Introduction

Biallelic complete loss-of-function variants in ZFYVE19 (MIM *619635) are associated with a type of progressive familial intrahepatic cholestasis (MIM #619849) characterised by conjugated hyperbilirubinemia, hypercholanemia, high serum activity of γ-glutamyl transpeptidase (GGT) and portal tract abnormalities that include increased numbers of bile ducts, cholangiocyte disarray with abnormal bile duct contours, portal venule hypoplasia and fibrous expansion, as encountered in congenital hepatic fibrosis.1 These abnormalities constitute the ductal plate malformation (DPM). Over time, portal inflammation, portal fibrosis and bile duct loss develop.1 2 The presentation may be associated with neonatal cholestasis3 or abnormal biomarker values that reflect hepatobiliary injury in older children or adults.4

ZFYVE19 (also known as ANCHR) and CHMP4C recruit VPS4 into the mid-body ring to regulate cytokinesis.5 Patient-derived cells that harbour biallelic complete loss-of-function variants in ZFYVE19 show ciliary and centriolar abnormalities.1 3 How the absence of a protein regulating cell division contributes to portal tract fibrosis is not understood. The developmental consequences of ZFYVE19 deficiency and their mechanisms need further study.

In the work described here, we generated a Zfyve19 −/− mouse model to investigate in mechanistic detail the role of loss of Zfyve19. Though the fetal development of intrahepatic bile ducts (IHBD) is normal, the knockout mice show histopathologic features characteristic of ZFYVE19-deficient patients on α-naphthyl isothiocyanate (ANIT) challenge after birth. The absence of ZFYVE19/Zfyve19 expression causes failure of cell division, with ciliary and centriolar abnormalities, and to cell death, thus initiating the process of biliary fibrosis. Our findings demonstrate that ZFYVE19 deficiency may be a ciliopathy with hepatobiliary manifestations that begin after birth.

Materials and methodsAnimal care

Mice used in this study were housed at an ambient temperature of approximately 22°C and subjected to 12-hour light-dark cycles. No more than five mice were housed in one cage, and free access to food and water was given. Mice were sacrificed by terminal anaesthesia with isoflurane, followed by cervical dislocation. All experiments were performed using age-matched littermate mice.

Generation of Zfyve19 knockout (Zfyve19−/−) mice

Since to date all patients with ZFYVE19 deficiency investigated harbour biallelic null mutations in ZFYVE19, we decided to produce a knockout mouse with complete loss of Zfyve19 function (Zfyve19−/− mouse), using CRISPR/Cas9 technology on a C57BL/6N background (figure 1A). Small guide RNAs (sgRNAs) were 5′-CCTTGTGGCCTTGTGCGCCCTGG-3′ and 5′-GGAGCGGGCAACTGCACGGTAGG-3′. Deletion of exon 3–6 of the Zfyve19 (GenBank ID: NM 028054.3; Ensembl: ENSMUSG00000068580) was predicted, leading to mRNA nonsense-mediated decay. Successful depletion of Zfyve19 was confirmed at the DNA (online supplemental figure 1A), mRNA (online supplemental figure 1B,C) and protein levels (online supplemental figure 2A,B).

Figure 1

Zfyve19 −/− mice show no defects in embryonic bile duct development and early postnatal phenotype. (A) Zfyve19 locus targeting scheme. (B) Immunofluorescence costaining of SOX9 and HNF4α in both Zfyve19 −/− and wild-type (WT) mouse livers at embryonic day 16.5 (E16.5). PV, portal vein. *Bile-duct lumen; dashed line, nascent tubules. Bars: 20 μm. (C) Body weight (g) measurements over 40 weeks in both female and male Zfyve19 −/− mice compared with WT littermates. Data are shown as mean±SD. (D) Kaplan-Meier survival curves over 80 weeks in Zfyve19 −/− mice compared with WT littermates of both sexes. (E,F) Mouse liver, age 8 weeks, stained with H&E and Sirius Red. Bars: 100 µm. ns, not significant.

