Mannose Phosphate Isomerase and Mannose Regulate Hepatic Stellate Cell Activation and Fibrosis in Zebrafish and Humans

Abstract

The growing burden of liver fibrosis and lack of effective antifibrotic therapies highlight the need for identification of pathways and complementary model systems of hepatic fibrosis. A rare, monogenic disorder in which children with mutations in mannose phosphate isomerase (MPI) develop liver fibrosis led us to explore the function of MPI and mannose metabolism in liver development and adult liver diseases. Herein, analyses of transcriptomic data from three human liver cohorts demonstrate that MPI gene expression is down-regulated proportionate to fibrosis in chronic liver diseases, including nonalcoholic fatty liver disease and hepatitis B virus. Depletion of MPI in zebrafish liver in vivo and in human hepatic stellate cell (HSC) lines in culture activates fibrotic responses, indicating that loss of MPI promotes HSC activation. We further demonstrate that mannose supplementation can attenuate HSC activation, leading to reduced fibrogenic activation in zebrafish, culture-activated HSCs, and in ethanol-activated HSCs.

Conclusion:

These data indicate the prospect that modulation of mannose metabolism pathways could reduce HSC activation and improve hepatic fibrosis.

The dearth of effective antifibrotic therapies highlights the need for identification of pathways and complementary model systems of hepatic fibrosis. Activation of hepatic stellate cells (HSCs) is central to liver fibrosis and is the main focus of antifibrotic drug discovery efforts. The functional contribution of glycosylation pathways to HSC activation and fibrosis has been largely overlooked.

Children with a congenital disorder of glycosylation (CDG) who lack the enzyme, mannose phosphate isomerase (MPI), have altered protein glycosylation and develop early, progressive liver fibrosis requiring transplantation.^(1–3) This monogenic disease poignantly demonstrates that loss of MPI provokes liver fibrosis; however, the mechanisms underlying this effect are not known. MPI interconverts fructose 6-phosphate and mannose 6-phosphate (Fru6P ←→ Man6P; Supporting Fig. S1), key intermediates in glycolysis and protein N-linked glycosylation, respectively. Importantly, mannose supplements can bypass MPI deficiency and directly provide Man6P for glycan synthesis (Supporting Fig. S1), leading to improvement in almost all symptoms within a few weeks.^(4,5) We previously established a zebrafish model of MPI-CDG with clinical cornerstones similar to MPI-CDG, including phenotypic rescue with mannose supplementation.⁽⁶⁾ Here, we utilize this model to investigate how loss of MPI leads to liver fibrosis in vivo and, more broadly, explore the potential contribution of reduced MPI in common adult fibrotic liver diseases. This disorder provides a unique opportunity to study the molecular basis of an early form of hepatic fibrosis, because the underlying genetic defect is clearly established.

HSC activation is a central event in response to chronic hepatocyte injury, leading to secretion of collagen and other extracellular matrix (ECM) proteins that promote development and progression of liver fibrosis to cirrhosis.⁽⁷⁾ Drug discovery efforts to target hepatic fibrosis have focused on known pathways and regulators of HSC activation, yet few candidates have entered early-phase clinical trials, and none have been approved for treatment.⁽⁷⁾ Based on initial studies in zebrafish, we hypothesized that loss of MPI in HSCs can amplify their activation, even in the presence of little or no liver injury.

Here, we investigate the broader role of MPI and mannose metabolism in liver development and adult liver diseases and integrate findings using human liver samples, in vivo zebrafish models, and HSCs in culture to investigate an important role of MPI and mannose in regulating HSC activation and hepatic fibrosis. By extrapolating therapy from a rare disease, our data suggest that mannose supplementation could potentially be a simple and effective antifibrotic therapy that merits further evaluation.

Materials and Methods

ZEBRAFISH MAINTENANCE, INJECTION, AND TREATMENT OF EMBRYOS

Procedures were performed in accord with the Icahn School of Medicine at Mount Sinai (ISMMS) Institutional Animal Care and Use Committee. Adult fish were maintained on a 14:10 light/dark cycle at 28°C. Wild-type (WT; AB and Tab 14), Tg(fabp10a: nls-mcherry)^mss4(8), Tg(hand2:EGFP)^pd24,⁽⁹⁾ Tg(TP1: βglobin-GFP)^um14,⁽¹⁰⁾ Tg(kdrl:ras-mCherry)^s896,⁽¹¹⁾ and mpi^{mss7 (12)} zebrafish strains were used. Fertilized embryos were collected and cultured in fish water (0.6 g/L of Crystal Sea Marinemix; Marine Enterprises, Baltimore, MD) containing methylene blue and maintained at 28°C. Embryos were injected with 4 ng of standard control or mpi morpholino⁽⁶⁾ at the one-to four-cell stage. Embryos were treated with D-mannose or 2% ethanol as described.^(6,13)

HISTOLOGY

Whole livers were dissected from 19-month-old, male adult zebrafish and fixed overnight at 4°C in Bouin’s fixative. Four-micron serial sections of paraffin-embedded fish were stained with Masson’s trichrome and imaged on an Olympus BX41 microscope (Olympus Corporation, Waltham, MA) and Nikon DS-Ri1 digital camera (Nikon Instruments, Inc., Melville, NY).

