Inhibition of HSD17B13 protects against liver fibrosis by inhibition of pyrimidine catabolism in nonalcoholic steatohepatitis

Significance

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease, in which the prognosis is determined by liver fibrosis. A loss-of-function variant in hydroxysteroid 17-beta dehydrogenase 13 (HSD17B13) is associated with decreased liver fibrosis, but the underlying mechanisms remain unclear. Here, we demonstrate that protection against liver fibrosis conferred by the variant in humans and by Hsd17b13 knockdown in mice is associated with decreased pyrimidine catabolism at the level of dihydropyrimidine dehydrogenase. Two common mouse models of NAFLD are characterized by a marked hepatic pyrimidine depletion. Furthermore, pharmacological inhibition of pyrimidine catabolism by a dihydropyrimidine dehydrogenase inhibitor phenocopies the protection against liver fibrosis. Our data suggest pyrimidine catabolism as a therapeutic target in NAFLD.

Abstract

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease, in which prognosis is determined by liver fibrosis. A common variant in hydroxysteroid 17-beta dehydrogenase 13 (HSD17B13, rs72613567-A) is associated with a reduced risk of fibrosis in NAFLD, but the underlying mechanism(s) remains unclear. We investigated the effects of this variant in the human liver and in Hsd17b13 knockdown in mice by using a state-of-the-art metabolomics approach. We demonstrate that protection against liver fibrosis conferred by the HSD17B13 rs72613567-A variant in humans and by the Hsd17b13 knockdown in mice is associated with decreased pyrimidine catabolism at the level of dihydropyrimidine dehydrogenase. Furthermore, we show that hepatic pyrimidines are depleted in two distinct mouse models of NAFLD and that inhibition of pyrimidine catabolism by gimeracil phenocopies the HSD17B13-induced protection against liver fibrosis. Our data suggest pyrimidine catabolism as a therapeutic target against the development of liver fibrosis in NAFLD.

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease, affecting more than 25% of the global population (1–3). It is strongly associated with obesity and predisposes to type 2 diabetes and cardiovascular disease (1, 4). Within the liver, NAFLD covers a spectrum from simple steatosis to nonalcoholic steatohepatitis (NASH), fibrosis, and cirrhosis (1). The interindividual liability of NAFLD to progress is highly variable and the underlying mechanisms remain unclear.

A recent study identified a genetic variant in hydroxysteroid 17-beta dehydrogenase 13 (HSD17B13, rs72613567-A) to be associated with a reduced risk of chronic liver disease and progression of steatosis to steatohepatitis and fibrosis in NAFLD, alcohol-related liver disease and chronic hepatitis C infection (5). The variant appears to specifically protect against advanced forms of liver disease as it does not alter liver fat content or insulin sensitivity (5–7). HSD17B13 is a lipid droplet protein, which is mainly expressed in the liver (5). The A variant encodes a prematurely truncated protein, consistent with loss-of-function (5). HSD17B13 is suggested to have enzymatic activity against multiple lipid species such as steroids, eicosanoids, and retinoids in vitro (5, 7, 8). We have previously shown that the variant associates with increased phospholipids in the human liver (6). However, the mechanism(s) by which the variant confers protection against the progression of liver disease and liver fibrosis remains poorly understood.

In the present study, we investigated the effects of the HSD17B13 rs72613567-A variant in the human liver and Hsd17b13 knockdown (KD) in mice by using a state-of-the-art metabolomics approach.

Results

HSD17B13 rs72613567-A Protects against Liver Fibrosis in Humans.

To investigate the effect of the HSD17B13 rs72613567-A on liver histology, we recruited 21 obese carriers (“A/AA”) and 27 non-carriers (“0”) of the A allele (Table 1). The groups were similar with respect to age, sex, body mass index (BMI), homeostatic model assessment of insulin resistance (HOMA-IR) and PNPLA3, TM6SF2, MBOAT7, and MARC1 genotypes (Table 1). The prevalence of hepatocellular steatosis, ballooning, and inflammation were comparable between the groups, while carriers had markedly lower prevalence of liver fibrosis and a lower overall severity of NAFLD as determined by the Steatosis-Activity-Fibrosis (SAF) score (Table 1 and Fig. 1 A–D).

Table 1.

