Polysialic acid/neural cell adhesion molecule modulates the formation of ductular reactions in liver injury


  • Potential conflict of interest: Nothing to report.

  • This work was supported by the Daiichi-Sankyo Foundation of Life Science and The Medical Research Council, UK. The German partners (R.G.S., D.S., and B.W.) acknowledge financial support from the cluster of excellence REBIRTH.

  • See Editorial on Page 1469


In severe liver injury, ductular reactions (DRs) containing bipotential hepatic progenitor cells (HPCs) branch from the portal tract. Neural cell adhesion molecule (NCAM) marks bile ducts and DRs, but not mature hepatocytes. NCAM mediates interactions between cells and surrounding matrix; however, its role in liver development and regeneration is undefined. Polysialic acid (polySia), a unique posttranslational modifier of NCAM, is produced by the enzymes, ST8SiaII and ST8SiaIV, and weakens NCAM interactions. The role of polySia with NCAM synthesizing enzymes ST8SiaII and ST8SiaIV were examined in HPCs in vivo using the choline-deficient ethionine-supplemented and 3,5-diethoxycarbonyl-1,4-dihydrocollidine diet models of liver injury and regeneration, in vitro using models of proliferation, differentiation, and migration, and by use of mouse models with gene defects in the polysialyltransferases (St8sia 2+/4+/, and St8sia2/4/). We show that, during liver development, polySia is required for the correct formation of bile ducts because gene defects in both the polysialyltransferases (St8sia2+/4+/ and St8sia2/4/ mice) caused abnormal bile duct development. In normal liver, there is minimal polySia production and few ductular NCAM+ cells. Subsequent to injury, NCAM+ cells expand and polySia is produced by DRs/HPCs through ST8SiaIV. PolySia weakens cell-cell and cell-matrix interactions, facilitating HGF-induced migration. Differentiation of HPCs to hepatocytes in vitro results in both transcriptional down-regulation of polySia and cleavage of polySia-NCAM. Cleavage of polySia by endosialidase (endoN) during liver regeneration reduces migration of DRs into parenchyma. Conclusion: PolySia modification of NCAM+ ductules weakens cell-cell and cell-matrix interactions, allowing DRs/HPCs to migrate for normal development and regeneration. Modulation of polySia levels may provide a therapeutic option in liver regeneration. (Hepatology 2014;60:1727–1740)






alkaline phosphatase


alanine aminotransferase


aspartate aminotransferase


choline-deficient ethionine-supplemented


cytokeratin 19


chronic liver diseases


collagen 1


carbamoyl phosphate synthetase 1


cytochrome P450




ductular reaction




epithelial cell adhesion molecule


flow cytometry


glucose 6 phosphatase


glial fibrillary acidic protein


gamma-glutamyl transpeptidase


hematoxylin and eosin


hepatitis C virus


hepatocyte growth factor


hepatic progenitor cell


hepatic stellate cells












liver weight/body weight ratio


myristoylated alanine-rich C kinase substrate


messenger RNA


neural cell adhesion molecule


nonhepatic progenitor cell neural cell adhesion molecule positive cell


oncostatin M




primary biliary cirrhosis


polymerase chain reaction


polysialic acid






alpha-smooth muscle actin


SRY (sex determining region Y)-box 9


tyrosine aminotransferase


tryptophan-2, 3-dioxygenase


wild type.

The normal liver regenerates efficiently through hepatocyte division. However, in chronic or severe liver injury, a second regenerative compartment is activated, termed the ductular reaction (DR).[1] In severe liver injury, DRs spread from the periportal area into the hepatic parenchyma. Hepatic progenitor cells (HPCs) are thought to reside in the canal of Hering and expand after severe or chronic liver damage and regenerate hepatocytes and cholangiocytes. Clonogenic bipotential HPCs can be isolated from mice with DRs,[2] and recent lineage tracing in mouse models of liver injury demonstrated that HPCs within the DR can regenerate parenchyma.[3] A stereotypical niche of cellular components (myofibroblasts and macrophages), which direct HPC-mediated liver regeneration by Notch and Wnt signaling,[4] surrounds the DRs.[5] Laminin matrix surrounds the DRs and may influence HPC behavior. There is little information regarding the mechanisms controlling migration of HPCs/DRs or the detachment of cells from the cellular or acellular components of the niche.[5]

