Self‐Renewal Capability of Hepatocytic Parental Progenitor Cells Derived From Adult Rat Liver Is Maintained Long Term When Cultured on Laminin 111 in Serum‐Free Medium

Abstract

In this study, we investigated how the ability of hepatocytic parental progenitor cells (HPPCs) to self‐renew can be maintained and how laminin (LN) isoforms play an important role in their self‐renewal and maturation. Hepatocytes isolated from adult rat livers were cultured on hyaluronic acid to form colonies consisting of CD44⁺ small hepatocytes, which could be passaged on dishes coated with Matrigel. When second‐passage cells were plated on Matrigel, LN111, or LN511, HPPCs appeared on Matrigel and LN111 but not on LN511. We identified two types of cells among the second‐passage cells: Small, round cells and large, flat ones were observed on Matrigel, whereas the former and latter ones were specifically attached on LN111 and LN511, respectively. We hypothesized that small and round cells are the origin of HPPC colonies, and the binding to LN111 could be key to maintaining their self‐renewal capability. Among the integrins involved in LN binding, integrins α3 and β1 were expressed in colonies on LN111 more than in those on LN511, whereas β4 was more strongly expressed in colonies on LN511. Integrin α3^highα6β1^high cells could form HPPC colonies on LN111 but not on LN511, whereas integrin α6β1^low cells could not on either LN111 or LN511. In addition, neutralizing anti‐integrin β1 and anti‐LN111 antibodies inhibited the passaged cells’ ability to attach and form colonies on LN111 by HPPCs. Matrigel overlay induced second‐passage cells growing on LN111 to increase their expression of hepatic functional genes and to form 3‐dimensional colonies with bile canalicular networks, whereas such a shift was poorly induced when they were grown onLN511. Conclusion: These results suggest that the self‐renewal capability of HPPCs depends on LN111 through integrin β1 signaling.

Hepatocytic parental progenitor cells (HPPCs) exist among CD44+ small hepatocytes. Integrin α3high α6β1high cells could form HPPC colonies on LN111, but not on LN511. The self‐renewal capability of HPPCs depends on LN111 through integrin β1 signaling.

Abbreviations

3D

3‐dimensional

Ab

antibody

ALB

albumin

ANOVA

analysis of variance

BC

bile canaliculi

BM

basement membrane

C/EBPα

CCAAT/enhancer binding protein α

ECM

extracellular matrix

FACS

fluorescence‐activated cell sorting

FD

fluorescein diacetate

HA

hyaluronic acid

HiPSC

human‐induced PSC

HNF

hepatocyte nuclear factor

HPPC

hepatocytic parental progenitor cell

ICAM‐1

intercellular cell adhesion molecule‐1

Itg

integrin

LN

laminin

MGOL

Matrigel overlay

MH

mature hepatocyte

PCR

polymerase chain reaction

PSC

pluripotent stem cell

qRT‐PCR

quantitative reverse‐transcription PCR

SC

satellite cell

SH

small hepatocyte

Whole or segmental liver transplantation is widely chosen as the last option to save patients suffering from severe liver diseases, yet a persistent shortage of donor organs prevents most patients from receiving the benefits of transplantation. Thus, cell‐based therapies, such as cell transplantation, engineered hepatocellular tissue constructs, and bio‐artificial liver devices, have been considered as alternatives to whole‐liver transplantation.1, 2, 3 A large number of healthy hepatocytes are required for these therapies to compensate for their insufficient hepatic function. However, it is quite difficult to routinely obtain healthy human hepatocytes because of severe liver donor shortages and a lack of methods to expand functional hepatocytes.