IMMUNOFLUORESCENCE

Zebrafish were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) overnight at 5 days postfertilization (dpf) and transferred to 30% sucrose in PBS overnight. Larvae were embedded in optimum cutting temperature (OCT) compound (Tissue-Tek), and 10-μm serial sections were obtained using the Leica CM3050 S Research Cryostat. Sections were washed in PBS + 0.1% Tween-20 (PBST). Tissue sections were blocked with 5% fetal bovine serum (FBS) and 2% bovine serum albumin (BSA) in PBST for 1.5 hours at room temperature (RT). A custom polyclonal antibody recognizing zebrafish Mpi was commercially prepared in rabbit (Biomatik, Inc, Wilmington, DE) using the sequence DPQAHIEHSKPYAE as the target antigen. Sections were stained with Mpi antibody (1:200) overnight at 4°C and then with 1:250 goat antirabbit Alexa Fluor 488 (A11008; Life Technologies, Waltham, MA) for 1.5 hours in the dark at RT. Sections were mounted with ProLong Gold Antifade Mountant with 4′,6-diamidino-2-phenylindole (DAPI; Life Technologies) and imaged using a Leica SP5 DMI at 63×. Biliary network reconstruction was performed as described.⁽¹⁴⁾ Mouse liver sections (Lrat-CRE/ZsGreen^Flox/Flox)⁽¹⁵⁾ were fixed in 4% PFA in PBS overnight and incubated for 24 hours each in 10%, 20%, and then 30% sucrose in PBS before embedding in OCT, and sectioned in 10 μM. Tissue sections were rehydrated in PBS and incubated with anti-MPI antibody (Sigma SAB2700835, 1:100; Sigma-Aldrich, St. Louis, MO), anti-CD31 antibody (550274, 1:200; BD Pharmingen, San Jose, CA), and anti–glial fibrillary acidic protein (GFAP; ab4674, 1:200; Abcam, Cambridge, MA) diluted in PBS overnight at 4°C. Sections were washed with PBS and incubated with appropriate secondary antibodies (A10520, A11007, A21245, and A21437; ThermoScientific, Waltham, MA) at 1:500 in PBS for 1 hour at RT. Sections were washed 3× in PBS and mounted in DAPI Fluoromount-G Mounting Medium (SouthernBiotech, Birmingham, AL). Imaging was performed on a Zeiss Axio Observer 7 (Zenblue software).

CELL CULTURE

TWNT-4 and LX-2 human HSC lines^(16,17) were maintained in complete medium (Dulbecco’s modified Eagle’s medium [DMEM] supplemented with 10% FBS and L-glutamine) and routinely tested for mycoplasma using the Venor GeM Mycoplasma Detection Kit (Sigma-Aldrich). For MPI knockdown experiments, small interfering RNAs (siRNAs) targeting human MPI (siMPI; see a previous work⁽¹²⁾) were transfected using Lipofectamine RNAiMAX transfection reagent (ThermoFisher, Waltham, MA), following the recommended protocol. Cells were collected 48 hours after transfection for total RNA or protein. For ethanol and mannose treatments, cells were grown in complete medium, replaced with starvation medium (DMEM + 0.1% BSA and 2 mM of L-glutamine) overnight (16–20 hours), replaced with starvation medium supplemented without or with ethanol and mannose, as described in the text, and collected 24 hours later for total RNA or protein. For proteomics analysis, man-nose-treated or untreated LX-2 cells were collected by scraping in 8 M of urea/20 mM of Hepes buffer (pH 7.1) with phosphatase inhibitors (Roche PhosSTOP, Sigma-Aldrich). Like samples (10 mL total) were combined, flash-frozen, and stored at –80°C until processing.

TRICHLORACETIC ACID PRECIPITATION OF MEDIA PROTEIN

Conditioned media were collected from cultures, and media proteins were precipitated with an equal volume of ice-cold 10% trichloracetic acid (TCA) for 1–2 hours on ice, centrifuged, and washed with ice-cold acetone for 30 minutes on ice and centrifuged again. Protein pellets were resuspended in nonidet P-40 (NP-40) cell lysis buffer (50 mM of Tris.Cl [pH 8.0], 150 mM of NaCl, and 1% NP-40) + protease inhibitor (Roche Complete, Sigma-Aldrich) to achieve 5× concentration and sonicated to resuspend. Twenty microliters of concentrate were separated by sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS-PAGE) for analysis.

MPI ENZYME ACTIVITY ASSAY

MPI activity assay was performed according to our published protocol in zebrafish and mammalian cells extract.^(6,12) Briefly, zebrafish or cell lysates were homogenized and MPI activity was assessed in 15 ug of protein extract using a coupled-fluorometric assay.

GENE EXPRESSION ANALYSIS IN ZEBRAFISH LIVERS AND HUMAN CELL LINES

Livers were microdissected from anesthetized zebrafish larvae and collected in 20 μL of RLT Buffer (Qiagen, Germantown, MD). Total RNA was isolated from >10 livers per sample by TRIzol extraction (Life Technologies). Human HSCs were lysed in TRIzol, following the supplied protocol. RNA was reverse transcribed using the SuperScript complementary DNA (cDNA) synthesis kit (Quantabio, Beverly, MA). qRT-PCR was performed using PerfeCTa SYBR Green Fast Mix (Quantabio). Samples were run in triplicate on the Roche LightCycler 480 in a reaction volume of 10 μL. Gene expression levels were normalized to ribosomal protein large P0 (zebrafish) or ribosomal protein S18 (human) using the comparative threshold cycle (ΔΔCt) method. Primer sequences are listed in Supporting Table S1.