In this report, we have examined the functional role of the DR/HPC marker, NCAM (neural cell adhesion molecule), and its typical posttranslational modification, polysialic acid (polySia), in the DR/HPC niche. NCAM is a prototypic member of the immunoglobulin (Ig) family of adhesion molecules.[6] NCAM has three major isoforms: NCAM-120 (120 kDa); −140 (140 kDa), and −180 (180 kDa). NCAM mediates cell-cell adhesion by multiple modes, including homo- and heterophilic interactions and cell-matrix contact.[7] These interactions are modified by the posttranslational modification of NCAM with polySia. Two Golgi-resident polysialyltransferases (polySTs), ST8SiaII and ST8SiaIV, exist and transfer polySia onto complex N-glycans located in the fifth Ig-like domain of NCAM.[8] The enzymes, ST8SiaII and ST8SiaIV, are independently able to transfer polySia onto NCAM.[9]

Because of the size of the polySia chains and their high water-binding capacity, polySia on NCAM (polySia-NCAM) changes NCAM functions from adhesive to antiadhesive.[10] Accordingly, polySia-NCAM has been implicated in dynamic processes, such as cell migration and plasticity in the nervous system.[11] The crucial role of polySia in ontogeny is highlighted by the lethal phenotype of mice lacking two polySTs.[12] Mice with only partially depleted polySia synthetic capacity exhibit distinct, but milder, phenotypes.[13] The only known polySia-degrading enzyme is endosialidase (endoN), a phage-born enzyme.[14] Because of its stability and selectivity, endoN has been used in numerous animal models without producing adverse effects, and we have therefore used endoN here to cleave polySia.[15]

The liver has been described as an organ with low NCAM and polySia expression.[12] However, NCAM can be detected on specific cells in liver injury. Knittel et al. and Nakatani et al. reported on NCAM+ hepatic stellate cells (HSCs) and portal fibroblast in rat and human liver.[16-18] Fabris et al. reported that human reactive ductules with atypical morphology coexpressed NCAM.[19, 20] Zhou et al. provided evidence that NCAM+ cells in DRs represent bipotent HPCs in human.[21] This study and our previous study[22] showed NCAM expression in bile ducts and a wide range of human DRs. Importantly, those DRs immediately adjacent to mature hepatocytes were NCAM negative.[21] The biological functions of NCAM have not been previously reported on in the liver. Given that cell proliferation and migration from the niche is observed in liver regeneration, we hypothesized that the antagonistic functions of NCAM and polySia-NCAM in the regulation of cell adhesion and cell migration (proven for the nervous tissue) may play a significant role during liver development and regeneration.

Here, we show the expression of polySia in the liver. PolySia was found to increase markedly after liver damage and facilitated the migration of the DRs/HPCs away from the periportal niche during liver regeneration. Furthermore, polySia and its carrier, NCAM, were reduced during the differentiation of HPCs to mature hepatocytes.

Materials and Methods


Animals were housed in a specific pathogen-free environment and kept under standard conditions with a 12-hour day/night cycle and access to food and water ad libitum. All animal experiments had local ethical approval and were conducted according to UK Home Office Legislation. Male mice were fed either CDE (choline-deficient diet with 0.15% ethionine diet; Sigma-Aldrich, St Louis, MO) or DDC diet (3,5-diethoxycarbonyl-1,4-dihydrocollidine; 0.1% Purina 5015 mouse chow) for 11 days to C57BL/6 mice (n = 6; 7 weeks old) and S129P2 mice (10 weeks old, n = 6; Harlan Laboratories, Inc. Indianapolis, IN), respectively, for immunohistochemistry (IHC) and real-time polymerase chain reaction (PCR). To analyze the effect of endoN during liver development, 6 wild-type (WT) mice, 7 double heterozygous St8sia2+/4+/ mice and 12 double knockout (KO) St8sia2/4/ mice (provided by Prof. Rita Gerardy-Schahn, Hannover University, Hannover, Germany) were analyzed. All WT and St8sia mutant mice were sacrificed at day P8.5-P9.5. To analyze the effect of endoN during liver regeneration, S129P2 mice (more than 10 weeks old) were fed a DDC diet for 16 days. Induction of polySia was significantly higher in DDC diet mouse liver, compared to CDE diet mouse liver (Fig. 1C,D); thus, we employed the DDC diet mice in this experiment. To determine an effective dose of endoN that would work in vivo, we injected 3.14-19.69 μg of endoN/g of mouse in a single intraperitoneal (IP) injection on day 15 of a 16-day DDC injury protocol. Whereas all of the control injected mice were positive for polySia in western blots of whole liver tissue lysates, polySia could not be detected in endoN-injected mouse liver tissue lysates (Supporting Fig. 8B). Because side effects were not observed with any endoN dose, a 5-μg/g injection of endoN was used in the subsequent experiment and was injected four times (days 8, 10, 12, and 14) i.p. during the 16-day DDC experiment (Supporting Fig. 8A). Mice were divided into two groups: One group received endoN (n = 10) and control mice received vehicle (n = 9). Histological analysis, liver weight/body weight (LW/BW) ratio, serum blood tests, and real-time PCR were performed for analysis. Serum was analyzed using commercial kits for alanine aminotranspherase (ALT; Alpha Laboratories Ltd, Eastleigh, UK), albumin (ALB; Olympus Diagnostics Ltd, Southend-on-Sea, UK), aspartate aminotransferase (AST), and alkaline phosphatase (ALP; both Randox Laboratories, London, UK), according to manufacturer instructions.