Clinical characteristics of the participants grouped by the HSD17B13 rs72613567 genotype

	0 (n = 27)	A/AA (n = 21)	P-value
Age, y	47.6 ± 1.9	45.6 ± 2.0	0.46
Women, n	18	16	0.48
Body mass index, kg/m2	44.1 ± 1.2	42.0 ± 1.4	0.26
Waist–hip ratio	0.96 ± 0.02	0.96 ± 0.03	0.87
fP-glucose, mmol/L	5.5 ± 0.1	5.7 ± 0.1	0.30
fS-insulin, mU/L	14.3 ± 1.6	13.1 ± 1.5	0.56
HOMA-IR	3.5 ± 0.4	3.3 ± 0.4	0.64
Hemoglobin A1C, %	5.3 ± 0.1	5.5 ± 0.1	0.19
fP-total cholesterol, mmol/L	4.2 ± 0.2	4.7 ± 0.2	0.04
fP-HDL cholesterol, mmol/L	1.22 ± 0.06	1.17 ± 0.08	0.61
fP-LDL cholesterol, mmol/L	2.6 ± 0.2	3.1 ± 02	0.08
fP-triglycerides, mmol/L	1.17 ± 0.09	1.26 ± 0.11	0.52
P-ALT, U/L	33 ± 3	38 ± 5	0.41
P-AST, U/L	29 ± 2	31 ± 2	0.69
P-GGT, U/L	35 ± 4	39 ± 7	0.66
P-albumin, g/L	38.8 ± 0.5	38.5 ± 0.7	0.74
B-platelets, 109/L	248 ± 10	261 ± 14	0.44
Macrovesicular steatosis, %	10.7 ± 3.8	9.8 ± 3.4	0.85
Steatosis (0/1/2/3), n	13/11/1/2	14/4/3/0	0.17
Ballooning (0/1), n	25/2	21/0	0.10
Inflammation (0/1), n	25/2	21/0	0.10
Activity (0/1/2), n	25/0/2	21/0/0	0.10
Fibrosis (0/1/2), n	18/6/3	18/3/0	0.04
SAF score (0/1/2/3/4/5/6/7), n	8/12/4/1/0/1/0/1	13/3/5/0/0/0/0/0	0.04
HSD17B13 rs72613567 (TT/TTA/TATA), n	27/0/0	0/19/2	<0.001
PNPLA3 rs738409 (CC/CG/GG), n	21/4/2	14/6/1	0.50
TM6SF2 rs58542926 (CC/CT/TT), n	20/6/1	18/3/0	0.50
MBOAT7 rs641738 (CC/CT/TT), n	9/16/2	9/6/4	0.14
MARC1 rs2642438 (GG/GA/AA), n	15/9/2	13/6/2	0.90

Categorical data are in numbers and continuous data in means ± SE. Significances were determined by Student’s t test for continuous variables, Pearson’s chi-squared test for categorical variables and Mantel–Haenszel test for histological variables. f, fasting; P, plasma; S, serum; B, blood; HOMA-IR, homeostatic model assessment of insulin resistance; HDL, high-density lipoprotein; LDL, low-density lipoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyltransferase; SAF score, Steatosis-Activity-Fibrosis score.

Fig. 1.

The HSD17B13 rs72613567-A protects against liver fibrosis and increases hepatic glycerolipids, pyrimidines, and sex steroids in humans. Hepatic (A) steatosis, (B) ballooning, (C) inflammation, and (D) fibrosis in the non-carriers (0) and carriers (A/AA) of the HSD17B13 rs72613567-A variant. Panel E shows the log₂ fold change of liver metabolites in the AA/A as compared to the 0 group. The y-axis denotes the significance [−log₁₀(P)]. Phosphoglycerolipids are highlighted in purple, pyrimidines in red and sex steroids in magenta. Panel F summarizes the changes in hepatic sex hormones between the groups. The concentrations of metabolites with a red upward arrow were increased in the AA/A vs. the 0 group. The pale blue vertical line denotes the step that is catalyzed by the HSD3B1 enzyme in the adrenals and suggests that HSD17B13 might catalyze a similar step in the liver. Panel G localizes the other metabolites, i.e., glycerophospholipids and pyrimidines, on the pyrimidine metabolic pathway. Data are represented as percentages (A). N = 27 in 0 and 21 in A/AA group. P values were determined by Mantel–Haenszel trend test (A–D) and by Student’s t test (E).

HSD17B13 rs72613567-A Increases Hepatic Glycerolipids, Pyrimidines, and Sex Steroids in the Human Liver.