Human Samples (In Silico)

To assess the association of MPI depletion and fibrosis in human chronic liver disease, we selected three different cohorts of human samples, which encompass different degrees of fibrosis and two etiologies for which genome-wide expression data are available: hepatitis B virus (HBV; ArrayExpress E-GE0D-84044); nonalcoholic fatty liver disease (NAFLD; see a previous work⁽¹⁸⁾, E-GEOD-49541); and Heptromic (see a previous work⁽¹⁹⁾, GEO Omnibus GSE63898). In the HBV cohort, patients were categorized in two groups according to their Scheuer staging (i.e., scarring and architectural distortion): low (value ≤ 2) and high (value > 2). Differences in MPI expression were tested with the Mann-Whitney U test. Subsequent functional characterization of samples with various MPI expression levels was studied using gene set enrichment analysis (GSEA; also see a previous work⁽²⁰⁾). We ranked patients based on MPI expression level (regardless of their Scheuer staging and grading) and compared the expression of all genes for patients <10th percentile in MPI expression with those >90th percentile. Differentially expressed genes were identified using linear models as implemented in the limma package,⁽²¹⁾ and P values were corrected for multiple testing using the Benjamini-Hochberg procedure. Volcano plots show data of one probe per gene. For those genes with multiple probes, the one with the highest fold-change expression is depicted in the plot. The resulting fold changes were used to rank the genes and test enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways using GSEA.⁽²⁰⁾ The NAFLD cohort was analyzed in a similar way, with the patients categorized by degree of fibrosis: advanced NAFLD (stages 3 and 4) versus mild NAFLD (stages 0 and 1). For this cohort, because the total number of patients was significantly lower than for the other two data sets, we compared expression between patients with MPI values <35th percentile and those with >65th percentile. In the Heptromic cohort, the comparison was made between normal liver samples and adjacent cirrhotic tissues from patients with hepatocellular carcinoma (HCC). The rest of the analyses are analogous to the HBV cohort.

WESTERN BLOTTING

Zebrafish samples were homogenized in lysis buffer (20 mM of Tris [pH 7.5], 150 mM of NaCl, 1% NP-40, 2 mM of ethylenediaminetetraacetic acid, 10% glycerol, 1% β-mercaptoethanol, and protease inhibitors [Roche Complete]). Lysates were centrifuged with one volume sample buffer added to each sample. Cell samples were collected by trypsinization and lysed by sonication in NP-40 cell lysis buffer + protease and phosphatase (Roche PhosSTOP) inhibitors. Lysates were resolved by 12% SDS-PAGE. Anti-PDGFRB (sc-432 at 1:1,000; Santa Cruz Biotechnologies, Dallas, TX), anti–Collagen 1a (BS-10432R at 1:1,000; Bioss, Woburn, MA), and anti-Tubulin (12G10 at 1:2,000; DSHB, Iowa City, IA) were used.

PROTEOMICS ANALYSIS

Urea lysates from mannose-treated or untreated LX-2 cells were processed for proteomic analysis by the Stony Brook University (SBU) Biological Mass Spectrometry Shared Resource Facility. Briefly, samples were reduced, alkylated, proteolytically cleaved, and desalted using C18 Sep-Pak Plus Cartridges (Waters WAT020515). On-column peptides were desalted and eluted with sequential 10%, 20%, and 50% acetonitrile, 0.1% trifluoroacetic acid, and lyophilized. Phospho-and total peptide fractions were isolated and labeled with isobaric tagging for relative and absolute quantitation (iTRAQ) stable isotope tags (Sigma 4381662), run on a Sciex 5600+ quad-rupole quantitative time-of-flight liquid chromatography with tandem mass spectrometry instrument, and analyzed using Analyst QS 1.1 to determine peptide masses, sequence information, and abundance. For differentially regulated pathway analysis, we first calculated the ratio of protein or phosphoprotein in Mannose-treated versus control experiments and then selected the experimental replicate with the highest number of observed protein or phosphoprotein for downstream analyses. For phosphoproteomics data, we collapsed the phosphosites to gene-level phosphorylation values by averaging the phosphosite expressions observed for each protein. We used the Wilcoxon rank-sum test to test for changes in protein expression or phosphorylation ratio ranks of members in each KEGG signaling pathway against all the other observed protein in that sample. Resulting P values were adjusted using the Benjamini-Hochberg procedure to false discovery rate (FDR).

STATISTICAL ANALYSIS

Data were analyzed using GraphPad Prism software (version 8; GraphPad Software Inc., San Diego, CA). Data are expressed as mean ± SD. Differences between experimental and control groups were analyzed by two-tailed, paired Student t test or one-way analysis of variance, followed by Dunnet’s or Tukey’s post-hoc correction, when more than two groups were compared. P values <0.05 were considered statistically significant.