Small hepatocytes (SHs) are a subpopulation of mature hepatocytes (MHs) that can act as hepatocytic progenitor cells.4, 5 Importantly, approximately 2.5% of MHs in young adult rat liver have the potential to become SHs, and this proportion gradually decreases with age.6 Rat and human SHs can proliferate to form colonies in serum‐free medium when cultured on hyaluronic acid (HA)‐coated dishes.7, 8 In addition, SHs consistently and specifically express CD44, an HA receptor.9 We recently reported that hepatocytic parental progenitor cells (HPPCs) exist among CD44⁺ SHs.10 HPPCs continue to proliferate and generate daughter cells after passages, which indicates that they possess the ability to self‐renew. In fact, rat HPPCs can potentially divide more than 50 times in a period of 17 weeks and over four passages. Although primary SHs require HA to attach to and grow on, under serum‐free conditions, most passaged cells do not attach to HA‐coated dishes. On the other hand, HPPCs were reported to expand on dishes coated with Matrigel derived from Engelbreth–Holm–Swarm sarcoma and containing components of basement membrane (BM)11; this suggests that Matrigel’s extracellular matrix (ECM) component is crucial for HPPCs to maintain their ability to self‐renew.

Adhesive interactions of epithelial cells with underlying BM are known to be instrumental for the development, differentiation, and maintenance of tissues. The major components of BM are laminins (LNs), type IV collagen, nidogen, and heparan sulfate proteoglycans.12 Among these components, LNs serve as the major adhesive proteins and mediate the adhesion of cells to BM. LNs are composed of three polypeptide chains, designated as α, β, and γ, and five α (α1‐5), three β (β1‐3), and three γ (γ1‐3) chains are recognized in mammals.13 Matrigel contains LN111(α1, β1, γ1) as a major constituent.11 The LNα1‐chain is expressed in fetal and early‐adult rat liver lobules but is not found in mature adults; instead, the LN α5‐chain is present in the portal triads.14 However, the transient expression of LNα1 has been observed in regenerating liver after partial hepatectomy.14 Integrins play central roles in the adhesion of cells to LNs.13 They are composed of noncovalently associated α and β subunits. To date, at least 24 separate integrins consisting of distinct combinations of α and β subunits have been identified in mammals. Among these, integrins α3β1, α6β1, α6β4, and α7β1 function as major LN‐binding receptors.12 The specificity of LN–integrin interaction is dependent primarily on LN α‐chains, and the integrin ligand specificity is determined primarily by the α subunits. Integrin β subunits play an auxiliary role in the specificity.12, 13

Preliminary experiments have shown that only 15% to 20% of CD44⁺ SHs attach to LN111‐coated or Matrigel‐coated dishes, although more than 40% attach to LN511‐coated ones.10 In the present study, we investigated how HPPCs’ self‐renewal capability can be maintained and how LN isoforms play a role in their self‐renewal and maturation. Although a population of HPPCs maintains a smaller size and the ability to self‐renew, cells that have a large cytoplasm and lose the ability to self‐renew can also be generated. Among integrins involved in LN binding, integrins α3 and β1 were expressed in colonies on LN111 more than in those on LN511, whereas β4 was more strongly expressed in colonies on LN511. Integrin α3^highα6β1^high but not α6β1^low cells could form HPPC colonies on LN111 but not on LN511. In addition, both neutralizing anti‐integrin β1 and anti‐LN111 antibodies inhibited the cells’ attachment to LN111 as well as colony formation by HPPCs. These results suggest that HPPCs’ self‐renewal capability depends on LN111 through integrin β1 signaling. Furthermore, Matrigel overlay (MGOL) induced second‐passage cells growing on LN111 to increase the expression of hepatic functional genes and to form 3‐dimensional (3D) colonies with bile canalicular networks, whereas such a shift was poorly induced when they were grown on LN511. Thus, HPPCs’ abilities to self‐renew and mature depend on LN111.

Materials and Methods

Animals

Six‐week‐old to 10‐week‐old male F344 rats (dipeptidylpeptidase IV⁺ strain; Sankyo Lab Service Corporation, Inc., Tokyo, Japan) were used for cell isolation. The animals were kept at a constant temperature of 23°C ± 1°C under a 12‐hour light/dark cycle with standard chow and water ad libitum. The experimental protocol was approved by the Committee on Laboratory Animals and complied with Sapporo Medical University guidelines (Approval Nos. 13‐132 and 13‐134).

Cell Isolation, Culture, and Passages

Primary SHs were isolated by the collagenase perfusion method, as described.7 The details of the methods for cell culture and passage have been reported.10 In the experiments using LNs, second‐passage cells were plated on 12‐well plates coated with LN111, LN511, LN521, and Matrigel (see Supporting Information for details). The medium was replaced every other day, and the cells were passaged every 4 weeks.