ANIT treatment in mice

20 male mice (6–8 weeks of age, 18–22 g) were used, 10 Zfyve19−/− (groups 1 and 2, 5 mice each) and 10 wild type (WT) (groups 3 and 4, 5 mice each). On days 0, 7 and 14, mice in groups 1 and 3 were given the liver toxin ANIT (Sigma-Aldrich), 60 mg/kg, in olive oil vehicle via intragastric sonde; mice in groups 2 and 4 were similarly given vehicle at the same dose. Between 36 and 48 hours after the final dose, mice were fasted for 4 hours and euthanised; samples were collected (figure 2A).

Figure 2

Biliary fibrosis and disordered cholangiocyte polarity in liver of Zfyve19 −/− mice after α-naphthyl isothiocyanate (ANIT) challenge. (A) Schematic experimental outline. Mice per treatment group, n=5–6. (B) Appearance of representative livers and gallbladders. The inset shows a region with severe liver lesions. The enlarged gallbladder is marked by a red arrow. (C) Serum alanine aminotransferase (ALT), alkaline phosphatase (ALP), total bile acid (TBA), and total bilirubin (TBIL) levels. (D) Representative H&E and Sirius Red-stained liver sections. Bars: 100 µm. (E) Plot of per cent area of Sirius Red staining in livers from Zfyve19 −/− and wild-type (WT) mice after ANIT treatment. (F) RT-qPCR of α-Sma and Col1a1 in livers from Zfyve19 −/− and WT mice (n=5–6). (G) Immunofluorescence costaining of laminin and acetylated α-tubulin (ac-α-Tub) in mouse livers. Bars: 50 µm. *Bile duct lumen. (H) Immunofluorescence costaining of laminin and ac-α-Tub in liver from patients with ZFYVE19 variants (ZFYVE19mutant) or ABCB4 variants (ABCB4mutant) and from healthy controls. Bars: 50 µm. *bile duct lumen. Data are shown as mean±SD. ns, not significant, *p<0.05, **p<0.01, ***p<0.001 or ****p<0.0001.

Human samples

Tissues from three ZFYVE19-deficient patients (P14, P29 and P30) and three patients harbouring biallelic ABCB4 variants—ABCB4 deficiency, such as ZFYVE19 deficiency, presents as high-GGT cholestasis with cholangiopathy—all obtained from explanted livers, were studied, with normal donors as controls. All subjects or parents provided written informed consent.

Generation of ZFYVE19 knockout and knockdown retinal pigment epithelial-1 (RPE-1) cells

Human RPE-1 cells (‘hTERT RPE-1’, CRL-4000; ATCC, Manassas, Virginia, USA) were cultured in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12) (Biological Industries) with 5% Fetal Bovine Serum (FBS) (Gibco) under 5% CO2. sgRNAs (5′-CACCGCGACCCTGGACGCAACCCCG-3′ and 5′- AAACCGGGGTTGCGTCCAGGGTCGC-3′) targeting ZFYVE19 were inserted into a modified lentiCRISPRv2 vector (AddGene #98293). Lentivirus were produced by cotransfection of lentiCRISPRv2-sgRNA plasmids pCMV-VSVG (AddGene #8454) and psPAX2 (AddGene #12260). Media containing lentivirus were added to the RPE-1 cells’ culture medium when cells were grown to 50% confluence. Blasticidin (Meilunbio) was added to the culture medium 48 hours after infection, to a final concentration of 30 µg/mL. After 2 weeks of blasticidin selection, colonies formed by single cells were picked out and assessed for ZFYVE19 knockout by western blot.

To knock down the endogenous expression of ZFYVE19, siRNAs specific for ZFYVE19 were designed (siRNA1: 5′-GGAUGAGGCAAGUGGCUUUTT-3′ and 5′-AAAGCCACUUGCCUCAUCCTT-3′; siRNA2: 5′-UCACCACCUCAGAACUAUATT-3′ and 5′-UAUAGUUCUGAGGUGGUGACC-3′). The non-targeting siRNAs (5′-GCGACGAUCUGCCUAAGAUdTdT-3′ and 5′-AUC UUAGGCAGAUCGUCGCdTdT-3′) were used as negative controls. siRNA oligonucleotides were complexed with Lipofectamine RNAiMAX transfection reagent (Invitrogen) in Opti-MEM medium (Invitrogen) and were added to the medium when the cells were grown to ~50% confluence in penicillin-free and streptomycin-free DMEM/F12 complete medium (Gibco). The final siRNA concentration was 20 nM. After 72 hours, cells were fixed for immunofluorescence (IF) microscopy or harvested to assess ZFYVE19 knockdown efficiency by western blot (figure 3A).