RESULTS

MPI EXPRESSION IS INVERSELY CORRELATED WITH ADVANCED FIBROSIS IN HUMAN LIVER DISEASE

MPI loss leads to early fibrosis in children with MPI-CDG, suggesting a link between MPI deficiency and liver fibrosis. To determine whether changes in MPI correlated with common etiologies of liver disease, we analyzed publicly available transcriptome expression data sets encompassing fibrotic samples from three human liver disease cohorts totaling 376 samples and arising from different etiologies: HBV cohort (n = 124), NAFLD cohort (n = 72), and the Heptromic cohort (normal liver and cirrhotic samples of multiple etiologies, HCC samples excluded; n = 180). In all three cohorts, MPI was significantly down-regulated in samples with advanced fibrosis (Fig. 1A), regardless of the etiology. We found significant heterogeneity in expression values of MPI across tissues, suggesting that MPI is not the only factor contributing to this phenotype. Samples with low expression levels of MPI were enriched in fibrotic genes, such as collagen type I alpha 1 chain (COL1A1), collagen type IV alpha 3 chain, and collagen type I alpha 2 chain (Fig. 1B; Supporting Table S2). GSEA in samples with low MPI, regardless of level of fibrosis, revealed cell and focal adhesion clusters to be consistently ranked first among the gene sets significantly enriched in these samples across data sets (Fig. 1C; Supporting Table S3), a finding consistent with deranged cell adhesion as a hallmark of liver fibrosis.⁽²²⁾ These data demonstrate an inverse correlation between MPI gene expression and fibrosis in adult human liver disease.

FIG. 1.

MPI expression is decreased in human liver samples with more advanced fibrosis. (A) Box plots for MPI expression in samples with advanced degree of fibrosis (red) versus milder or normal phenotypes (green). Analysis conducted in three different data sets with genome-wide genomic data available. (B) Volcano plots of genes differentially expressed among samples above and below the 90th and 10th percentile, respectively, of MPI expression. (C) Enrichment plots of the top-ranked gene sets significantly enriched between samples with high and low MPI gene expression in the three data sets. Abbreviations: DGE, differential gene expression.

MPI LOSS BLOCKS HEPATIC DEVELOPMENT IN ZEBRAFISH

Zebrafish are an excellent system to study liver development and disease, including steatosis, fibrosis, and liver cancer.⁽²³⁾ We previously established a zebrafish model of Mpi depletion using morpholino oligonucleotides (MOs), which led to a significant decrease in Mpi enzymatic activity and hallmarks of hypoglycosylation comparable to those of MPI-CDG patients.⁽⁶⁾ To use this model to study the role of MPI in liver disease, we first confirmed that Mpi is abundantly expressed in zebrafish liver using a custom antibody raised against zebrafish Mpi, followed by immunofluorescent imaging in a transgenic line that expresses a fluorescent protein under a hepatocyte-specific promoter, Tg(fabp10a:nls-mCberry)⁽⁸⁾ (Supporting Fig. S2A). To assess for potential liver defects following Mpi loss, Mpi was efficiently depleted in this transgenic line using morpholino targeting as described⁽⁶⁾ (Supporting Fig. S2B). Severity of the whole-body phenotype in Tg(fabp10a:nls-mCberry) larvae correlated with Mpi activity and protein level (Fig. 2A,B and Supporting Fig. S2B). Importantly, consistent with our previous reports,⁽⁶⁾ mpi morphants developed smaller livers, which also correlated with severity of whole-body phenotype (Fig. 2C), suggesting a link between acute Mpi depletion and liver pathology. In addition, using a transgenic reporter for Notch activity to identify biliary cells (Tg(TP1:βglobin-GFP))⁽¹⁰⁾ mpi morphants developed an abnormal biliary system characterized by dilated branches of the biliary tree compared to control livers (Fig. 2D,E).

FIG. 2.

mpi morphant zebrafish have defects in liver development. (A) Overlay of fluorescent and bright-field images of 5 dpf Tg(fabp10:nls-mCherry) zebrafish larvae injected with standard control or mpi morpholino. Representative images, liver shown in red. (B) Mpi activity measurements from pooled 5-dpf zebrafish larvae separated into gross body phenotype groups as indicated, and reported as percent residual activity relative to standard (std) MO controls as mean ± SD from five independent clutches. **P < 0.01. (C) Liver size correlates with severity of gross body phenotype (classification previously described in detail in Chu et al.⁽⁶⁾). (D) Reconstruction of biliary epithelium in 5-dpf mpi morphant and control (siblings) Tg(fabp10:nls-mcherry; TP1:βiglobin-egfp) larvae. N = 3 per group. Representative images. (E) Biliary branches in mpi morphants were dilated when compared to controls. *P < 0.05, N = 3 standard control, 3 mpi morphant larvae.

We have previously reported on the overt mpi MO phenotype and did not identify defects that are suggestive of delayed embryonic development.⁽⁶⁾ However, to determine whether biliary abnormality was secondary to overall developmental delay, we quantified biliary length as a measure of developmental progression, and despite the abnormal branching, we found similar lengths between mpi morphants and controls (Supporting Fig. S2C). Furthermore, Annexin A4 staining with 2F11 antibody⁽²⁴⁾ identified the presence of differentiated cholangiocytes in mpi MO livers at 100 high-power field (hpf; Supporting Fig. S2D), at which point full differentiation and development of well-defined ducts are complete.⁽²⁵⁾ These data demonstrate the capacity for biliary cell differentiation in these livers, and suggest that biliary phenotypes are not attributed to global developmental delay. Together, these findings show that Mpi depletion in zebrafish larvae causes developmental defects in the liver, including small liver size and abnormal biliary development.

