A human liver cell-based system modeling a clinical prognostic liver signature for therapeutic discovery

Abstract

Chronic liver disease and hepatocellular carcinoma (HCC) are life-threatening diseases with limited treatment options. The lack of clinically relevant/tractable experimental models hampers therapeutic discovery. Here, we develop a simple and robust human liver cell-based system modeling a clinical prognostic liver signature (PLS) predicting long-term liver disease progression toward HCC. Using the PLS as a readout, followed by validation in nonalcoholic steatohepatitis/fibrosis/HCC animal models and patient-derived liver spheroids, we identify nizatidine, a histamine receptor H2 (HRH2) blocker, for treatment of advanced liver disease and HCC chemoprevention. Moreover, perturbation studies combined with single cell RNA-Seq analyses of patient liver tissues uncover hepatocytes and HRH2⁺, CLEC5A^high, MARCO^low liver macrophages as potential nizatidine targets. The PLS model combined with single cell RNA-Seq of patient tissues enables discovery of urgently needed targets and therapeutics for treatment of advanced liver disease and cancer prevention.

Fig. 1. Modeling the clinical prognostic liver signature (PLS) in a cell-based system.

a Experimental approach. b Analysis of HCV infection by qRT-PCR (log10 of relative RNA quantity normalized to GAPDH mRNA; mean ± SD, n = 3). Immunodetection of HCV E2 protein 10 days post infection. Scale bar: 50 µm. c PLS assay in Huh7.5.1^dif cells. Heatmap shows expression of poor- and good-prognosis genes (186 gene signature). Results are representative of one out of three independent experiments performed in triplicate. d Analysis of the PLS after HCV Jc1 infection and antiviral treatment. PLS induction was determined by GSEA analysis using “mock” non-infected cells as reference. Simplified heatmaps show PLS global status and PLS poor- and good-prognosis gene expression. Results from two experiments performed in triplicate are shown. FDR false discovery rate, NES normalized enrichment score. e scRNA-Seq profiling reveals HCV-dependent induction of the PLS in Huh7.5.1^dif cells. Heatmaps show: PLS global status and modulation of the poor-prognosis (FDR = 0.013) and good-prognosis genes (FDR = 0.016). White: non-infected cells, n = 17; black: infected cells, n = 23. f Enrichment scores of PLS gene modulation for their correlation with the HCV viral load measured at the single-cell level (Pearson correlation tests). g PLS assays in the cell-based models. Cell-based models were infected with different viruses or exposed to metabolic cues. HBV infection was performed in HepG2-NTCP cells and free fatty acids (FFA) treatment was performed in Huh7.5.1^dif cells cocultured with LX2. Results are from three independent experiments. h PLS in clinical liver tissues from patient cohorts. HCV-related cirrhosis (Italy, n = 216 GSE156546; US, n = 145 GSE541027), HBV-related liver cancer (n = 199, GSE14520), alcoholic hepatitis (n = 22, GSE28619), and NASH (n = 72, GSE49541). Induction of the PLS was assessed by comparing diseased tissues with non-diseased tissues. i Induction of the PLS in cell-based systems was compared to liver transcriptome profiles from clinical cohorts using Subclass Mapping. j PLS deconvolution showing mean gene expression of the poor- and the good-prognosis genes in the main cell compartment within the liver. The scRNA-Seq dataset was extracted from GSE115469. Source data are provided as a Source data file.

Fig. 2. A high-throughput screen with the clinical PLS as a readout identifies nizatidine as candidate compound for treatment of chronic liver disease and HCC prevention.

a Effect of computationally selected compounds on the PLS. Simplified heatmaps show the HCV viral load after drug treatment (log 10), PLS global status, and PLS poor- and good-prognosis gene expression. PLS induction was determined by GSEA analysis using “mock” non-infected cells as reference control. Reversal of the PLS was assessed by comparing “HCV-infected cells treated with compounds” versus “HCV-infected cells (CTRL)”. The compounds were then ranked depending on the FDR (significant FDR < 0.25). Results from drug screening performed in triplicate are shown. b Screen hit target genes modulated by HCV infection at the single-cell level. Gene signatures comprising the targets of each specific compound were extracted from the LINCS database and tested for their enrichment in association with the HCV viral load. Heatmaps show the mean expression of the leading-edge genes for the significantly enriched signatures (p < 0.05) in each of the single cells (columns), row normalized as z-score. The single cells are ordered by infection status (white: non-infected cells, n = 17; black: infected cells, n = 23), and by increasing HCV viral load (blue bar plot on top). c Global modulation of the compound-specific gene signatures by HCV infection and drug treatment. Source data are provided as a Source data file.