Figure 3

Effect of ZFYVE19 defects on proliferation, maturity and differentiation of cholangiocytes in ZFYVE19-variant patients. Immunofluorescence photomicrographs of liver, PCNA and CK19 ((A) scale bar: 100 µm); SOX9 and CK7 ((C) scale bar: 50 µm); and CK19 and CK7 ((E) scale bar: 50 µm), ZFYVE19-deficient patient and normal donor (control). DAPI, 4′,6-diamidino-2-phenylindole. Histograms, CK19+PCNA+ cells as a percentage of total CK19+ cells (B); SOX9+CK7+ cells as a percentage of total CK7+ cells (D); and CK19+CK7+ cells as a percentage of total CK7+ cells (F). Data are shown as mean±SD. ns, not significant.

Mouse embryonic fibroblasts (MEFs)

MEFs were cultured in DMEM (Sigma) with 10% FBS (Gibco). WT MEFs and Zfyve19−/− MEFs were generated from WT and Zfyve19−/− mice, respectively. Mice were mated at 8–12 weeks of age and embryos were allowed to develop for 13.5–15.5 days (E13.5–E15.5). Embryonic thoracic wall tissues were used to generate an MEF cell suspension by incubation in 0.25% trypsin (Ethylenediaminetetraacetic acid (EDTA) free) at 37°C for 30 min. The MEF cell suspension was centrifuged at 1000 rpm for 5 min, washed in phosphate-buffered saline once and resuspended in complete culture medium.

DNA content assay

For DNA content assay by flow cytometry, cells were suspended in a 1:9 mixture of RNase A and propidium iodide (PI) reagent (Keygentec). The cells were immediately sorted by flow cytometry (Beckman). Flow cytometry data were analysed using ModFit LT V.5.0.

Measuring cell death by flow cytometry and transferase‐mediated deoxyuridine triphosphate nick‐end labelling (TUNEL) assay

Fluorescence-activated cell sorting with flow cytometry analysis of PI-stained cells was used to measure cell death as described.6

Fluorescent terminal deoxynucleotidyl TUNEL assayed dead cells in paraffin-embedded sections (4 µm) of liver tissues. Dead cells in portal areas were quantified by counting TUNEL+ cells per viewfield in at least 10 random microscope fields (15×) of portal tract areas, using a Panoramic MIDI histoscanner (3DHistech).

Serum biochemistry analysis, quantitative real-time PCR, RNA-Seq, western blotting analysis, histopathological staining and IF staining

See online supplemental material.

Statistical analysis

Statistical analysis used GraphPad Prism V.6.0 software (GraphPad). All data are presented as means±SD. Continuous data were tested for normality and analysed by unpaired Student’s t-tests, Mann-Whitney U testing or one-way analysis of variance (ANOVA), as appropriate. Differences within and between groups were evaluated using two-way ANOVA. Survival was assayed by Kaplan-Meier and log-rank analyses. Statistical significance is displayed as ns (not significant), *p<0.05, **p<0.01, ***p<0.001 or ****p<0.0001, unless specified otherwise.

ResultsZfyve19−/− mice and WT littermates show no obvious differences in bile duct development and postnatal growth

Zfyve19−/− mice were generated by CRISPR/Cas9 deletion of exons 3–6 of Zfyve19 (figure 1A). The development of IHBD is characterised by asymmetric expression of the expanded bile duct and hepatoblast cell markers, respectively, SOX9 on the internal side and HNF4α on the external side of the ductal plate (~E16 P2).6–8 IF staining of SOX9 and HNF4α in E16.5 Zfyve19−/− mice showed normal bile duct lumina. Ductal plate cells exhibited typical asymmetry in nascent tubules, with SOX9-expressing cells near the portal vein and HNF4α-expressing cells contralaterally (figure 1B). These developmental features did not differ from those of IHBD in WT littermates.