Next, we determined the association of the hepatoprotective A allele with the metabolomic profile of human liver biopsies obtained from the participants (Fig. 1E and SI Appendix, Table S1). The concentrations of glycerolipids (such as 2-palmitoyl-glycerophosphatidylcholine (GPC), 1-palmitoyl-GPC, 1-dihomo-linoleoylglycerol and 1-arachidonoyl-glycerophosphatidylethanolamine (GPE)), pyrimidines (such as uridine 3′-monophosphate, uracil and uridine-2′,3′-cyclic monophosphate), and sex steroids (such as 21-hydroxypregnenolone disulfate, pregnenolone sulfate, dehydroepiandrosterone sulfate and androstenediol monosulfate) were strikingly increased in the livers of the carriers as compared to non-carriers (Fig. 1E).

The differences in hepatic sex steroid concentrations are summarized in Fig. 1F. The observed pattern resembles that seen in deficiency of 3-beta hydroxysteroid dehydrogenase type 1 (HSD3B1) which catalyzes the rate-limiting step in the peripheral conversion of pregnenolone to progesterone, hydroxypregnenolone to hydroxyprogesterone, dehydroepiandrosterone to androstenedione, and androstenediol to testosterone (highlighted in pale blue in Fig. 1F) (9).

The differences in hepatic metabolites are summarized in Fig. 1G, highlighting an increase in uridine metabolism. This pathway supplies nucleotides that are required in the de novo synthesis of phospholipids, RNA, and DNA (10).

Hsd17b13 KD Protects against Liver Fibrosis in Choline-Deficient Mice.

The rs72613567-A variant allele decreases the stability of the HSD17B13 protein and thereby likely confers a loss of function (5). Thus, to model the effect of the hepatoprotective A allele in mice, we compared the Hsd17b13 KD and control (“Ctrl”) C57BL/6J mice during 14 wk of a choline-deficient, amino acid-defined high-fat diet (CDAHFD) (Fig. 2A). Fig. 2B confirms the KD of hepatic HSD17B13 protein by ~60% in the KD group as compared to the Ctrl group. There were no differences in body weight, fat mass, lean mass, plasma glucose or insulin, respiratory exchange ratio (RER), energy expenditure, activity, feeding, or transaminase activities between the groups (Fig. 2 C–M). The histological scores of hepatocellular steatosis, ballooning, and inflammation were comparable between the groups, while the KD group had significantly lower liver fibrosis as compared to the Ctrl group (Fig. 2 N–Q). The decrease in liver fibrosis in the KD group was confirmed by liver hydroxyproline assay (Fig. 2R). The canonical Yap/Taz targets Cyr61, Ctgf, Tgfb1, and Spp1 were decreased in the KD group as compared to the Ctrl group (SI Appendix, Fig. S1).

Fig. 2.

Hsd17b13 KD protects against liver fibrosis in choline-deficient mice. (A) Study design, (B) hepatic Hsd17b13 protein expression, (C) body weights, (D) fat mass, (E) lean mass, (F) plasma glucose concentration, (G) plasma insulin concentration, (H) RER, (I) energy expenditure, (J) physical activity, (K) food intake, (L) plasma ALT, (M) plasma AST activities, (N) hepatic steatosis, (O) ballooning, (P) inflammation, (Q) fibrosis, and (R) liver hydroxyproline concentrations in the Ctrl and the ‘KD’ groups. The mouse in A is from TogoTV (©2016 DBCLS TogoTV/CC-BY-4.0). Data are represented as mean ± SEM. N = 6 to 7 per group (B) or 12 to 13 per group (C–G and N–Q) or 8 per group (H–K) or 10 to 16 per group (N–M) or 10 per group (R). P values were determined by repeated-measures ANOVA test (C) or by Student’s t test (B and D–R). See also SI Appendix, Fig. S1.

Hsd17b13 KD Increases Hepatic Glycerolipids and Pyrimidines in the Mouse Liver.

Next, we determined the effect of the Hsd17b13 KD on the metabolomic profile of the mouse liver (Fig. 3A and SI Appendix, Table S2). The concentrations of hepatic glycerolipids (such as 1-oleoyl-2-linoleoyl-glycerophosphatidylinositol (GPI), 1-linoleoyl-GPC and 1-palmitoyl-2-oleoyl-glycosyl-GPE) and pyrimidines (such as uridine and uridine-2′,3′-cyclic monophosphate) were markedly increased in the KD group as compared to the Ctrl group (Fig. 3A). Conversely, the concentrations of hepatic acetyl-CoA, a product of uridine catabolism, were decreased in the KD group as compared to the Ctrl group (Fig. 3A).

Fig. 3.