Statistical Analyses

Statistical analyses were performed using GraphPad Prism5 software (GraphPad Software, Inc., La Jolla, CA). Data are presented as mean ± SD. The results were assessed using Mann-Whitney test. Differences were considered significant when the P value was less than 0.05.

Figure 1.

NCAM expression of normal and damaged livers in mouse by IHC. mRNA expression change of Ncam, St8sia2, and St8sia4. (A and B) NCAM expression in normal mouse livers and CDE (A) and DDC (B) diet-induced damaged livers. In normal mouse livers, NCAM+ cells could be detected only on bile ducts or bile duct surrounding cells. After liver damage, NCAM+ DRs and their surrounding cells expanded aggressively from the portal tract in both liver damage models. (C and D) mRNA change of Ncam, St8sia2, and St8sia4 before and after CDE and DDC diet-induced liver damage. After liver damage, whereas mRNA of Ncam and St8sia4 increased significantly in both models, mRNA levels of St8sia2 did not change significantly. Scale bar = 100 μm.

Further description of the materials and methods used are provided in the Supporting Information.


NCAM+ DRs Expand From the Periportal Area After Liver Injury

We studied IHC for NCAM from mouse models of hepatocellular injury and regeneration (CDE diet) and from biliary injury and regeneration (DDC diet). In normal livers, a few bile duct cells and their surrounding cells were positive for NCAM; however, after liver damage, NCAM+ DRs and closely associated NCAM+ niche cells began to expand from the portal area and spread into the parenchyma in both hepatocellular and biliary injury (Fig. 1A,B). These results suggested that NCAM+ cells are at least two populations, the epithelial cells of the DR and their surrounding niche cells after liver damage.

ST8SiaIV Is the Main PolyST Expressed in Damaged Mouse Livers

Two enzymes, the polySTs ST8SiaII and IV, are independently able to transfer polySia onto NCAM. Levels of polyST transcripts have been demonstrated to tightly correlate with the level of polySia expression.[23] To identify the polySTs responsible for NCAM polysialylation in the liver, gene expression levels were determined in parallel to Ncam expression in the CDE- and DDC-damaged livers. Both diets (CDE shown in Fig. 1C and DDC in Fig. 1D) led to a significant increase in expression of Ncam and St8sia4, whereas expression of St8sia2 remained at the level of control tissue. Because the product formation by polySTs has been described to be tightly linked to the messenger RNA (mRNA) expression level, these data provide strong evidence that polySia production in damaged livers results from ST8SiaIV activity23.23

Ductular Cells and Myofibroblasts Surrounding DRs Are NCAM+ and PolySia Is Predominantly Expressed at the Lateral Cell Surface of Bile Ducts and DRs

To precisely define the NCAM+ cells associated with the DRs in 11-day DDC-damaged livers, double fluorescence IHC was performed using NCAM together with markers for bile ducts and DRs (pancytokeratin [panCK] and SRY (sex determining region Y)-box 9 [sox9]; panCK can stain bile ducts and DRs/HPCs in the same manner with cytokeratin 19 [CK19] as shown in Supporting Fig. 1A,B; both anti-panCK and -CK19 antibodies [Abs] were used in this study), HSCs (desmin and glial fibrillary acidic protein [GFAP]), myofibroblasts (alpha-smooth muscle actin [α-SMA] and desmin), and hematopoietic cells (CD45). Double immunofluorescence (IF) between NCAM and panCK (Fig. 2A and Supporting Fig. 2A) and sox9 (Fig. 2B and Supporting Fig. 2B) revealed that bile duct and DRs (epithelial cells) frequently expressed NCAM, together with surrounding (nonepithelial) cells. α-SMA+/NCAM+ cells and desmin+/NCAM+ cells were identified to associate with the DRs; however, α-SMA/NCAM+ and desmin/NCAM+ cells were also detected in this area (Fig. 2C and Supporting Fig. 2C). We could not detect GFAP+/NCAM+ cells (Supporting Fig. 2D) or CD45+/NCAM+ cells (Supporting Fig. 2E). These results revealed that there are NCAM+ myofibroblasts associated with the DRs. When the relationship between NCAM+ cells and laminin (Supporting Fig. 2F) was analyzed in the DRs, it was found that almost all NCAM+ cells made contact with laminin. PolySia was predominantly located at the lateral cell surface of bile ducts and DRs (Fig. 2D). NCAM and polySia also could be detected in the same manner in the long-term (8 weeks) DDC-damaged livers (Supporting Fig. 2G,H), suggesting that the polySia-NCAM system is perpetuated during liver damage. Furthermore, to analyze polySia-NCAM in human chronic liver diseases (CLDs), we immunostained for CK19, NCAM, and polySia using serial sections of cirrhotic liver tissues caused by hepatitis C virus (HCV), alcohol, and primary biliary cirrhosis (PBC). All of these human diseases expressed NCAM and polySia in DRs, suggesting that polySia-NCAM is present in human CLDs (Fig. 3E).