Adhesive Properties of Passaged Cells to LNs

To examine whether two different populations are generated during the culture of passaged cells, first‐passage cells, which had been cultured on Matrigel‐coated dishes for 21 days, were collected by trypsinization and plated onto LN111‐coated 35‐mm dishes. Three hours after plating, culture medium containing unattached cells was collected and transferred to LN511‐coated 35‐mm dishes. The medium was replenished for the cells on LN111, and 1 hour later, all cells on LN111, LN511, and Matrigel were harvested. The collected cells were used for gene‐expression analysis using oligo microarrays and quantitative reverse‐transcription polymerase chain reaction (qRT‐PCR) (see Supporting Information for details).

Morphological Analyses of Cultured Cells

Colonies were photographed with a phase‐contrast microscope equipped with a charge‐coupled device camera (Olympus Optical Co., Tokyo, Japan) to assess the number of colonies, the number of cells per colony, and the sizes of cells and colonies. At least 10 fields per dish or well were randomly selected, and three dishes or wells were examined per experiment. At least two independent experiments were performed. All captured images were processed using specialized software (Olympus cellSens Dimension Desktop 1.12; Olympus Optical Co.).

Comprehensive Gene Analysis and PCR

Differences in gene expression related to hepatic functions and stem cell properties among the cells attached to LN111, LN511, and Matrigel were analyzed using an oligo microarray spotted with 30,584 probes (SurePrint G3 Rat Gene Expression v2 G4853B; Agilent Technologies, Santa Clara, CA) and by qRT‐PCR (see Supporting Information for details).

Fluorescent Immunocytochemistry

For details on the fluorescent immunocytochemistry, see the Supporting Information.

Measuring Labeling Index

The cells were treated with 40 µM 5‐bromo‐2′‐deoxyuridine (BrdU) at 18 hours before fixation and the immunocytochemistry for BrdU (see Supporting Information for details).

Flow Cytometry

First‐passage cells, cultured on the Matrigel‐coated dishes for 28 days, were collected and incubated with mouse antirat integrin α6β1 or goat antirat integrin α3 antibodies in Dulbecco’s modified Eagle’s medium/F12 medium containing 10% fetal bovine serum for 30 minutes at 4°C. Integrin α3^highα6β1^high, α3^lowα6β1^high, or α3^lowα6β1^low cells were sorted using a fluorescence‐activated cell sorting (FACS) Aria cell sorter (BD Biosciences, San Jose, CA) and plated onto LN111‐coated, LN511‐coated, and Matrigel‐coated 12‐well plates. Data were analyzed using Kaluza Flow Cytometry Software version 1.1 (Beckman Coulter, Inc., Brea, CA) (see Supporting Information for details).

Inhibition Assay by Neutralizing Antibodies

First‐passage cells, cultured on Matrigel‐coated dishes for 28 days, were collected by trypsinization. To study integrin β1 and LN111 dependence, neutralizing antibodies against integrin β1 and LN111 were used, respectively (see Supporting Information for details).

Maturation of HPPCs by MGOL

First‐passage cells, cultured on Matrigel for 28 days, were separated and plated on LN111‐coated, LN511‐coated, and Matrigel‐coated dishes. The cells were cultured for 14 days and then either overlaid with Matrigel or untreated and then cultured for an additional 7 days. To assess the formation of bile canaliculi (BC), the cells were incubated for 30 minutes following the addition of 0.25% fluorescein diacetate (FD) (Sigma‐Aldrich Co., St Louis, MO) to the medium and then rinsed 3 times with phosphate‐buffered saline (PBS). Fluorescent images were immediately taken using a phase‐contrast microscope equipped with a fluorescence device (Olympus Optical, Co.). Gene expression by these cells was analyzed by qRT‐PCR.

Scanning Electron Microscopy

To observe the morphology of cells initially attached on LN111‐coated, LN511‐coated, and Matrigel‐coated dishes, the cells were fixed in PBS containing 2.5% glutaraldehyde solution (see Supporting Information for details).