Hsd17b13 KD increases hepatic glycerolipids and pyrimidines in the mouse liver. (A) Log₂ fold change in liver metabolites in the KD vs. Ctrl group. The y-axis denotes the significance [−log₁₀(P)]. Phosphoglycerolipids are highlighted in purple, pyrimidines in red and acetyl CoA, a pyrimidine catabolic intermediate, in blue. Panels B–G show hepatic expression of pyrimidine-metabolizing enzymes relative to loading controls (GAPDH and HSP90) in the groups. The hepatic expression of (B) uridine-cytidine kinase (UCK), (C) pyramidine 5′-nucleotidase (NT5C3A), (D) uridine phosphorylase 1 UPP1, (E) dihydropyrimidine dehydrogenase DPYD, (F) dihydropyrimisinase DPYS, and (G) beta-ureidopropionase UPB1. Panel H summarizes the findings in the HSD17B13 KD mouse livers and highlights decreased pyrimidine catabolism as a common finding in both the Hsd17b13 KD mice and HSD17B13 variant carrier humans (Fig. 1). Data are represented as mean ± SEM (B–G). N = 4 per group (A) or 6 to 7 per group (B–G). P values were determined by Student’s t test.

The increase in uridine metabolites and the decrease in acetyl-CoA suggested that uridine catabolism might be altered in the KD group as compared to the Ctrl group, so we next assessed hepatic expressions of uridine-metabolizing enzymes in the groups (Fig. 3 B–G). The protein expression of dihydropyrimidine dehydrogenase (DPYD, an enzyme in the pyrimidine degradation pathway) was ~58% decreased (P = 0.0007) and that of uridine kinase (UCK, an anabolic enzyme in the pyrimidine pathway) was ~52% increased (P = 0.03) in the KD group as compared to the Ctrl group. The expressions of pyrimidine 5′-nucleotidase (NT5C3A), uridine phosphorylase (UPP1), dihydropyrimidinase (DPYS), and beta-ureidopropionase (UPB1) were comparable between the groups (Fig. 3 B–G). Fig. 3H summarizes the findings and highlights decreased uridine catabolism in the livers of the Hsd17b13 KD mice.

Overlap in the Human and Mouse Liver Metabolites Highlights Pyrimidine Metabolism in HSD17B13-Mediated Hepatoprotection.

To further characterize the hepatic metabolite profiles, we performed pathway-level over-representation analyses in the human HSD17B13 rs72613567-A (SI Appendix, Table S3), the mouse Hsd17b13 KD (SI Appendix, Table S4), and their overlap (SI Appendix, Table S5). These analyses identified Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways such as “pyrimidine metabolism,” “aminoacyl-transfer ribonucleic acid (tRNA) biosynthesis,” “mineral absorption,” “protein digestion and absorption,” and “central carbon metabolism in cancer” as common features of the human HSD17B13 rs72613567-A and the mouse Hsd17b13 KD (SI Appendix, Tables S3–S5). Next, we assessed the overlap of individual metabolites in the human HSD17B13 rs72613567-A and the mouse Hsd17b13 KD and identified six common metabolites (SI Appendix, Fig. S2A). Among these six metabolites, uridine-2′,3′-cyclic monophosphate and uridine 3′-monophosphate were more markedly increased both in the human HSD17B13 rs72613567-A (SI Appendix, Fig. S2B) and the mouse Hsd17b13 KD (SI Appendix, Fig. S2C) as compared to the other metabolites. Moreover, these two metabolites were the only ones to correlate inversely with histological fibrosis of the liver (SI Appendix, Fig. S2 D and E).

Hepatic Pyrimidine Deficiency Characterizes Two Mouse Models of NASH.

Since hepatic pyrimidine (e.g., uridine) concentrations were increased in the human carriers of the hepatoprotective HSD17B13 variant allele and in the Hsd17b13 KD mice and correlated inversely with hepatic fibrosis, we asked whether altered pyrimidine metabolism might characterize commonly used mouse models of NASH.

In the first model, we fed C57BL/6J mice either a “CDAHFD” or regular chow (“Chow”) for 14 wk. At the end of the study, the CDAHFD group had marked increases in all features of NASH, including steatosis, inflammation, ballooning, and fibrosis (Fig. 4 A–D). The metabolomics analyses of their livers showed a marked deficiency in virtually all pyrimidines and an increase in the pyrimidine catabolic intermediate, 3-ureidopropionate (Fig. 4E and SI Appendix, Table S6).

Fig. 4.