Fig. 3. Nizatidine reverses the cPLS by inhibiting the HRH2/CREB signaling pathways.

a Detection of HRH2 in Huh7.5.1^dif cells by immunofluorescence. HRH2 is shown in magenta (Alexa Fluor^TM 647) and nuclei in blue (DAPI). CTRL = cells incubated with AF647-labeled secondary antibody. Scale bar: 10 µm. b PLS assessment upon perturbation of the HRH2/cAMP/PKA axis. PLS induction was determined by GSEA analysis using “Mock” non-treated cells as reference. Results from two experiments performed in triplicate are shown. Simplified heatmaps show PLS global status and PLS poor- and good-prognosis gene expression. FDR false discovery rate. c Intracellular levels of cAMP were assessed by ELISA. Results from two experiments are expressed in % ±SD compared to Mock. (n = 5; *p < 0.05, unpaired t test). d Expression of histidine decarboxylase (HDC) in Huh7.5.1^dif cells analyzed by qRT-PCR. Results from three independent experiments are expressed in % ±SD compared to Mock cells and normalized to GAPDH mRNA. (n = 7: **p < 0.01, two-tailed Mann–Whitney U test). e Nizatidine inhibits HCV-mediated CREB1 activation and decreases CREB5 expression. Western blot analysis of phospho-CREB1 (pCREB1) (Ser133), total CREB1, and total CREB5. NT non-treated. Results are representative of one out of two experiments (see Supplementary Fig. 7). f CREB5 is a driver of the HCV-induced cPLS. CREB5 KO was performed as described in “Method”. PLS induction was determined by GSEA analysis using “Mock” non-infected cells as reference. Reversal was assessed by analyzing sgCREB5 VS sgCTRL. g Activation of the HRH2 signaling pathways during chronic liver injury. AC: adenylate cyclase, CRE CREB response element. h Nizatidine reverts the FFA-induced poor-prognosis PLS in culture of patient-derived multicellular spheroids (three patients without history of chronic liver disease). PLS induction was determined by GSEA analysis using “Mock” non-treated spheroids as reference. Simplified heatmaps show PLS global status and PLS poor- and good-prognosis gene expression. i HRH2 expression measured by qRT-PCR in healthy multicellular spheroids. Each experiment shows mean ± SD in percentage compared to Mock spheroids (n = 4), *p < 0.05, unpaired t test. j Nizatidine reverts the histamine-induced poor-prognosis PLS in culture of patient-derived multicellular spheroids. Source data are provided as a Source data file.

Fig. 4. In vivo proof-of-concept of nizatidine for treatment of liver fibrosis progression and HCC chemoprevention.

a–d Nizatidine alleviates carcinogenesis in cirrhotic livers. a DEN-injured male Wistar rats received vehicle control or nizatidine for 10 weeks. PBS, n = 3; DEN + vehicle, n = 7; DEN + nizatidine, n = 8. b Representative morphometric analysis (hematoxylin and eosin (H&E)) of liver slices are shown (original magnification ×5). The number and size of macroscopic tumors were reported in c. Surrogates used for tumor burden include liver/body weight ratio and number of surface tumor nodules. Cell proliferation was assessed by the IHC evaluation of the proliferation marker Ki-67 in liver tissues. d Measurement of serum transaminases (alanine aminotransferase (ALT), aspartate aminotransferase, (AST) alkaline phosphatase (ALP), gamma-glutamyl transferase (γGT) activity, and of total bilirubin are shown. e, f Nizatidine efficiently reduces fibrosis. Liver specimens were stained with Sirius red and fibrosis stage was evaluated through quantitative digital analysis of whole-scanned liver sections (collagen proportional area, CPA). αSMA staining, hydroxyproline quantification and fibrotic gene expression are shown. g RNA-Seq analysis of the liver from three animals subjected to either vehicle or nizatidine treatment. Commonly dysregulated pathways in both severely fibrotic NASH liver (compared to mildly fibrotic NASH liver) and DEN-treated rat liver, and pathways reversed by nizatidine were determined by GSEA. h Heatmaps show the mean expression of the 186 gene signature (z-scores of log2-normalized data). Gene expression was normalized according to six different housekeeping genes. i–k Expression of nizatidine target genes. i Liver specimens were stained with anti-HRH2 or anti-pCREB1 antibodies. Original magnification ×10 or ×20 (hepatocytes). j HRH2 expression and CREB1 phosphorylation were assessed by quantification of IHC. Levels of cAMP were assessed by ELISA. Creb1, Creb5, and Prkacb expression were extracted from RNA-Seq analyses. k Western blot analysis of HRH2, phosphorylated (Ser 133) CREB1, total CREB1, and CREB5. Beta-actin was used as a loading control. Quantification of western blot intensities (arbitrary units) was performed using image J software. All graphs show mean ± SD. # denotes p < 0.05 and ## denotes p < 0.01 comparing PBS-treated rat to DEN + vehicle. * denotes p < 0.05 and ** denotes p < 0.01 comparing DEN + vehicle to DEN + nizatidine (One-way ANOVA, followed by Tukey’s multiple comparisons tests). Source data are provided as a Source data file.