LOSS OF MPI ACTIVATES HSCs AND PROMOTES FIBROGENESIS

HSCs are the main driver of fibrosis, so we examined whether reduction of MPI is sufficient to drive HSC activation. Using a zebrafish transgenic reporter line for HSCs (Tg(hand2:EGFP)⁽⁹⁾), we first confirmed Mpi expression in zebrafish HSCs using immunofluorescence and found that Mpi colocalized with Hand2 (Fig. 3A). However, there was no difference in total HSC number between control and mpi morphants at 5 dpf, as calculated by the HSC/DAPI ratio (Supporting Fig. S3A–SC). To investigate the effects of MPI loss on HSC activation, we isolated livers from control and mpi MO embryos at 5 dpf and used qRT-PCR to measure expression of genes reflecting activation of fibrotic response.⁽²⁶⁾ Gene expression for the ECM proteins, collagen, type I, alpha 1b (col1a1b), actin, alpha 2, smooth muscle (acta2), platelet-derived growth factor receptor beta (pdgfrb), laminin subunit beta 4A (lamb4a), and tissue inhibitor of metalloproteinases 2 (timp2b), were significantly increased in mpi MO livers (Fig. 3B). There was no induction of transforming growth factor beta 1 (tgfb1) or platelet-derived growth factor subunit B (pdgfb; Fig. 3B), which encode growth factor ligands secreted by hepatocytes and biliary cells to induce HSC activation and proliferation⁽²⁷⁾ This suggested autonomous induction of HSCs following MPI depletion, which was not dependent on hepatocyte interactions. We previously established an mpi mutant zebrafish line (ZFIN:mpi^mss7),⁽¹²⁾ in which residual Mpi activity at 5 dpf is comparable to that found in MPI-CDG patients with hypomorphic mutations^(6,12) Like the embryonic lethal phenotype of mpi morphants, homozygous mpi^mss7/mss7mutants do not survive larval development.⁽¹²⁾ Therefore, to determine whether a sustained fibrogenic effect persisted past embryonic development and into adult stages, we analyzed livers from mpi^+/mss7 heterozygous adults (liver Mpi activity 69%; Supporting Fig. S3D). The mild Mpi depletion in heterozygous mutants demonstrated a persistent elevation in many fibrogenic genes at >1 year, including pdgfrb, collagen, type I, alpha 1a (col1a1a), col1a1b, and acta2 (Fig. 3C). Heterozygous mutant adult livers were sectioned and stained with Masson trichrome to visualize collagen as an indicator of fibrosis. Heterozygous mutant adult livers demonstrated abnormal bile ducts embedded within fibrotic stroma when compared to age-and sex-matched WT siblings with no fibrosis and normal bile duct profiles (Fig. 3D). These results suggest that even subtle loss of Mpi is correlated with increased fibrosis, and, collectively, that both acute and chronic loss of Mpi in vivo is sufficient to induce a sustained fibrogenic response in zebrafish liver during development and in adulthood.

FIG. 3.

MPI depletion induces fibrotic gene expression and fibrosis in zebrafish liver. (A) Immunofluorescent staining for zebrafish Mpi (blue) in 5-dpf Tg(hand2:EGFP) larvae. Images are representative stainings from at least two independent samples. (B) qRT-PCR on whole-liver cDNA from 5-dpf control and mpi morphant larvae. Statistical significance was determined by unpaired Student t test (*P < 0.05; **P < 0.01; N = 5–10 clutches analyzed per gene). Data are presented as mean ± standard deviation. (C) qRT-PCR on whole-liver cDNA prepped from adult WT (N = 9) and mpi^+/mss7 MT (N = 12) zebrafish. Statistical significance was determined by unpaired Student t test (*P < 0.05). Data are presented as mean ± standard deviation. (D) Representative images of Masson trichrome staining on paraffin-embedded whole-liver sections prepped from adult male WT (N = 5) and mpi^+/mss7 MT zebrafish (N = 4). Abbreviation: ctrl, control.

Because our transcriptomic analysis of HBV and NAFLD liver tissues demonstrated a correlation between lower MPI levels and worsened fibrosis, we asked whether Mpi depletion may sensitize the liver to injury and fibrosis in vivo. To test this, we treated zebrafish embryos from either WT or mpi^+/mss7 heterozygous adult incrosses (pooled embryos with 58% average Mpi activity⁽¹²⁾) with 2% ethanol from 96 to 120 hpf.^(13,26) Livers with Mpi depletion demonstrated a clear trend toward induction of fibrogenic genes (col1a1a, acta2, pdgfrb, and timp2b), but not for ligands tgfb1a or pdgfb, when treated with 2% ethanol, compared to ethanol-treated WT embryos (Supporting Fig. S4). These results suggest that loss of Mpi may exacerbate the fibrogenic response to known stressors such as ethanol, HBV, or NAFLD.