Hepatic pyrimidine deficiency characterizes two mouse models of NASH. (A) Hepatic steatosis, (B) ballooning, (C) inflammation, and (D) fibrosis in C57BL/6J mice fed regular chow (Chow) or a choline-deficient, amino acid-defined, high-fat diet (CDAHFD). (E) Log₂ fold changes in liver pyrimidines in the CDAHFD vs. Chow group. The y-axis denotes significance [−log₁₀(P)]. Metabolites highlighted in blue were decreased, while those highlighted in red were increased in the CDAHFD vs. Chow. (F) Hepatic steatosis, (G) ballooning, (H) inflammation, and (I) fibrosis in Ob/Ob mice fed regular chow (Chow) or the “GAN” diet. (J) Log₂ fold changes in liver pyrimidines in the GAN vs. Chow group. The y-axis denotes significance [−log₁₀(P)]. Metabolites highlighted in blue were decreased, while those highlighted in red were increased in the GAN vs. Chow. (K) Summarizes the changes in hepatic pyrimidines in the CDAHFD and GAN groups as compared to their respective control groups. CMP, cytidine monophosphate; dCMP, deoxycytidine monophosphate. Data are represented as mean ± SEM (A–D and F–I). N = 8 to 13 per group. P values were determined by Student’s t test.

In the second model, we fed ob/ob mice either the high-saturated fat, high-sugar, high-cholesterol Gubra Amylin NASH (GAN) diet or regular chow for 16 wk. At the end of the study, the GAN group had marked increases in all features of NAFLD, including steatosis, inflammation, ballooning, and fibrosis (Fig. 4 F–I). The metabolomics analyses of their livers showed a marked deficiency in virtually all pyrimidines and an increase in the pyrimidine catabolic intermediate, 3-ureidopropionate (Fig. 4J and SI Appendix, Table S6). Fig. 4K summarizes the findings of the two mouse models of NASH and highlights pyrimidine deficiency as a common feature of NAFLD.

Inhibition of Pyrimidine Catabolism by Gimeracil Protects against Liver Fibrosis.

Finally, we asked whether a pharmacological inhibition of pyrimidine catabolism by gimeracil could prevent CDAHFD-induced liver fibrosis in mice. Gimeracil is a specific inhibitor of DPYD (Fig. 5A) (11), i.e., the same pyrimidine catabolic enzyme that was specifically downregulated in the Hsd17b13 KD mice (Fig. 3E). During 12 wk of treatment, there were no differences in body weight, fat mass, lean mass, plasma glucose or insulin concentrations, RER, energy expenditure, or activity between the groups (Fig. 5 B–I). The food intake was slightly lower in the gimeracil-treated group (Fig. 5J), but this difference did not confer any meaningful effects on, e.g., body composition. With respect to liver histology, the scores of steatosis, inflammation and ballooning were comparable between the groups, while liver fibrosis tended to be ~12% (P = 0.053) lower in the gimeracil-treated group as compared to control (Fig. 5 K–N). A more sensitive method to assess liver fibrosis, the hydroxyproline assay, identified an ~25% (P = 0.02) decrease in the gimeracil-treated group as compared to the Ctrl group (Fig. 5O).

Fig. 5.

Inhibition of pyrimidine catabolism by gimeracil protects against liver fibrosis. (A) Gimeracil inhibits pyrimidine catabolism at the level of DPYD. (B) Body weights, (C) fat mass, (D) lean mass, (E) plasma glucose concentration, (F) plasma insulin concentration, (G) RER, (H) energy expenditure, (I) physical activity, (J) food intake, and histologically determined grades of (K) hepatic steatosis, (L) inflammation, (M) ballooning, and (N) fibrosis in C57BL/6J mice fed the CDAHFD with gimeracil (“Gime”) or without it (Ctrl). (O) Liver hydroxyproline measurements were used as an independent biochemical readout of liver fibrosis. Data are represented as mean ± SEM. N = 8 to 12 per group. P values were determined by repeated-measures ANOVA test (B) or by Student’s t test (C–O).

Discussion

In the present study, we investigated the hepatoprotective effects of the HSD17B13 rs72613567-A variant in human liver and in a Hsd17b13 KD mouse model by applying a state-of-the-art metabolomics approach.

Consistent with previous reports (5, 6), we found that the HSD17B13 variant protected specifically against liver fibrosis independent of steatosis, BMI, HOMA-IR, or other genetic variants (Table 1 and Fig. 1). We also found an increase in hepatic phospholipids, such as 1- and 2-palmitoyl-GPC, in variant carriers as compared to non-carriers (Fig. 1), consistent with our previous study in an independent cohort (6). We extend the previously reported data by showing that in the carriers as compared to non-carriers, hepatic sex steroids and uridine metabolites are markedly altered (Fig. 1).