Fig. 5. Genetic loss-of-function studies confirm a functional role of HRH2 in hepatocarcinogenesis.

a Validation of sgRNA targeting mouse Hrh2 and Tp53. KO efficacy was assessed by TIDE analysis. Plasmid constructs are described in method. b Hrh2 (sgHrh2) and Tp53 (sgTp53) KO and Kras constitutive expression (KRAS^G12D) or CTRLs were induced by injecting different plasmids constructs by HTVI in C57BL/6 males. Graph shows survival curves of injected mice. Median survival from injection to death is shown. Log-rank Matel–Cox test. **p < 0.01. c Pictures of representative livers. Scale bars, 1 cm. d Liver weight body ratio and total number of tumor nodules per mouse are shown (mean ± SD; CTRL mice, n = 5; Hrh2 KO mice, n = 6). p > 0.05 (two-tailed Mann–Whitney U test). e Cell proliferation was assessed by the IHC evaluation of Ki-67 in liver tissues (mean ± SD; CTRL mice, n = 5; Hrh2 KO mice, n = 6). Scale bar, 300 µM. **p < 0.01 (two-tailed Mann–Whitney). f, g HRH2 KO decreases cancer cell proliferation in cell culture. f HRH2 KO validation at genetic level using T7 endonuclease assay. g Effect of HRH2 KO on cancer cell proliferation. EdU incorporation assay by FACS showing % ±SD of proliferative EdU-positive cell from three independent experiments in CTRL and HRH2 KO cells (n = 8; **p < 0.01, two-tailed Mann–Whitney U test). h Effect of HRH2 KO on cancer cell apoptosis induced by oxidative stress (H₂0₂). Cleaved caspase 3 is shown in green. Nuclei were counterstained in blue (DAPI). Scale bar, 100 µM. Graph shows integrated cleaved caspase 3 intensity/total cell number from two independent experiments (n = 12; **p < 0.01, two-tailed Mann–Whitney U test) measured using Celigo Cytometer. Western blot analysis of cleaved- and total caspase 3 is shown. i–k Effect of HRH2 knockdown on cancer cell proliferation. i siRNA efficacy was assessed by measuring mRNA by qRT-PCR. j EdU incorporation assay by FACS showing % ±SD of proliferative EdU-positive cell from three independent experiments in cell transfected with siCTRL and siHRH2 (n = 6; **p < 0.01, two-tailed Mann–Whitney U test). k Cell proliferation was assessed daily in Huh7.5.1 transfected with a siCTRL or a siHRH2 by cell counting (TC20 Automated Cell Counter). Two representative and independent experiments are shown. l Full cPLS induction is impaired by HRH2 KO. PLS induction was determined by GSEA analysis using “Mock” non-infected cells as reference. Source data are provided as a Source data file.