We next explored whether MPI depletion promotes fibrogenesis through direct activation of human HSCs. First, we verified that MPI is expressed in mammalian HSCs. We found MPI to be highly expressed in mammalian HSCs using immunofluorescence for the HSC marker, GFAP,⁽²⁸⁾ in normal mouse liver tissue (Fig. 4A), as well as robust MPI protein expression in human HSCs, as visualized by western blotting (Supporting Fig. S5A). In addition to HSCs, MPI expression was found in other liver cell types, including hepatocytes and liver sinusoidal endothelial cells (Supporitng Figs. S5B and S6). To determine the effects of MPI loss in HSCs, MPI was depleted in the human LX-2 HSC line using a specific siRNA against human MPI (MPI activity depleted to 44% of control). Downstream transcriptional responses were characterized through qRT-PCR. MPI depletion in these cells increased expression of COL1A1 and PDGFRB, whereas ACTA2 expression was unchanged (Fig. 4B). Similar to Mpi-depleted zebrafish whole livers, PDGFB and TGFB ligand expression remained unchanged from controls (Fig. 4B). These results were confirmed by depleting MPI using short hairpin RNA in LX-2 cells, which resulted in increased expression of COL1A1, PDGFRB, and ACTA2 (Supporting Fig. S5C–E). In addition, COL1A levels were increased in media from MPI-depleted HSCs compared to controls (Fig. 4C), suggesting that loss of MPI leads to enhanced HSC secretion of matrix protein COL1A to drive fibrosis. Changes in fibrogenic gene expression observed in MPI depletion were found to be in similar ranges to that of standard TGFB activation with the exception of ACTA2, and ligands PDGFB and TGFB (Fig. 4B), suggesting a potential difference in mechanism of HSC activation.

FIG. 4.

MPI depletion induces fibrotic gene expression. (A) Immunofluorescence staining in WT mouse liver for MPI and GFAP (HSC marker). Scale bar indicates 20 μM. Small outline in first panel represent magnified area. (B) qRT-PCR from siNC and siMPI LX-2 cDNA. Statistical significance was determined by paired Student t test (*P < 0.05; **P < 0.01; n = 5–9 independent tranfection experiments per gene, as indicated). Data are presented as mean ± standard deviation. (C) Western blotting analysis demonstrating Collagen 1A secretion into conditioned media collected from LX-2 cells transfected with siNC or siMPI. Shown are two independent sets of experiments, representative of five total experiments and with 2% FBS media as nonconditioned media control. Twenty microliters of TCA-precipitated media used per lane. Fold change in secreted COL1A signal intensity (relative NC controls) are shown, and total cell protein (μg) at time of collection are indicated. Abbreviations: NC, non-targeting control; TIMP1, tissue inhibitor of matrix metalloproteinase 1.

Given the expression of MPI in hepatocytes (Supporting Figs. S5B and S6A), we next investigated the paracrine effects of MPI depletion in hepatocytes on HSC activation. Cell mixing experiments demonstrated that the increased expression of fibrogenic genes in MPI-depleted LX-2 cells was neither significantly enhanced by conditioned media from MPI-depleted hepatocyte cultures (THLE5B and Huh7 cells; Supporting Fig. S7), nor did conditioned media induce HSC activation alone, suggesting that MPI loss directly activates HSCs (Supporting Fig. S7). These data demonstrate that acute and sustained MPI depletion are sufficient to activate HSCs and induce fibrogenesis in human HSC cells in vitro and in a whole vertebrate model.

MANNOSE SUPPRESSES HSC ACTIVATION AND FIBROSIS

Our data point to a role for MPI in HSC activation and liver fibrosis. Mannose supplementation is the standard treatment for MPI-CDG and works by directly producing Man6P by phosphorylation, bypassing MPI deficiency (Supporting Fig. S1).⁽⁴⁾ Based on results of therapy for MPI-CDG, we explored whether mannose supplementation attenuated liver phenotypes in vivo in zebrafish, and whether mannose reduced human HSC activation in culture. We have previously shown that mannose can rescue the mpi MO phenotype, including restoration of liver size. Therefore, we treated Mpi-depleted zebrafish embryos with 50 mM of D-mannose immediately following morpholino injection and assessed expression of profibrotic genes in the liver. Mannose mitigated induction of fibrogenic genes (except col1a1a) in Mpi-deficient zebrafish (Fig. 5A), demonstrating that early mannose supplementation is effective at protecting against fibrosis in the zebrafish model of MPI-CDG.

FIG. 5.

Mannose mitigates fibrotic gene expression. (A) qRT-PCR on liver cDNA prepared from 5 dpf control and mpi morphant ± 50 mM of mannose. Data are presented as fold change of mpi MO + 50 mM of mannose/mpi MO samples (N = 2). (B,C) qRT-PCR on cDNA from TWNT-4 cells (B) or 2% BCS or serum-activated LX-2 cells (C) treated with increasing concentrations of mannose. (D) Western blotting analysis of LX-2 cells following 24-hour serum starvation, then exposure to 50 mM of ethanol for 24 hours in the absence or presence of mannose. Band intensity quantified with ImageJ (NIH, Bethesda, MD), normalized to Tubulin. Average fold change compared to untreated samples is indicated. Statistical significance shown in (A–C) determined by paired Student t test, *P < 0.05; **P < 0.01. Abbreviations: BCS, bovine calf serum; EtOH, ethanol.