The observed increases in pregnenolone, hydroxypregnenolone, dehydroepiandrosterone and androstenediol resemble those seen in deficiency of HSD3B1 which catalyzes the rate-limiting step in the peripheral conversion of pregnenolone to progesterone, hydroxypregnenolone to hydroxyprogesterone, dehydroepiandrosterone to androstenedione, and androstenediol to testosterone (highlighted in pale blue in Fig. 1F) (9). HSD17B13 has been reported to have enzymatic activity against multiple lipid species such as steroids, eicosanoids, and retinoids in vitro (5, 7, 8). The present data suggest that HSD17B13 might also play a role in hepatic sex steroid metabolism in humans.

Pathway analyses of the liver metabolites identified changes in multiple pathways, including pyrimidine metabolism, aminoacyl-tRNA biosynthesis, mineral absorption, protein digestion and absorption, and central carbon metabolism in cancer (SI Appendix, Tables S3–S5). However, only pyrimidine metabolites correlated inversely with histological fibrosis stage in the liver (SI Appendix, Fig. S2), highlighting pyrimidines as a potential mediator of HSD17B13-mediated hepatoprotection. This is consistent with previous studies in mice showing protection against NAFLD by uridine-boosting therapies. Decreasing hepatic uridine by overexpressing uridine phosphorylase induced NAFLD in mice, which was prevented by uridine supplementation and by 5-benzylacyclouridine, an inhibitor of uridine phosphorylase, and 5-(2-bromocinyl)uracil, an inhibitor of dihydropyridine dehydrogenase (12). Another study showed that increasing liver uridine by CPBMF65, an inhibitor of uridine phosphorylase, protected against carbon tetrachloride-induced liver fibrosis in mice (13). A third study reported that uridine supplementation prevented tamoxifen-induced NAFLD and stimulated hepatic phospholipid biosynthesis (14).

Since the HSD17B13 variant confers a loss of function, we modeled its effect in mice by an adeno-associated virus (AAV)-mediated Hsd17b13 KD. In addition, we chose the CDAHFD as the NASH-inducing diet based on our previous data which suggested that the variant plays a role in phosphatidylcholine metabolism (Fig. 1) (6). The Hsd17b13 KD in mice very consistently phenocopied the human variant carriers. For instance, the KD mice had: 1) a specific protection against liver fibrosis, 2) no change in body composition or markers of insulin sensitivity, 3) an increase in distinct hepatic phospholipids, and 4) an increase in hepatic uridine metabolites. With the mouse model, we were able to extend the human data by identifying downregulation of DPYD as the key step altered on the pyrimidine catabolic pathway. Interestingly, DPYD expression is under regulation of multiple nuclear receptors, including constitutive androstane receptor (CAR) and pregnane X receptor (PXR) (15), which are activated by sex steroids.

Next, we investigated whether uridine metabolism might play a role in well-established mouse models of NAFLD, i.e., the CDAHFD on C57BL/6J background and the GAN diet on ob/ob background. Strikingly, we found that virtually all pyrimidines were depleted and markers of pyrimidine catabolism were increased in association with severe NAFLD in these two distinct models.

Finally, we asked whether pharmacological inhibition of DPYD-mediated pyrimidine catabolism by gimeracil would phenocopy the HSD17B13 variant and protect against liver fibrosis and indeed, we found this to specifically protect against liver fibrosis despite unchanged body composition and measures of insulin resistance.

Hepatic HSD17B13 is thought to express in hepatocytes, but not in stellate cells which are the ultimate producers of hepatic fibrosis (7). Future studies are needed to investigate whether pyrimidines might have direct effects on stellate cells or whether their anti-fibrotic effects are mediated indirectly.

In summary, we demonstrate that the protection against liver fibrosis conferred by the HSD17B13 rs72613567-A variant in humans and by the Hsd17b13 KD in mice is associated with decreased DPYD-mediated pyrimidine catabolism. We furthermore show that hepatic pyrimidines are depleted in two distinct mouse models of NASH and the inhibition of pyrimidine catabolism by gimeracil phenocopies the HSD17B13 variant-induced protection against liver fibrosis. These findings suggest pyrimidine catabolism as a potential therapeutic target against the development of liver fibrosis in NAFLD.

Materials and Methods

Human Participants.

We recruited a total of 48 participants (Table 1) amongst those undergoing laparoscopic bariatric surgery at the Helsinki University Hospital, Finland. The following inclusion criteria were employed: 1) age between 18 and 75 y; 2) alcohol consumption <20 g/d for women and <30 g/d for men; 3) no known acute or chronic disease other than obesity, type 2 diabetes, NAFLD or hypertension on the basis of medical history, physical examination and blood count, plasma creatinine, and electrolyte concentrations; 4) no clinical or biochemical evidence of liver disease other than NAFLD, or clinical signs or symptoms of inborn errors of metabolism; 5) no history or current use of hepatotoxic compounds; 6) no pregnancy or breastfeeding in women. The study protocol was approved by the ethics committee of the Hospital District of Helsinki and Uusimaa (Helsinki, Finland). The study was conducted in accordance with the Declaration of Helsinki and each participant provided written informed consent after being explained the nature and potential risks of the study.

