CRAN	RRID: SCR_016293
Other
Masterflex L/S Variable-Speed Economy Modular Drive	VWR	Cat# MFLX07559-00
High-fat diet	Research Diets	Cat# D12492
Low-fat diet	Research Diets	Cat# D12450J
Vectra Polaris	Akoya Biosciences	RRID:SCR_025508

Highlights.

• Individuals with Down syndrome and Dp16 mice have elevated plasma bile acids

• Dp16 mice show baseline liver pathology, with ductular reaction and inflammation

• Cell-type-specific alterations define a dysregulated Dp16 liver transcriptome

• A high-fat diet exacerbates liver pathology and bile acid dysregulation in Dp16 mice