To assess growth, mice were weighed once a week during postnatal weeks 1–8 after birth and once every 4 weeks thereafter until age 40 weeks. Weights of Zfyve19−/− mice of either gender did not differ from those of WT littermates (figure 1C). Survival rates of Zfyve19−/− mice of either gender (up to 80 weeks) also did not differ from those of WT littermates (figure 1D). Hepatobiliary development as assessed by light microscopy did not differ at age 8 weeks between male Zfyve19−/− mice and WT littermates; cholestasis, bile duct proliferation, necrotic hepatocytes and hepatic fibrosis were not seen (figure 1E,F). Loss of Zfyve19 expression in mouse thus led to no identified systemic or liver-restricted abnormalities.

Features of liver injury, with biliary fibrosis and cholangiocyte disarray resembling lesions in ZFYVE19-deficient patients, are prominent after ANIT treatment in Zfyve19−/− mice

Zfyve19−/− mice were exposed to different liver toxins. With ANIT administration (figure 2A, see Material and methods for details), Zfyve19−/− mice trended towards greater loss of body weight and greater liver: body weight ratio at endpoint than those in WT littermates, although without differences that reached statistical significance (online supplemental figure 3A, B). However, livers of Zfyve19−/− mice were firmer than those of WT littermates, with more visible liver lesions (nodularity distributed throughout the liver), larger gallbladders and darker bile (figure 2B). The levels of serum ALT and ALP activity and total bile acid (TBA) concentrations, but not total bilirubin (TBIL) concentrations, were significantly higher in Zfyve19−/− mice than in WT littermates after ANIT challenge, while no differences were observed between Zfyve19−/− and WT mice with oil vehicle control (figure 2C).

Though histopathologic study showed no hepatobiliary anomalies in Zfyve19−/− mice and no differences from WT mice treated with vehicle controls, after ANIT treatment, more numerous necrotic hepatocytes, denser inflammatory infiltration, increased numbers of bile duct profiles on H&E staining (figure 2D), significantly more severe fibrosis on Sirius Red staining (figure 2D,E) and higher (approximately fourfold) mRNA expression of profibrosis genes α-Sma and Col1a1 (figure 2F) were seen in Zfyve19−/− mice than in WT mice. The expression of α-SMA and COL1A1 mRNA was consistently higher in livers of ZFYVE19-deficient patients than in normal controls (online supplemental figure 4), consonant with findings in mice.

Patients with biallelic complete loss-of-function ZFYVE19 variants exhibit DPM.1 DPM is generally associated with cholangiocyte-polarity disorders.7 8 IF costaining for laminin, a basally expressed antigen, and acetylated (ac)-α-tubulin (ac-α-Tub), an apically expressed antigen, accordingly was used to assess cholangiocyte apical–basal polarity. In vehicle-treated Zfyve19−/− and WT mice and in ANIT-treated WT mice, laminin adjoined the basement membrane of bile ducts and ac-α-Tub adjoined bile duct lumina (figure 2G). However, ANIT-treated Zfyve19−/− mice showed abnormal apical–basal cholangiocyte polarity: laminin localisation was disordered and ac-α-Tub signalling was obviously reduced or even absent at bile duct lumina (figure 2G).

To learn if the polarity disruption seen in ANIT-treated Zfyve19−/− mice was present in patients with ZFYVE19 deficiency, IF staining for laminin and ac-α-Tub was conducted in the livers of three such patients and of three patients with ABCB4 disease, a disorder clinically characterised by cholestasis with high serum GGT and abnormally numerous bile duct profiles, and in the livers of healthy controls. We found that the healthy controls without biliary fibrosis and the patients with ABCB4 variants, whose livers showed biliary fibrosis, had normal laminin and ac-α-Tub distributions in bile ducts (figure 2H). In contrast, laminin signalling in patients with ZFYVE19 deficiency lay principally at non-basal sites within cholangiocytes, while ac-α-Tub at bile duct lumina was markedly less strongly expressed than normal or even absent entirely (figure 2H). These findings demonstrate that the polarity disruption seen in ANIT-treated Zfyve19−/− mice is also present in patients with ZFYVE19 deficiency, suggesting that the Zfyve19−/− mouse may, as a model, track human disease.