Human Metabolic Study.

One week prior to surgery, the participants were invited to a separate clinical visit for detailed metabolic phenotyping. After an overnight fast, body weight, height, and circumferences of the waist and hip were measured as described (16). Blood samples were withdrawn from an antecubital vein for measurement of circulating blood count, glucose, insulin, hemoglobin A_1C, total cholesterol, high-density lipoprotein (HDL) cholesterol and low-density lipoprotein (LDL) cholesterol, triglyceride, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), and albumin concentrations and for genotyping as described (17). Plasma glucose concentrations were measured using the hexokinase method, plasma total cholesterol, HDL and LDL cholesterol and triglyceride concentrations were assessed using enzymatic kits and activities of plasma ALT, AST, and GGT were determined using photometric International Federation of Clinical Chemistry methods on an autoanalyzer (Roche Diagnostics Hitachi 917, Hitachi). Serum insulin concentrations were measured by time-resolved fluoroimmunoassay using an Insulin Kit (AUTOdelfia, Wallac). Hepatic insulin resistance was assessed by using the HOMA-IR, calculated using the following formula: serum insulin (mU/L) × plasma glucose (mmol/L)/22.5 (18). Hemoglobin A_1C was determined using an immunoturbidometric method (Abbott Laboratories). Blood counts and platelets were measured using impedance, flow cytometric and photometric assays (XN10, Sysmex). Plasma alb min concentrations were determined by using a photometric method on an autoanalyzer (Modular Analytics EVO, Hitachi). The genotyping of HSD17B13 rs72613567, PNPLA3 rs738409, TM6SF2 rs58542926, MBOAT7 rs641738, and MARC1 rs2642438 was performed using the TacMan polymerase chain reaction method (Applied Biosystems) (17).

Human Liver Biopsies and Liver Histology.

Wedge biopsies of the human liver were obtained at the beginning of the surgery. One part of the biopsy was sent to the pathologist for histological examination and the remaining tissue was flash-frozen in liquid nitrogen and stored at −80 °C for subsequent analysis of the hepatic metabolome. The histological examination was performed by an experienced liver pathologist (J.A.) in a blinded manner according to the SAF score (19).

Liver Metabolomic Analysis.

The non-targeted metabolomics analyses of the human and mouse (see below) liver tissue samples were performed at Metabolon, Inc. Briefly, snap-frozen liver samples were methanol extracted and analyzed by four complementary ultra-high performance liquid chromatography-tandem mass spectrometry methods: 1) acidic positive ion conditions optimized for more hydrophilic compounds, 2) acidic positive ion conditions optimized for more hydrophobic compounds, 3) basic negative ion optimized conditions, and 4) negative ionization following elution from a hydrophilic interaction chromatography (HILIC) column. Metabolites were identified by automated comparison of the ion features in the experimental samples to a reference library. The human and mouse liver datasets comprised a total of 959 and 914 compounds of known identity, respectively. Sample accessioning, preparation, quality control, ultra-high-performance liquid chromatography, bioinformatics, data extraction, compound identification, curation, metabolite quantification, and data normalization were performed by Metabolon, Inc. as described (20). Normalized metabolomics data were analyzed using Student’s t tests between the groups and visualized as volcano plots. The results from the human and mouse liver analyses were compared for overlap in order to validate each other and to avoid false positive findings. Pathway-level over-representation analysis was performed using the Integrated Pathway-Level Analysis of Transcriptomics and Metabolomics data with IMPaLa (21).

Mouse Models.

All mouse studies were approved by the Yale University Institutional Animal Care and Use Committee. Male C57BL/6J and ob/ob mice (Jackson Laboratories) were group housed at the animal care facility at Yale University Animal Research Center and maintained under controlled temperature (23 °C) and lighting (12:12 h light/dark cycle, lights on at 7 a.m.) with ad lib access to water and food.

The KD of mouse Hsd17b13 was performed by tail vein injection of an AAV (AAV8-GFP-U6-m-HSD17B13-shRNA for the KD group and AAV8-GFP-U6-scrmb-shRNA for the control group, both from Vector Biolabs) at a dose of 5e11 GC per mouse.