Figure 2.

Characterization of NCAM+ cells and localization of polySia+ cells in damaged livers by IF staining. NCAM and polySia expression in human damaged livers. (A and B) Double IF staining between NCAM and the ductular/HPC marker panCK (A) and sox9 (B) revealed that some cells of the bile duct and DR expressed NCAM (A and B; yellow arrows). Some cells surrounding the DR also expressed NCAM (A and B; green arrows). (C) Double IF staining between NCAM and myofibroblast marker α-SMA revealed that some of the nonductular NCAM+ cells were α-SMA+ myofibroblasts (C, yellow arrow; green arrow shows α-SMA/NCAM+ cells). (D) IF staining for polySia revealed that polySia was predominantly expressed at the lateral cell surface of DRs (CK19+ cells were showed by serial section). (E) IHC using serial sections showed that chronic damaged human livers caused by HCV, alcohol, and PBC also expressed NCAM and polySia on CK19+ DRs. Scale bar = 100 μm (A and C) and 50 μm (B-E).

Figure 3.

Analysis of the role of polySia in vitro. (A) Double IF staining for NCAM and polySia revealed that BMOL cells were positive for NCAM and polySia. FCM analysis showed that almost all cells are positive for NCAM, some of which were positive for polySia with different levels. Western blotting analysis for NCAM without endoN revealed that anti-NCAM Ab detected diffuse signal, corresponding to polySia with NCAM, on lanes of P1.5 WT mouse brain and BMOLs. After cleavage of polySia by endoN, the bands were concentrated at 180 and 140 kDa in P1.5 WT mouse brain lane and at 140 kDa in BMOLs lane. In lanes of NCAM KO mouse brain, no band could be detected. (B) Adhesion assay. After addition of endoN, the number of adherent cells to laminin increased. (C) Cleavage of polySia by endoN did not affect cell expansion, even in the presence of HGF. (D) Aggregation assay. After the cleavage of polySia by endoN, the frequency of single cells/total cells decreased significantly indicating greater cell-cell aggregation. (E) Migration assay. Cleavage of polySia by endoN decreased HGF-induced cell migration significantly, compared to control (no endoN; dotted lines show the distal end of migrated cells). Scale bar = 100 μm.

PolySia Weakens Cell-Matrix Interactions, Prevents Cell Aggregation, and Promotes HGF Induced Migration of HPCs in HPC Culture Models