Fig. 6. ScRNA-Seq analyses of patient liver tissue uncover pro-inflammatory liver macrophages as nizatidine target.

at-SNE map of single-cell transcriptomes from normal liver tissue of donors without history of chronic liver disease highlighting the main liver cell compartments. Data extracted from ref. . Cells sharing similar transcriptome profiles are grouped by clusters and each dot represents one cell. Expression t-SNE map of HRH2, CLEC5A, CD163L1, and MARCO are shown. The color bar indicates Log2-normalized expression. b–e Perturbation of gene expression by nizatidine in liver tissue from patient with chronic liver disease and HCC identifies liver macrophages as therapeutic target. b Experimental approach. CD45⁺ leukocytes from patient liver tissue were enriched by flow cytometry and were treated with nizatidine or vehicle control (DMSO). Single cells were sorted and analyzed as described. c t-SNE map of single-cell transcriptomes showing control (blue) and nizatidine-treated cells (yellow), d the t-SNE map indicating the main cell compartments (MAFB: macrophages, CD8: CD8⁺ T lymphocytes). e Expression t-SNE map of HRH2, macrophage markers, and Siglec-10 are shown. The color bar indicates log2-normalized expression. f–h GSEA for differentially expressed genes between nizatidine-treated and CTRL macrophages depicted in d. f Normalized enrichment score (NES) of genes related to macrophage activation (classical M1 vs alternative M2). g NES of the pathways significantly enriched after nizatidine treatment (FDR ≤ 0.05). h Expression heatmap of differentially expressed genes in individual nizatidine- and control-treated macrophages. Each row representing a single cell. Markers of inflammation, fibrogenesis/cancer, and antigen presentation are shown. All genes are normalized by row from their own min to max (Log2 fold; p value ≤ 0.05). Source data are provided as a Source data file.

Fig. 7. Expression of the liver HRH2/CREB5 pathway is associated with chronic liver disease progression and hepatocarcinogenesis in patients.

a, b HRH2 (a) and CREB5 (b) expression in liver tissues of clinical cohorts with various liver disease etiologies. In box and whisker plots, boxes represent the 75th and 25th percentiles, the whiskers represent the most extreme data points within interquartile range × 1.5, and the horizontal bar represents the median. Open circle indicates actual observation for each sample. Exact p values are indicated for each panel (Wilcoxon’s rank-sum test for all panel excepted for GSE94660, paired Wilcoxon’s rank-sum test). GSE84346: normal uninfected, n = 6, HCV infected, n = 40; GSE48452: healthy obesity, n = 27; NASH, n = 18; GSE94660, paired samples, n = 21; GSE48452: healthy obesity, n = 27; NASH, n = 18; GSE49541: mild fibrosis, n = 40, advanced fibrosis, n = 32; GSE14520: non-cirrhosis, n = 14, cirrhosis, n = 185; GSE10143: non-cirrhosis, n = 15, cirrhosis, n = 32. GSE94399: good prognosis, n = 23; poor prognosis, n = 15. c Higher expression of HRH2 in patient liver biopsies is associated with poor-prognosis and decreased probability of overall survival. Exact p values are indicated for each panel (log-rank test for comparisons of Kaplan–Meier survival). d Higher expressions of CREB5 in the adjacent livers were significantly associated with decreased survival and tumor recurrence after curative resection among patients with HCC. Exact p values are indicated for each panel (log-rank test for comparisons of Kaplan–Meier survival). Source data are provided as a Source data file.

Fig. 8. Proof-of-concept for therapeutic impact of nizatidine in patient-derived tissues and cell culture models.

a Nizatidine reverts the poor-prognosis PLS in culture of patient-derived tissue that were surgically resected from five patients diagnosed with chronic hepatitis C (HCV). Detailed PLS gene expression profiles: heatmaps show the mean expression of the 186 gene signature. PLS was determined by GSEA analysis using DMSO treated tissues as reference. Simplified heatmaps show: (top) the classification of PLS status as poor (orange) or good (green) prognosis; (bottom) the significance of induction (red) or suppression (blue) of poor- or good-prognosis genes. FDR false discovery rate. b Absent effect on cell viability in PHH, assessed 4 days after nizatidine treatment in 3D culture. Each experiment shows mean ± SD in percentage compared to DMSO treated cells (n = 4). c Nizatidine decreases HCC cell viability in a 3D patient-derived tumorspheroid model. HCC spheroids were generated from patient HCC tissues with different etiologies. Cell viability was assessed 4 days after treatment by measuring ATP levels. Each experiment shows mean ± SD in percentage compared to DMSO treated spheroids (n = 4). *p < 0.05; **p < 0.01, unpaired t test. The pictures show representative image of patient-derived tumorspheroids (magnification ×40). NASH nonalcoholic steatohepatitis, ALD alcoholic liver disease. Source data are provided as a Source data file.