To induce NAFLD, two distinct experimental designs were used. In C57BL/6J mice, a CDAHFD (60% fat, 0.1% methionine and no choline, A06071302, Research Diets) was used to induce progressive NAFLD (22). In ob/ob mice, the GAN diet was used to induce progressive NAFLD (23) (40% saturated fat, 20% fructose, 2% cholesterol, D09100310, Research Diets, Inc.). The control mice were fed regular chow (18% fat, 58% carbohydrate, 24% protein, TD2018, Harlan Teklad, Envigo). In the gimeracil study, gimeracil (Sigma) was added in the CDAHFD diet at Research Diets at 260 mg gimeracil per kg diet, estimated to result in a dosing of 25 mg/kg/d in the C57BL6/J mice.

Catheters were placed in the jugular vein 7 to 9 d prior to metabolic studies in all mouse studies except the gimeracil study in which no catheters were placed. After an overnight fast, the awake mice were placed under gentle tail restraint for 20 min after which plasma samples were withdrawn from the tail veins. At the end of the study, mice were euthanized with pentobarbital sodium (150 mg/kg) except the gimeracil study in which isoflurane was used for euthanasia. Immediately after euthanasia, tissues were harvested and frozen with Wollenberger tongs precooled in liquid nitrogen. Liver and plasma were stored at −80 °C for subsequent analyses.

Mouse Body Composition and Metabolic Cage Studies.

Indirect calorimetry and assessment of physical activity and food intake were measured in Comprehensive Laboratory Animal Monitoring System metabolic cages (Columbus Instruments). The body composition was assessed by ¹H magnetic resonance spectroscopy (Minispec, Bruker).

Mouse Plasma Biochemical Analysis.

Plasma glucose concentrations were determined using a YSI 2700 analyzer (Yellow Springs Instruments). Plasma insulin concentrations were measured using radioimmunoassay (EMD Millipore) at the Yale Diabetes Research Center. Plasma ALT and AST activities were determined by using a COBAS Mira Plus analyzer (Roche).

Mouse Liver Histology and Biochemical Analysis.

One part of the mouse liver was fixed in formalin for histological analyses and the remaining part was snap frozen in liquid nitrogen using Wollenberger tongs and stored at −80 °C for subsequent metabolomic analyses. The liver histological examinations were performed by an experienced liver pathologist (X.Z.) in a blinded manner according to the NASH Clinical Research Network scoring system (24). Liver fibrosis was quantified biochemically using a hydroxyproline colorimetric assay (Biovision).

Quantitative PCR.

Whole-liver total RNA was isolated using TRIzol reagent (Thermo-Fisher) according to the manufacturer’s directions. This RNA was reversed transcribed using iScript (Bio-Rad) and was used as material to perform reverse-transcription qPCR on a StepOnePlus using Fast Taqman reagents (Thermo-Fisher). messenger ribonucleic acid (mRNA) expression levels were estimated using the 2ΔΔC_T method by using the following probes: Cyr61 (M00487498_m1), connective tissue growth factor (CTGF) (Mm01192933_g1), TGFb1 (Mm01178820_m1), Spp1 (Mm00436767_m1), yes-associated protein 1 (YAP) (Mm01143263 m1), Wwtr1 Tafazzin (TAZ) (Mm01289583_m1), Adgre1 (Mm00802529_m1), Acta2 (Mm00725412_s1), and Col1a1 (Mm_00801666_g1).

Western Blotting.

Proteins from liver lysate were resolved by sodium dodecyl sulfate polyacrylamide gel electroforesis (SDS-PAGE) using a 4 to 14% gradient gel and electroblotted onto polyvinylidiene difluoride membranes (EMD Millipore). Polyclonal primary antibodies were used to detect HSD17B13 (abcam), DPYD (abcam), UPP1 (abcam), UCK (Invitrogen), NT5C3A (Invitrogen), UPB1 (Invitrogen), DPYS (Invitrogen), glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (Cell Signaling), and HSD90 (Cell Signaling). Secondary antibodies were horseradish peroxidase-conjugated and detection was carried out using enhanced chemiluminescence (Cell Signaling Technology).

Statistical Analyses.

Statistical analyses were performed using GraphPad Prism for macOS version 9.2.0 (GraphPad) and SPSS Statistics version 27.0.1.0 (IBM). Groups were compared by using the Student’s unpaired t test, repeated-measures ANOVA test, Pearson’s chi-squared test, or Mantel–Haenszel trend test, as appropriate. All data are reported as mean ± SEM or n. P values less than 0.05 were considered statistically significant.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.