Statistical Analysis

All data were analyzed using analysis of variance (ANOVA) with Tukey’s multiple comparison test or Kruskal–Wallis with Dunn’s multiple comparison test. Experimental results are expressed as the geometric mean ± SEM.

Results

We first confirmed our preliminary results.10 First‐passage cells were plated on dishes coated with ECMs, including LN‐α1 (LN111 and Matrigel) or α5 (LN511 and LN521). The adhesion rates of cells on day 1 were 15% to 20% and 40% to 50% on LN111 and LN511/521, respectively (data not shown), which was consistent with our preliminary results.10 The sizes of the attached cells on day 1 were measured, and those on LN511/521 were significantly larger than those on LN111 (Supporting Fig. S2A). Moreover, the features of cells on each ECM differed quite markedly in the early period after plating (Fig. 1 and Supporting Fig. S2B). At 3 hours, most cells attached to LN511/521 rapidly spread their cytoplasm to produce flattened features, whereas those on LN111 retained their spherical shapes. The different features of cells on each ECM were maintained even at day 1. Both types of cell that appeared on LN111 and LN511/521 were observed on Matrigel (Fig. 1A). Scanning electron microscopy revealed fine structures of cells on LN111, LN511, and Matrigel at day 1 (Fig. 1B). Cells on LN521 showed features similar to those on LN511 (Supporting Fig. S2A). Colonies with different shapes were formed on Matrigel at day 7 and later: One was a colony in which small cells compactly gathered, and the other featured cells with a large cytoplasm loosely contacting each other. The former was often found on LN111, whereas the latter was only on LN511. These results indicate that the typical features of colonies consisting of HPPCs (i.e., small, polygonal‐shaped cells) occurred on LN111 and Matrigel that contained LN111. This suggested that interactions between cells with LN111 but not with LN511 may contribute to maintaining the characteristics of HPPCs. Because the behavior of cells on LN521 was similar to that on LN511, subsequent experiments were performed using LN511‐coated culture devices.

Figure 1.

Phase‐contrast and scanning electron microscopy images of second‐passage cells cultured on Matrigel‐coated, LN111‐coated, or LN511 (E8 fragment)‐coated plates. The SHs sorted by anti‐CD44 Ab were cultured on Matrigel‐coated dishes for 28 days. (A) The first‐passage cells were trypsinized and plated on Matrigel‐coated, LN111‐coated, or LN511‐coated 35‐mm dishes. Photos were taken at 3 hours, 1 day (day 1), and 7 days (day 7) after plating. Scale bars: 200 µm. (B) Images of the second‐passage cells attached to LN111 (a,b), LN511 (c,d), or Matrigel (e‐g) at day 1 were obtained by scanning electron microscopy. Enlarged versions of the boxed regions of the cells on LN111 (a), LN511 (c), and Matrigel (e) are shown in (b), (d), (f), and (g), respectively. Scale bars: 120 µm (a,c,e) and 30 µm (b,d,f,g).

To examine whether the first‐passage cells consisted of different types of cells exhibiting distinct abilities to bind to different LN isoforms, we collected the cells that did not attach to LN111‐coated dishes and transferred them to LN511‐coated ones (Supporting Fig. S3A). Three hours after plating cells onto LN111‐coated dishes, the medium containing unattached cells was collected and poured into LN511‐coated dishes. One hour after plating, many cells had attached and exhibited a flattened shape on the LN511‐coated dishes, whereas the cells attached to the LN111‐coated ones still maintained spherical features even though 4 hours had passed since plating (Supporting Fig. S3B). Four hours after separation, the cells on dishes coated with each ECM were harvested. Their gene expression was comprehensively analyzed using oligo microarrays. Cells were also cultured for 7 days before assessing the efficiency of colony formation by counting the numbers of cells in images taken 7 days after plating. Colonies typical of HPPCs were formed in the LN111‐coated and Matrigel‐coated dishes, but not in the LN511‐coated ones (Supporting Fig. S3B). Colony sizes and the number of cells forming the colonies were quantified and are given in Supporting Table S3. The largest number of cells per colony was observed in the dishes coated with LN111. The average sizes of the cells on LN111 were significantly smaller than those on LN511, although the labeling index on day 7 did not differ among the three ECMs. Furthermore, the cells cultured on LN111 retained the ability to form HPPC colonies after passaging (Supporting Fig. S4A). The level of albumin (Alb) and Cd44 expression was maintained in the cells cultured on LN111, even after four passages (Supporting Fig. S4B). However, when cells cultured on LN511 were separated and replated onto LN111‐ and Matrigel‐coated dishes, they could not form HPPC colonies in any dishes.