To analyze the role of polySia, we stained three HPC lines—the previously characterized HPC line, BMOL,[24] and our two established cell lines (HPC lines 2 and 3)—for NCAM and polySia. All of the HPC lines expressed NCAM and polySia (Fig. 3A and Supporting Fig. 3A,B). BMOLs had six times the expression level of polySia, compared to the other two HPC lines (Supporting Fig. 3C), and showed typical gene changes observed during a differentiation protocol induced by Wnt3A or OSM with a decrease in Ck19 and an increase in the hepatocyte marker, Alb (Supporting Fig. 4A,C). HGF induced relatively weak differentiation (Supporting Fig. 4B), but induced HPC expansion and migration in vitro (Fig. 3C,E). Therefore, BMOL cells are an ideal HPC line for further analysis of polySia and NCAM. Flow cytometric (FCM) analysis revealed that BMOLs are almost all positive for NCAM, but have varying levels of polySia positivity (Fig. 3A). To investigate the polySia-NCAM relationship further, western blotting analysis for NCAM was performed using WT P1.5 neonatal mouse brain (known to contain abundant polySia-NCAM as positive control), P1.5 NCAM KO neonatal mouse brain (negative control), and BMOLs with or without endoN. Without endoN, a diffuse signal of approximately 200 kDa was observed, which showed polySia with NCAM in WT P1.5 neonatal mouse brain lane and BMOLs lane. After treatment with polySia-specific endoN, bands were concentrated at 180 and 140 kDa in the WT P1.5 neonatal mouse brain lane and at 140 kDa in BMOLs lane, respectively. In NCAM KO mouse lanes, no band could be detected (Fig. 3A). To confirm the relationship between NCAM and polySia, further coimmunoprecipitation was performed. After immunoprecipitation for NCAM, immunoprecipitates were blotted for NCAM and polySia. This showed polySia with an NCAM diffuse signal (approximately 200 kDa) in both groups. After endoN digestion, western blotting for NCAM showed that this signal was concentrated at 140 kDa (Supporting Fig. 5). All of these results indicate that NCAM-140 is the main carrier of polySia in the BMOL HPC cell line (Fig. 3A). As observed in whole mouse liver, St8sia4 was also the main polyST in BMOL by real-time PCR (Supporting Fig. 6A). We examined the laminin-BMOL cell-matrix interaction by seeding BMOL cells on laminin-coated plates for 30 minutes. Pretreatment with endoN significantly increased cell adhesion, compared to control (Fig. 3B), suggesting that polySia weakens cell-matrix interaction in the HPC niche.

Figure 4.

Fate of polySia-NCAM during differentiation of HPCs to hepatocytes in vitro. (A) FCM analysis using BMOLs cultured for 2 (A) and 4 days (B) in differentiation conditions (Wnt3A, HGF, and OSM vs. control) markedly decreased the frequency of polySia+ cells. Especially, 4 days after addition of OSM strikingly decreased polySia. (C) Whereas addition of Wnt3A and HGF decreased the mRNA of St8sia4, addition of OSM increased the mRNA of Ncam and St8sia4 at day 1, despite the frequency of being polySia decreased at day 2. (D) Western blotting analysis of BMOL cell lysate with endoN after addition of OSM showed loss of NCAM protein. (E) In contrast, western blotting analysis of BMOL culture media supernatant with endoN revealed that the addition of OSM increased NCAM protein in the supernatant.

Figure 5.

Characterization NCAM+ myofibroblasts and migration of HPCs in the presence of NCAM+ myofibroblasts. (A) We characterized two cells lines (NHNPC1 and NHNPC2) as myofibroblasts that expressed NCAM and α-SMA. (B) Both NHNPC1 and NHNPC2 expressed α-SMA, Fibulin-2, Col-1, and Desmin, but were negative for Gfap mRNA. (C) FCM analysis of NHNPCs revealed that the frequency of polySia expression in NHNPCs was extremely low. (D) Schematic of the coculture between NHNPC1 and BMOL. (E) Distance of migration (24 hours). Coculture between NHNPC1 and BMOL with time lapse photography showed that BMOL migration was inhibited by 30% when polySia was cleaved with endoN (Supporting Videos 1 [control] and 2 [with endoN]). Scale bar = 100 μm.

Figure 6.

Analysis of the role of polySia during liver development. (A and B) HPC from E13.5 fetal HPCs expressed NCAM (A) and polySia (B). (C) H&E staining of WT, St8sia2+/4+/ mice and St8sia2/4/ mice revealed grossly abnormal portal tracts in KO mice and normal portal tracts in WT mice. IHC for panCK revealed that the number of tube-forming bile duct (bile ducts with typical morphology; black arrows) was less in St8sia2+/4+/ and St8sia2/4/ mice. (D) Frequency of tube-forming bile ducts/portal area revealed a significant decrease in the number of tube forming bile duct in St8sia2+/4+/ and St8sia2/4/ mice, compared to WT mice. (E) Up-regulation of many biliary and HPC marker genes (Ck19, Ggt, Sox9, Prom-1, and Ncam) could be detected in St8sia2/4/ mice, compared to WT mice. Scale bar = 100 μm.

We examined whether polySia affects HPC expansion, aggregation, and migration with or without endoN. We assessed BMOL expansion with or without HGF, having confirmed that cell expansion induced by HGF for 3 days of culture did not affect NCAM expression levels in the BMOLs (Supporting Fig. 6B). In both groups, cleavage of polySia by endoN did not affect cell expansion (Fig. 3C). In contrast, aggregation assays revealed that the presence of endoN decreased (>10%) significantly the ratio of single cells/total cells (55.1% ± 3.1% vs. 44.9% ± 3.3%; Fig. 3D), suggesting that polySia inhibits formation of stable cell-cell aggregation. Migration assays with a strong migratory stimulation factor HGF showed that endoN significantly decreased cell migration (6.53 ± 0.28 vs. 5.93 ± 0.26e[5] pixel); however, in the absence of migratory stimulation, the difference was not obvious (Fig. 3E). Taken together, these results suggest that polySia inhibits cell-cell and cell-matrix interactions and promotes HPC migration in response to HGF.