IF staining for the mature cholangiocyte marker CK19 showed increased number of bile duct profiles in ZFYVE19-deficient patients (figure 3). IF staining for SOX9 (a marker of regeneration of hepatocytes and cholangiocytes) and CK7 (a mature and immature cholangiocyte marker) showed increased to similar degrees of CK19 expression (figure 3C–F). However, IF costaining did not reveal increased marking for PCNA, a proliferation-associated antigen (figure 3A,B). The expression of Pcna mRNA in ANIT-treated Zfyve19−/− mice did not differ from that in WT controls (online supplemental figure 5).

ZFYVE19 depletion/deletion causes failure of cell division and cell death

ZFYVE19 acts as a key regulator of cytokinesis.5 To examine the effects of ZFYVE19 deficiency on cell division, ZFYVE19-targeting siRNA was used to deplete ZFYVE19 in RPE-1 cells (online supplemental figure 6A). ZFYVE19 knockdown induced a significant increase in proportions of cells with DNA content>2N and <4N (online supplemental figure 6B,C). IF staining showed that the proportion of cells with three or four centrioles (CP110+) was significantly increased, by at least fourfold, in ZFYVE19 knockdown cells (online supplemental figure 6D,E). Similar results were observed in ZFYVE19 knockout RPE-1 cells (online supplemental figure 6A–E) and Zfyve19−/− MEFs (online supplemental figure 7A–C).

The failure of cell division can cause overduplication of centrioles. The duplicated centrioles express CEP170, a marker of mature centrioles. Costaining for CP110 and CEP170 was performed (online supplemental figure 8A). The proportion of cells with >2 centrioles of which >1 marked for CEP170 was significantly greater among ZFYVE19 knockout RPE-1 cells than among controls (online supplemental figure 8B). This indicates that centrioles are overduplicated in ZFYVE19 knockout cells, in agreement with greater numbers of double cilia in ZFYVE19 knockout RPE-1 cells than in controls (online supplemental figure 8C,D) and in Zfyve19−/− MEFs than in WT MEFs (online supplemental figure 8E,F).

Both defects of centrosome splitting and failure of cell division cause cell death.9–11 Since biliary fibrosis in livers with ZFYVE19 deficiency might reflect an inflammatory response induced by cell death, PI uptake in ZFYVE19 knockdown RPE-1 cells was measured by flow cytometry. Depletion of ZFYVE19 in RPE-1 cells led to more cell death (figure 4F,G). To determine the levels of cell death in vivo, TUNEL assays were performed in mouse and human liver sections. A number of dead cells in the portal areas of the liver in ANIT-treated Zfyve19−/− mice were significantly higher (greater than threefold) than in ANIT-treated WT mice. No difference between vehicle-treated Zfyve19−/− and WT mice was observed (figure 4H,I). Consistent with this, TUNEL assays found substantially more dead cells (>20-fold) in the portal areas of liver from ZFYVE19-deficient patients than in those from healthy controls (figure 4J,K).

Figure 4

Failure of cell division and cell death after ZFYVE19 depletion/suppression. (A) Western blot of ZFYVE19 in retinal pigment epithelial-1 (RPE-1) cells. (B) Representative cell cycle distributions in ZFYVE19 knockdown RPE-1 cells, measured by DNA content; flow cytometry analysis. (C) Histogram, cell cycle distribution in ZFYVE19 knockdown RPE-1 cells in (B). (D) Immunofluorescence (IF) costaining for CP110 and γ-tubulin (γ-Tub) in ZFYVE19 knockdown RPE-1 cells. Bars: 10 µm. (E) Percentages of ZFYVE19 knockdown RPE-1 cells with different numbers of centrioles (CP110+). (F) Cell death in ZFYVE19 knockdown RPE-1 cells measured by flow cytometry. (G) Percentages of propidium iodide-marked cells in ZFYVE19 knockdown RPE-1 cells. (H) IF, TUNEL+ cells (arrows) in portal tracts in the livers of WT and Zfyve19 −/− mice after oil or 1-naphthyl isothiocyanate treatment. (I) Number of TUNEL+ cells per viewfield (15×) from mouse tissues (n=3–5). (J) IF, periportal TUNEL+ cells (arrows) in liver from ZFYVE19-variant patients (n=3). (K) Number of TUNEL+ cells per viewfield (15×) in liver sections from ZFYVE19-variant patients. Bars: 50 µm. Data are shown as mean±SD. ns, not significant, *p<0.05, **p<0.01, ***p<0.001 or ****p<0.0001.