Figure 2 shows heat maps of selected genes related to hepatic function and hepatic stem cells that were expressed in the cells 4 hours after passage. The expression of genes related to hepatic function in the cells on Matrigel showed a tendency to be higher than in the cells on both LN111 and LN511, although the absolute values in the cells on Matrigel were much lower than those in MHs. Representative gene expression (Alb, hepatocyte nuclear factor 4α [Hnf4α], and CCAAT/enhancer binding protein α[C/ebpα]) was confirmed using qRT‐PCR (Fig. 2C). The expression of such genes in the cells on Matrigel, LN111, and LN511 was much lower than that in MHs, although the expression of Hnf4α in the cells on Matrigel was as high as that in MHs. Afp was expressed in the passaged cells (Supporting Fig. S5). Genes related to cholangiocytes, such as Krt19, Sox9, and EpCAM, were more expressed in cells grown on LN511 than on LN111, whereas the expression of genes related to hepatic stem cells, such as Thy1, c‐Kit, and CD24, appeared not to differ between the cells grown on LN111 and on LN511. The fact that the gene expression pattern clearly differed between the cells grown on LN111 and on LN511 may indicate that the colonies formed during culturing of the first‐passage cells consisted of two distinct populations.

Figure 2.

Comprehensive analysis of gene expression of second‐passage cells attached to L111, LN511 (E8 fragment), or Matrigel. The relative expression of selected genes related to hepatic function (A) and stem cell markers (B) was analyzed using oligo microarrays. The first‐passage cells cultured on Matrigel for 28 days were trypsinized and plated on LN111‐coated dishes (Supporting Fig. S3A). Unattached cells in the medium were collected and replated on LN511‐coated dishes. The cells were harvested 4 hours after passage. Primary hepatocytes (MH‐1‐3) were used as a control. (C) Differences in gene expression related to hepatic function between the cells attached on L111, LN511, and Matrigel were analyzed by qRT‐PCR. MHs were used as a reference. Bars depict means ± SEM of three dishes. Abbreviations: GAPDH, glyceraldehyde 3‐phosphate dehydrogenase; Mat, Matrigel.

The separate populations also expressed distinct integrins (Fig. 3A). The gene‐expression pattern of integrin subunits (α‐chains and β‐chains) differed between cells grown on LN111 and LN511. The cells grown on LN111 were close to those grown on Matrigel. These microarray results were confirmed by qRT‐PCR (Fig. 3B). Integrin β1 was expressed in all cells. Integrins α1, α5, αv, and β3 were highly expressed in cells grown on LN111, whereas α3, α6, and β4 were highly expressed in cells grown on LN511, of which integrins α3, α6, α7, β1, and β4 are known as those of the LN receptor.12 The cells grown on Matrigel and primary CD44⁺ cells expressed the genes by cells grown on both LN111 and LN511. The gene expression of integrin α3, α6, α7, β1, and β4 was confirmed using qRT‐PCR (Fig. 3C). Both α6 and β4 genes were expressed more in the cells grown on LN511 than in those grown on LN111, whereas the β1 gene was significantly more highly expressed in the cells grown on LN111 than in the cells on LN511. The expression of the α7 gene was extremely low in all cells examined. To investigate the production of the proteins encoded by these proteins, we performed double immunocytochemistry for the integrins α6β1/β4 and integrin α3/HNF4α in the colonies (Fig. 4A). HPPCs in dishes coated with LN111 and Matrigel were positive for integrin α6β1 and integrin α3 and weak or negative for integrin β4 (Fig. 4). However, anti‐integrin β4 was stained in the cytoplasm of relatively large cells observed in dishes coated with LN511. These results indicate the discrepancy between gene and protein expression in hepatocytic progenitor cells and that the expression pattern of integrins differed between hepatocytic progenitor cells on LN111 and LN511.