PolySia-NCAM Is Cleaved During Differentiation of HPCs to Hepatocytes by Oncostatin

Whereas DRs/HPCs themselves express polySia and NCAM, hepatocytes were negative for both factors, suggesting that, during differentiation, polySia-NCAM disappears from the cell surface of HPCs. Therefore, we checked the fate of polySia during differentiation to hepatocyte-like cells through the addition of Wnt3A, HGF, and oncostatin (OSM) to BMOLs[25] by FCM and found that the frequency of polySia+ cells decreased significantly, compared to untreated cells (Fig. 4A,B) 2 and 4 days after the addition of Wnt3A, HGF, and OSM. Four days after the addition of OSM, the frequency of polySia+ cells decreased markedly. We further checked the effect of Wnt3A, HGF, and OSM upon polySia transcription by looking for changes in Ncam and St8sia4 mRNA at day 1. Wnt3A and HGF decreased St8sia4 mRNA at day 1 and the frequency of polySia+ cells by FCM at day 2, suggesting that Wnt3A and HGF decreased polySia at the transcriptional level. However, OSM increased Ncam and St8sia4 gene expression at day 1 and decreased the frequency of polySia+ cells by FCM significantly at day 2, indicating that, during OSM-induced differentiation, the reduction in polySia+ cells was not mediated at the transcriptional level (Fig. 4C). Because the Ab used in FCM recognized polySia and not NCAM, to resolve the mechanism of polySia disappearance, western blotting for NCAM using 24-hour BMOL cultures separated into cell lysate and supernatant was performed. This revealed that the protein level of NCAM after endoN treatment in BMOL cell lysate decreased after addition of OSM (Fig. 4D), and the protein level of NCAM after endoN treatment in the BMOL supernatant increased after the addition of OSM (Fig. 4E). These results support the idea that NCAM was cleaved and released to the supernatant during differentiation.

Hepatic Myofibroblasts Contain NCAM+ Cells With Low PolySia Expression In Vitro

We previously identified NCAM+ DR-associated myofibroblasts and we therefore established two nonepithelial NCAM+ cell lines from CDE-damaged livers (nonhepatic progenitor cell neural cell adhesion molecule positive cells) NHNPC1 and NHNPC2. Both subfractions contained NCAM+ and α-SMA+ cells (Fig. 5A). Both fibroblastic cell lines contained hepatic myofibroblasts that expressed mRNA for myofibroblasts (α-sma, Fibulin-2, Collagen 1 [Col-1], and Desmin), but were negative for the inactivated HSC marker Gfap (Fig. 5B). Importantly, FCM analysis revealed that whereas polySia expression levels in HPC cell lines were high (Supporting Fig. 3C), polySia expression levels in NHNPCs were low (Fig. 5C).

EndoN Digestion of PolySia Reduces HPC Migration in Cocultures With Myofibroblasts

We analyzed BMOL migration with myofibroblasts to model cellular interactions in the damaged regenerating liver. Therefore, we modeled an “artificial niche” by culturing BMOLs upon a NHNPC1 cell layer with or without endoN (Fig. 5D). Time lapse photography revealed that red stained BMOLs readily migrate over the myofibroblasts; however, in the presence of endoN, mobility of BMOLs was significantly decreased (Supporting Videos 1 [control] and 2 [with endoN]). The average distance of migration dropped by >30% (from 97.4 ± 47.0 to 62.1 ± 31.6 μm) in endoN-containing cultures (Fig. 5E), indicating that polySia weakens HPC cell-cell interaction and thus aids migration of DRs/HPCs.