Moreover, gene set variation analysis (GSVA) on mRNA-seq in livers of ZFYVE19-deficient patients and livers from ANIT-treated Zfyve19−/− mice showed alterations in cell division-related and cell death-related signalling pathways (G2/M-checkpoint, DNA-repair and mitotic-spindle signalling pathways) (online supplemental figure 9 and figure 5A).

Figure 5

Changes in signal pathways regulating hepatic fibrosis and macrophage activity in ZFYVE19-deficient and Zfyve19-deficient liver. (A) Gene set variation analysis of signal pathways in mouse liver from 4 groups (n=5–6). (B,C) Expression of genes in the transforming growth factor-β (TGF-β) pathway tested using RT-qPCR in mouse liver (n=5–6) and human liver. (D) Relative mRNA expression levels of macrophage chemokine genes, including Cxcl1, Cxcl10, Cxcl12 and Ccl2 and proinflammatory cytokines, including Il-1β, Tnf-α and Il-6 by RT-qPCR in mouse liver (n=5–6). KO, knockout (Zfyve19 −/−). WT, wild-type. Data are shown as mean±SD. ns, not significant, *p<0.05, **p<0.01, ***p<0.001 or ****p<0.0001.

Transforming growth factor-β (TGF-β) signalling pathway and macrophage chemokine synthesis are up-regulated in ZFYVE19-variant livers

To explore the pathways and genes involved in fibrosis, we applied GSVA to mRNA-seq data from the livers of ZFYVE19-deficient patients and of Zfyve19−/− mice. GSVA showed significant upregulation of the TGF-β signalling pathway in ZFYVE19-deficient patients (online supplemental figure 9). In Zfyve19−/− mice gavaged with ANIT the TGF-β and JAK-STAT3 signalling pathways were up-regulated in comparison with control mice (figure 5A); these pathways are important in regulating liver fibrosis and macrophage activation.12 13 RT-qPCR of mouse liver tissue confirmed these results. Tgfβ1 mRNA, but not Tgfβ2 mRNA, was more abundant in ANIT-induced Zfyve19−/− mice than in ANIT-exposed WT mice (figure 5B). The expression of Itgb6/8 (activators of latent TGFβ1) and of Ctgf, downstream from Tgfβ in a desmoplastic cascade, was also increased at the mRNA level in ANIT-exposed Zfyve19−/− mice (figure 5B). TGFβ1, CTGF and ITGB6/8 mRNA levels in the livers of ZFYVE19-deficient patients were consistently increased as well (figure 5C). The levels of SMAD2 and its activated form, p-SMAD2, which are TGFβ signalling pathway downstream regulators, were increased in patient liver (online supplemental figure 10A).

In contrast, expression in mouse liver of other genes related to hepatic fibrosis, for example, Yap (Hippo signalling pathway), β-Catenin (Wnt signalling pathway), Jnk, Rock2 (planar cell polarity pathway) and PDGFα, did not vary from controls (online supplemental figure 11). We therefore conclude that the TGF-β and JAK-STAT3 signalling pathways may be principally involved in biliary fibrosis in ZFYVE19 deficiency.

Consistent with mRNA sequencing data, RT-qPCR results showed that after ANIT treatment expression of macrophage chemotactic factors (Cxcl1, Cxcl10, Cxcl12 and Ccl2) and of the proinflammatory cytokine Il-1β increased in liver of Zfyve19−/− mice (figure 5D), while no significant difference in Il-1β expression was observed in ZFYVE19-deficient patients (online supplemental figure 10B). Although the expression of inflammatory factor Tnf-α in Zfyve19−/− mice also increased, no statistical significance was attained. The expression of another inflammatory factor, Il-6, was similar among all four groups (figure 5D).