Figure 3.

Expression of integrin subunit genes by cells attached to L111, LN511(E8 fragment),or Matrigel. (A) Comprehensive expression was analyzed using oligo microarrays. Each sample is the same as used for Fig. 2. (B) Expression of integrin subunit genes was analyzed by qRT‐PCR; primary hepatocytes (MHs), SHs sorted by anti‐CD44‐Ab (CD44‐positive), and second‐passage cells attached to LN111, LN511, or Matrigel at 4 hours after passage. (C) Specific integrin subunits (α3, α6, α7, β1, and β4) were analyzed by qRT‐PCR. The bars represent the means ± SEM of the three dishes. *P < 0.05. Abbreviation: GAPDH, glyceraldehyde 3‐phosphate dehydrogenase.

Figure 4.

Fluorescent immunocytochemistry for integrins in second‐passage cells on LN111, LN511 (E8 fragment), or Matrigel. Photos of fluorescent immunocytochemistry for integrin α6β1 and integrin β4 (A), and integrin α3 and HNF4α (B), are shown for the second‐passage cells cultured on LN111, LN511, and Matrigel for 2 weeks. 4′,6‐diamidino‐2‐phenylindole was used as a nuclear marker. Scale bars: 100 µm.

It has been reported that pluripotent stem cell (PSC)‐derived hepatoblast‐like cells maintain their ability to self‐renew when grown on LN111 through integrin α6β1.15 In this study, small‐sized cells comprising the colonies on Matrigel/LN111 were positive for integrins α6β1 and α3. Therefore, we used FACS to analyze integrin α6β1 expression on first‐passage cells that had been cultured for 21 days. We found that integrin α6β1 expression varied from low to high (Fig. 5A). Therefore, we sorted two distinct populations, integrin α6β1^high and integrin α6β1^low, plated them onto LN111‐coated and LN511‐coated wells, and cultured them for 28 days. Approximately 30% of plated integrin α6β1^high cells attached to the LN111‐coated wells, whereas only approximately 3% of integrin α6β1^low cells did the same (Fig. 5B). The attachment ratio of integrin α6β1^high and integrin α6β1^low cells to LN511‐coated wells was approximately 45% and 15%, respectively. The attachment rates of integrin α6β1^high and integrin α6β1^low cells to Matrigel‐coated ones were approximately 40% and 10%, respectively. Integrin α6β1^high cells could form typical colonies in LN111‐coated and Matrigel‐coated wells but not in LN511, whereas integrin α6β1^low cells could not in any ECM‐coated wells (Fig. 5C). Although the attachment rate of the sorted cells was much higher to LN511‐coated wells than to LN111‐coated ones, colonies typical of HPPCs developed only in LN111‐coated and Matrigel‐coated wells, not in the LN511‐coated ones. The fate of integrin α6β1^high cells on LN111 versus LN511 suggests that HPPCs may alter their cellular features and lose their ability to self‐renew on LN511. Integrin α3 was also expressed more highly in the cells on LN111 than those on LN511 (Fig. 3C). Therefore, first‐passage cells were sorted by the expression of integrins α6β1 and α3 (Fig. 6A). The sorted cells were divided into three subpopulations: integrin (Itg) α3^highα6β1^high (30%), Itg α3^lowα6β1^high (34%), and Itg α3^lowα6β1^low (36%). The attachment rates of Itg α3^highα6β1^high, Itg α3^lowα6β1^high, and Itg α3^lowα6β1^low cells on LN111 were approximately 5%, 1.5%, and 1%, respectively (Fig. 6B). Only integrin α3^highα6β1^high cells could form HPPC colonies in the LN111‐coated wells (Fig. 6C), whereas integrin α3‐negative ones could not. These results suggest that HPPCs express integrin α3^highα6β1^high.

Figure 5.