PolySia Is Required for Correct Formation of Bile Ducts

We isolated and expanded fetal HPCs from C57BL/6 mice[25] and confirmed that they expressed NCAM (Fig. 6A) and polySia (Fig. 6B). To understand further the role of polySia in liver development, we analyzed neonatal (8.5-9.5 days after birth) livers of WT, heterozygous St8sia2+/4+/, and double KO St8sia2/4/ mice. Because most of the double KO mice die within 20 days after birth, we analyzed livers from 8.5- to 9.5-day-old neonatal mice. Hematoxylin and eosin (H&E) staining of these livers revealed an obvious difference between WT livers and St8sia2+/4+/ or St8sia2/4/ mice, which lacked proper bile duct structures (Fig. 6C). To confirm this, we performed IHC for panCK (Fig. 6C) and quantified the tube-forming bile ducts (bile ducts with typical morphology)/portal area in all groups. The frequency of tube-forming bile ducts/portal area in St8sia2+/4+/ and St8sia2/4/ mice groups was decreased significantly, compared to the WT group (Fig. 6D and Supporting Fig. 7). We performed real-time PCR using bile duct and HPC markers (Ck19, gamma-glutamyl transpeptidase [Ggt], Sox9, Prominin-1 [Prom-1], Ncam, and epithelial cell adhesion molecule [Epcam]) and hepatocyte markers (Alb, tyrosine aminotransferase [Tat], carbamoyl phosphate synthetase 1 [Cps1], tryptophan-2, 3-dioxygenase [To], glucose 6 phosphatase [G6p], cytochrome P450 [Cyp]1a2, and Cyp2a12) from each group. There were no significant differences between WT and St8sia2+/4+/ mice, except the increase of mRNA levels of Ck19, which is expressed in bile ducts and DRs, in St8sia2+/4+/ mice. In contrast, there were marked differences between WT and St8sia2/4/ mice, with up-regulation of many biliary and HPC marker genes in St8sia2/4/ mice (Fig. 6E).

Figure 7.

Analysis of the role of polySia during DDC-induced liver damage. (A) PanCK (3,3′-diaminobenzidine; DAB) staining revealed that DRs/HPCs migrating into the parenchyma (black arrows) were more obvious in the control group, compared to endoN-treated mice. (B) In endoN-treated mice, DRs/HPCs migrating into the parenchyma/portal area decreased significantly. (C) Anaysis of LW/BW ratio revealed a significant decrease of LW/BW ratio in endoN-injected animals. (D) Serum blood test revealed that only serum ALT levels in endoN injection group were significantly elevated. (E) Real-time PCR analysis revealed that mRNA levels of Ck19, Alb, Tat, Cps1, and Col-1 were significantly increased in endoN injection group. Scale bar = 100 μm.

Figure 8.

Overview of the role of NCAM and polySia-NCAM in the niche. (1) In niche, NCAM-NCAM homophilic cell adhesion and NCAM-matrix heterophilic interaction contribute the stable settlement of HPCs. (2) During migration in the HPC niche, polySia is produced mainly by ST8SiaIV from HPCs. PolySia weakens cell-cell and cell-matrix interactions. In this context, HPCs can move more easily. With polySia, HGF is able to accelerate migration of HPCs during regeneration. (3) During differentiation into hepatocytes, HGF and Wnt stimulation regulate polySia at the transcriptional level and OSM contributes to cleavage of NCAM and polySia-NCAM from the cell surface. Therefore, hepatocytes do not express NCAM.

PolySia Is Required for the Migration of DRs into the Liver Parenchyma in the DDC Diet Liver Damage Model

To confirm the role of polySia during liver regeneration in vivo, we injected mice i.p. with endoN during the DDC-induced liver injury protocol (Supporting Fig. 8A).

The most obvious difference was the formation of DRs. In the endoN-injected group, most of the DRs stayed in the portal area, in contrast to the control group, where DRs migrating into the liver parenchymal area were more obvious (Fig. 7A,B and Supporting Fig. 9). Furthermore, in the endoN injection group, the LW/BW ratio was lower and serum levels of ALT were higher, compared to the control injection group (Fig. 7C,D). Finally, we checked mRNA expression of bile duct and HPC markers (Ck19, Ggt, Sox9, Prom-1, Ncam, and Epcam), hepatocyte markers (Alb, Tat, Cps1, To, G6p, Cyp1a2, and Cyp2a12), and fibrosis-related markers (α-sma, Desmin, and Col-1). The mRNA of Ck19, Alb, Tat, Cps1, and Col-1 were increased in the endoN injection group, compared to controls (Fig. 7E). Overall, these results suggest that endoN injection cleaves hepatic polySia, resulting in less ductular migration and an increase in susceptibility to hepatocellular injury during DDC diet liver damage.


Regeneration of hepatocytes and biliary epithelia by HPCs has been demonstrated using lineage tracing techniques.[26] However, the trigger for the formation of the DRs and detachment of HPCs from the niche into hepatocytes or bile ducts is not well understood. PolySia, through its action in weakening cell-cell and cell-matrix interactions, contributes to this important step during liver regeneration.