Building upon the findings in zebrafish, we sought to determine whether mannose regulates human HSC activation. We treated two independent, serum-activated human HSC cell lines with increasing concentrations of D-mannose and assessed expression of fibrogenic genes. After 24 hours of treatment, there was a dose-dependent effect, with 1 mM of mannose more effective than 0.25 mM of mannose, but with an effective plateau or saturation at concentrations higher than 1 mM (Fig. 5B,C). These data demonstrate a role for mannose supplementation in attenuating fibrogenic gene expression in culture-activated human HSCs in a concentration-dependent manner.

We next asked whether mannose treatment was effective against a clinically relevant inducer of hepatic fibrosis using ethanol, a potent activator of human HSCs.⁽²⁹⁾ Serum-starved LX-2 cells were treated with ethanol across a range of mannose concentrations. Untreated cells produced very little COL1A1 or PDGFRB, but, as expected, both fibrotic markers were strongly induced upon exposure to 50 mM of ethanol for 24 hours (Fig. 5D). Remarkably, cotreatment with mannose mitigated the induction of these fibrogenic markers, in a dose-dependent manner (Fig. 5D).

To investigate pathways altered in mannose-mediated HSC attenuation, we performed unbiased proteomics analysis using iTRAQ with LX-2 cells treated with and without 1 mM of mannose supplementation (n = 2 replicates per group). Among the differentially regulated pathway analysis of total proteins and of phosphoproteins in mannose-treated versus control experiments, glycolysis and glutamate metabolism were among the top pathways altered in HSCs treated with mannose (Table 1). Regulation of glucose and glutamine metabolism are both important for HSC activation based on these data and previous studies^.(30,31) Together, our data suggest a model whereby reduction of MPI, either through an endogenous genetic mutation or during the process of HSC activation, is sufficient to promote HSC activation and indicate an antifibrotic role for mannose supplementation in HSC attenuation.

Table 1.

Differentially Regulated pathway Analysis of Total proteins and phosphoproteins in Mannose-Treated Versus Control Human HSCs

Pathway name	Count	Wilcox_P
Total Protein
KEGG_ALANINE_ASPARTATE_AND_GLUTAMATE_	8	0.0054
METABOLISM
KEGG_GAP_JUNCTION	19	0.0056
KEGG_GLUTATHIONE_METABOLISM	14	0.0135
KEGG_FOLATE_BIOSYNTHESIS	3	0.0466
Phosphoprotein
KEGG_VIRAL_MYOCARDITIS	8	0.0017
KEGG_GAP_JUNCTION	9	0.0174
KEGG_GLYCOLYSIS_GLUCONEOGENESIS	5	0.0182
KEGG_VASCULAR_SMOOTH_MUSCLE_	9	0.0212
CONTRACTION
KEGG_MELANOGENESIS	7	0.0229

Discussion

Here, we focus on MPI as an important mediator of HSC activation and extrapolate therapies from this rare disease to clarify mechanisms of HSC activation. In silico analysis of human fibrosis samples ranging from HBV to NAFLD demonstrate an inverse correlation between MPI expression and fibrosis stage across etiologies. To further dissect this relationship, we have used the zebrafish as a whole-animal model in which we can modulate the mannose metabolism pathway.⁽⁶⁾ Mpi depletion in zebrafish results in hepatic fibrosis and biliary defects, similar to those observed in humans with MPI mutations. Our finding that mannose, a simple hexose sugar, mitigates the fibrotic response in culture-activated and ethanol-treated HSCs to suggest that inhibition of this enzyme may contribute to fibrosis, but that a simple treatment with mannose could possibly reverse it.

Zebrafish livers showed a more robust induction of fibrogenic gene expression in response to MPI loss than HSCs in culture (Figs. 3 and 4). HSCs comprise ~10% of resident liver cells,⁽³²⁾ and it is possible that systemic loss of MPI can have effects on non-HSC liver cells given that mpi morphant livers are smaller than livers in control siblings. mpi morphant livers contained abnormal vasculature, as marked by expression of Tg(kdrl:ras-mCherry) and measured by vessel branches per liver area (Supporting Fig. S3E). Our data suggest that a cell-autonomous activation of HSCs is, at least partially, responsible for this phenotype, because there was no induction of tgfb1 or pdgfb ligands in whole zebrafish livers from mpi morphants or mpi^+/mss7 adults, or MPI-depleted human HSCs (Figs. 3B,C and 4B, respectively). In addition, conditioned media from MPI-depleted hepatocytes did not induce a fibrogenic response in normal HSCs and did not exacerbate the effect of MPI knockdown in HSCs transfected with siMPI (Supporting Fig. S7). However, there are other paracrine growth factors that may contribute, and future studies clarifying which cell type(s) contribute to MPI-and mannose-mediated fibrotic changes are important. Another possibility is that, in vivo, MPI depletion leads to developmental defects stemming from a failure to remodel the developing biliary tree, which initiates liver fibrosis. In this case, conditional deletion of MPI may be used to understand its role in biliary development.