Discussion

A recently identified association between biallelic loss-of-function ZFYVE19 variants and a type of high-GGT portal fibrosis histopathologically characterised by DPM has been confirmed.3 4 However, no animal model has existed. The work presented here—generation of a Zfyve19−/− mouse in which, on ANIT administration, the human ZFYVE19-deficiency phenotype is recapitulated—provides mechanistic insight into the causal relationship between ZFYVE19 deficiency and the observed disease.

Zfyve19 −/− mice without challenge did not manifest the liver phenotype of ZFYVE19-deficient patients. Traditionally, when phenotypes in knockout mice do not recapitulate phenotypes in patients with lesions in orthologous genes, toxins are often used to induce phenotype development. For example, in Abcb11 knockout mice, cholic acid feeding induces ABCB11-deficiency-like phenotypes.14 Zfyve19 −/− mice on cholic acid feeding did not have the liver phenotype of ZFYVE19-deficient patients (data not shown). ANIT is a biliary toxin used to study acute (high and single dose) and chronic (low dose, ≥4 weeks) bile duct injury in mice.15 With three times weekly low-dose intragastric administration of ANIT, Zfyve19−/− mice developed elevated serum ALT and ALP activities and TBA concentrations (figure 2C) with IHBD abnormalities (figure 2D), portal tract inflammation (figure 2D) and portal tract fibrosis (figure 2E,F), findings resembling those in ZFYVE19-disease patients1.

Anomalies of cholangiocyte polarity, often associated with DPM, were then identified in ANIT-treated Zfyve19−/− mice. These anomalies were further demonstrated in ZFYVE19-deficient patients, but not in explanted livers from patients with ABCB4 variants. These results indicate that disordered cholangiocyte polarity was associated with ZFYVE19/Zfyve19 defects rather than simply with inflammatory cholangiopathy.

ZFYVE19 (ANCHR) was originally identified as a key component of cytokinesis during cell division, associated with Aurora-B-dependent abscission checkpoint control.5 Failure of cell division presumably equips cells with supernumerary ‘mother centrioles’ that become basal bodies for supernumerary cilia.16 Consistent with identification of abnormal centrioles in ZFYVE19-deficient patient-derived cells,1 ZFYVE19 depletion induced an increase in the number of cells with DNA content>2N, cells with >2 centrioles and cells with >1 mature centriole, suggesting failure of cell division.17 Both defects of the centrosome and failure of cell division have been related to cell death.10 18 19 This study demonstrates increased cell death in ZFYVE19-deficient RPE-1 cells, in portal areas of ANIT-treated Zfyve19−/− mice and in ZFYVE19-deficient patients. The effects of ZFYVE19 deficiency on cell division and cell death are also supported by RNA-seq data in ZFYVE19-deficient patients and ANIT-treated Zfyve19−/− mice that demonstrate upregulation of cell division-related signalling pathways, such as those for the G2/M checkpoint, DNA repair, and mitotic spindle formation.

Increased cell death was demonstrated by ZFYVE19 depletion/deletion in vitro and in vivo (see Results). Various cytokines and chemokines released from dead cells may participate in activation of intrahepatic macrophages, promoting fibrogenesis.20 21 RNA-seq data in ZFYVE19-deficient patients and ANIT-treated Zfyve19−/− mice suggest upregulation of TGF-β and JAK-STAT3 pathways, which are related to fibrosis and to macrophage activation.22–24 Elevation of TGF-β1, CTGF and ITGB6/8 mRNA synthesis, and of SMAD2 and p-SMAD2 expression, in ZFYVE19-deficient patients and ANIT-treated Zfyve19−/−mice confirms activation of the TGF-β pathway. Moreover, we found that mRNA expression of macrophage chemokines Cxcl1, Cxcl10 and Cxcl12, which might participate in the activation of macrophages, increased in Zfyve19−/− mouse liver, similar to changes seen in the Pkhd1 del4/del4 mouse (a mouse model of congenital hepatic fibrosis).21