In the brain, the roles of NCAM and polySia have been extensively analyzed and the physicochemical properties of the large, highly hydrated polySia have been implicated in the modulation of cell-cell contact. The presence of polySia enhances migration of neural progenitor cells and the growth and targeting of axons.[11] Our studies have shown that, in the normal liver, NCAM-expressing cells are very limited in number and are confined to bile duct cells and fibroblasts near the portal tract. However, after liver damage, NCAM+ cells and polySia production begin to increase in the area of the DR/HPC niche. Both ductular and nonductular cells surrounding the DR express NCAM. HPCs have high levels of polySia and the main HPC polyST was ST8SiaIV, whereas myofibroblasts have low levels of polySia expression. In damaged livers, the main polyST was ST8SiaIV, which is consistent with a predominant HPC source. From these results, we conclude that HPCs produce polySia mainly through ST8SiaIV, and that this is the main mechanism by which polySia is produced in the DR/HPC niche of damaged livers. PolySia is predominantly expressed at the lateral cell surface of DRs and bile ducts. In vitro studies using a HPC cell line revealed that polySia weakens cell-matrix interaction and cell-cell aggregation of HPCs and increases hepatocyte growth factor (HGF)-induced HPC migration. Whereas NCAM expression in HPCs is described, the fate of NCAM is not known as their progeny; hepatocytes do not express NCAM. Here, we show that NCAM is cleaved during OSM-induced differentiation of HPCs to hepatocytes and suggest it likely that NCAM and polySia-NCAM is cleaved from HPCs as soluble NCAM during differentiation of HPCs into hepatocytes in damaged livers.

In mice lacking polySia expression (St8sia2/4/ double KO mice), severe phenotypic alterations have been described in the brain, suggesting a lack of neural plasticity resulting from a precocious presentation of polySia-free NCAM forms.[12] In liver of neonatal double KO mice, tube-forming bile duct formation was inhibited and many HPC markers and bile duct markers were up-regulated, suggesting that polySia is required for correct liver, as well as brain, development.

The role of polySia in liver regeneration was analyzed in the DDC-induced liver damage model by repeated endoN injection. DRs/HPCs in endoN-injected polySia cleaved mice migrated less, compared to controls, suggesting that polySia promotes effective migration of DRs/HPCs during liver regeneration. These endoN-injected mice showed increased susceptibility to liver damage with increasing serum levels of ALT, compared to controls. Adverse effects of endoN have not been reported to date; however, from these experiments, we found that endoN may exacerbate hepatocellular injury. Therefore, we would not advocate the clinical use of this compound in the setting of liver injury.

Data showing the presence of polySia-NCAM on short- and long-term mouse DRs and on human DRs in CLDs suggest that the polySia-NCAM system is perpetuated during liver damage, is not species specific, and has potential relevance in human liver disease. Recently, other functions of polySia have been reported on, including a role as a reservoir for neurotrophic factors and neurotransmitters.[27] It will be interesting, in future studies, to determine whether polySia has such roles in liver development and regeneration.

NCAM is, by far, the best-characterized and major polySia carrier28; however, recently other polySia-binding proteins, such as SynCAM1,[29] neuropilin-2,[30] and myristoylated alanine-rich C kinase substrate (MARCKS),[31] have been reported on. Of these binding proteins, polysialylation of SynCAM1 strikingly depends on the presence of ST8SiaII,[28] which is expressed at very low levels in the liver, compared to ST8SiaIV, making it highly unlikely that SynCAM1 is the major polySia-binding protein in the liver. We checked the immune expression of neuropilin-2 and MARCKS in CDE- and DDC-damaged liver tissues and in BMOL cells. Only neuropilin-2 could be detected and was restricted to endothelial cells of damaged livers, suggesting that these two proteins are not binding protein in the liver. Therefore, we concluded that NCAM is the major binding protein of polySia in the liver, a similar situation to that observed in the brain (Supporting Fig. 10).

Here, we propose a three-step mechanism for NCAM and polySia in liver regeneration: (1) NCAM contributes to cell-cell and cell-matrix interaction of HPCs in the unactivated HPC niche in normal liver; (2) during liver damage, polySia facilitates DR and HPC cell migration; and (3) during differentiation of HPCs to hepatocytes, NCAM is cleaved from the cell surface (Fig. 8). We believe this three-step NCAM/polySia-NCAM system modulates liver regeneration and, importantly, stops the further migration of cells once differentiated. Blockade of this “cellular lubrication system” can result in impaired DRs and liver regeneration. The therapeutic modulation of polySia levels and polySia-NCAM interactions may therefore be worth exploring in future studies.


The authors thank Prof. Frederic A. Troy II (University of California School of Medicine) and Prof. Ken Kitajima (Nagoya University) for providing critical comments on the study. The TROMA-III Ab developed by Rolf Kemler was obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD, and maintained by The University of Iowa, Department of Biology (Iowa City, IA).