Previously published transcriptomic analysis of fibroblasts from multiple CDGs demonstrated increased expression of genes encoding ECM proteins compared to control fibroblasts.⁽³³⁾ Consistent with these data, our transcriptomic analysis of low-MPI human liver fibrosis samples and MPI-depleted HSC and zebrafish liver tissues demonstrated similar affected genes and pathways. These findings could implicate loss of protein glycosylation as an initiating event in fibrogenic response or, alternatively, suggest another function of MPI in fibrosis. Whereas many proteins are affected when protein hypoglycosylation is reduced, one pathway that merits further study is platelet-derived growth factor (PDGF) signaling, which is the most potent mitogenic driver of HSCs.⁽³⁴⁾ PDGF signaling is directly increased by loss of glycosylation—PDGFB is glycosylated and blocking its glycosylation increases PDGFB levels at the cell surface, leading to enhanced receptor binding and mitogenic activity.⁽³⁵⁾ Alternatively, aberrant glycosylation in PDGFB could result in improved presentation of the receptor-binding interface to PDGFRB to promote PDGF signaling.⁽³⁵⁾ Thus far, these studies have only been carried out in vitro, and the mechanisms and impact of altered glycosylation in PDGF signaling and HSC activation in vivo require further investigation.

As an alternative proposed mechanism, MPI reduction may activate HSCs through alterations in glycolysis and inter-related biosynthetic pathways. Recent studies have demonstrated increased bio-energetic and biosynthetic demands required for HSC activation dependent on glycolysis and glutaminolysis, given that blocking either of these pathways halts HSC activation.^(30,31) Unbiased proteomics analysis suggests that these same pathways are altered in mannose supplementation of HSCs with decreased activation. Our group has indicated that MPI is a regulator of Warburg metabolism and directs glycolytic intermediates into the hexosamine biosynthetic pathway,⁽¹²⁾ a key metabolic pathway which requires both glucose and glutamine as fuel. Recent publications demonstrate that mannose supplementation in vivo can suppress tumor cell proliferation⁽³⁶⁾ and steatosis.⁽³⁷⁾ Mannose may function similarly to suppress HSC activation through enhancing cell death or through changes in energy metabolism. Ongoing studies are examining the role of MPI and mannose in generating key substrates required for HSC activation and proliferation.

MPI-CDG patients receive oral mannose supplements to produce Man6P in order to bypass MPI deficiency (Supporting Fig. S1). Remarkably, mannose supplementation ameliorates most symptoms with restoration of glycosylation status of marker glycoproteins,^(4,5) but, in some cases, liver fibrosis continues to progress.⁽³⁾ Potential explanations for this lack of therapeutic response include: (1) irreversible, prolonged liver damage in utero and delay in diagnosis (MPI-CDG is often diagnosed well after infancy); (2) a nonglycosylation function of MPI and mannose therapy; or (3) higher doses of mannose are required to reverse the fibrosis. Current therapeutic guidelines for oral mannose supplements to treat MPI-CDG range from 0.1 to 1.0 g/kg of body weight daily and has been demonstrated in MPI-CDG patients to result in dose-dependent serum mannose concentrations of 250–490 μM.^(1,3–5) In our study, this concentration has little effect on HSC expression of fibrogenic genes in culture (Fig. 5). However, mannose of ≥1 mM significantly reduced fibrogenic gene expression, suggesting that increasing the dosage of mannose supplements may have a beneficial effect on mitigating liver fibrosis and may be effective as an antifibrotic more generally.

This study indicates a role for glycosylation in HSC activation and underscores the potential of glycosylation in liver fibrosis. Most important, we integrate findings using zebrafish in vivo models, HSCs in culture, and in silico analysis of human fibrotic liver samples. Together, the data indicate that MPI loss is a driver of liver fibrosis and suggest that modulating mannose metabolism pathways could reduce HSC activation and improve hepatic fibrosis.

Abbreviations:

ACTA2

actin, alpha 2, smooth muscle

CDG

congenital disorders of glycosylation

cDNA

complementary DNA

COL1A1

collagen type I alpha 1 chain

col1a1a

collagen, type I, alpha 1a

col1a1b

collagen, type I, alpha 1b

dpf

days postfertilization

DA PI

4’,6-diamidino-2-phenylindole

ECM

extracellular matrix

FBS

fetal bovine serum

FDR

false discovery rate

GFAP

glial fibrillary acidic protein

GSEA

gene set enrichment analysis

HBV

hepatitis B virus

HSC(s)

hepatic stellate cell(s)

ISMMS

Icahn School of Medicine at Mount Sinai

lamb4a

laminin subunit beta 4A

Man6P

mannose 6-phosphate

MO

morpholino oligonucleotide

MPI

mannose phosphate isomerase

NAFLD

nonalcoholic fatty liver disease

NP-40

nonidet P-40

PBS

phosphate-buffered saline

PDGF

platelet-derived growth factor

PDGFB

platelet-derived growth factor subunit B

PDGFRB

platelet-derived growth factor receptor beta

RT

room temperature

siMPI

small interfering RNAs targeting human MPI

TGFB

transforming growth factor beta

tgfb1

transforming growth factor beta 1

timp2b

tissue inhibitor of metalloproteinases 2

WT

wild type