Attachment and growth of the cells sorted by anti‐integrin α6β1 antibody. (A) Cells isolated from first‐passage cells were cultured on Matrigel for 28 days and sorted by FACS. Expression of integrin α6β1 varied from low to high. Two distinct populations, integrin α6β1^high and integrin α6β1^low cells, were sorted. The sorted cells were plated on LN111‐coated, LN511(E8 fragment)‐coated, and Matrigel‐coated wells of 12‐well plates and cultured for 28 days. (B) The number of attached cells was counted 1 day after plating. Bars represent the means ± SEM of three wells. The data were analyzed by ANOVA with Tukey’s multiple comparison test. *P < 0.05. (C) Phase‐contrast images of sorted cells were taken, and colonies were observed on day 28. Integrin α6β1^low cells rarely formed a colony, and typical colonies observed at 28 days after plating are shown. Scale bar: 200 µm (day 1), 100 µm (day 28).

Figure 6.

Attachment and growth of the cells sorted by anti‐integrin α3 and integrin α6β1 antibodies. (A) Cells isolated from first‐passage cells were cultured on Matrigel for 28 days and sorted by FACS. Expression of integrin α3 varied from low to high in α6β1^high cells. The integrin‐α6β1^low/α3^low, α6β1^high/α3^low, and α6β1^high/α3^high cells were sorted. The sorted cells were plated on LN111‐coated wells of 12‐well plates and cultured for 28 days. (B) The number of attached cells was counted 1 day after plating. The data were analyzed by ANOVA with Tukey’s multiple comparison test. Bars represent the means ± SEM of three wells. *P < 0.05. (C) Immunocytochemistry for albumin/HNF4α was performed in the colony formed by sorted α6β1^high/α3^high cells that were cultured on LN111 for 28 days. Scale bar: 100 µm (day 28).

Because the integrin β‐chain plays an important role in signal transduction, an integrin β1‐neutralizing antibody (Ab) was used to determine whether the Ab inhibits colony formation by HPPCs. Although the Ab remarkably suppressed the attachment of cells to LN111, it did not inhibit attachment to LN511 (Fig. 7A). The initial 48 hours of treatment by the Ab was also enough to inhibit HPPCs from forming colonies on LN111 and Matrigel, and the number of colonies decreased 7 days after plating (Fig. 7B). In contrast, CD44⁺ HPPC colonies were not formed on LN511 in the presence or absence of neutralizing Ab. Next, we investigated whether HPPC colony formation depended on a signal from LN111; therefore, an inhibition assay using anti‐LN111 Ab was conducted (Fig. 7C). The attachment of cells plated onto LN111 and Matrigel in the presence of anti‐LN111 Ab was significantly inhibited, whereas that of the cells similarly plated onto LN511 was not. HPPC colony formation on LN111 and Matrigel was also clearly suppressed by the neutralizing Ab (Fig. 7D). This result demonstrates that a signal from LN111 is crucial for HPPCs’ growth.

Figure 7.

Integrin β1 and LN111 dependence of cell attachment and colony formation. First‐passage cells cultured on Matrigel for 28 days were collected by trypsinization. The isolated cells were incubated for 2 hours in the medium containing antirat integrin β1 hamster immunoglobulin M antibody or hamster immunoglobulin M isotype standard. The cells were plated on LN111‐coated or LN511 (E8 fragment)‐coated 12‐well plates. The rates of cell attachment (A) and colony formation (B) were assessed at day 1 and day 7, respectively. In the same way, the isolated cells were plated on LN111‐coated, LN511‐coated, or Matrigel‐coated 12‐well plates and cultured in the medium containing rabbit anti‐LN111 or goat antirabbit immunoglobulin G antibody. The rates of cell attachment (C) and colony formation (D) were assessed at day 1 and day 7, respectively. All data were analyzed using ANOVA with Tukey’s multiple comparison test. *P < 0.05. Bars represent the mean ± SEM of three wells.

HPPCs attached to both LN111 and Matrigel generated a different type of colony as well as HPPCs, the features of which were similar to those of the colonies observed on LN511 (Fig. 1A). Because LNα5 was initially not present in either the LN111‐coated or the Matrigel‐coated dishes, we examined the production of LN subunits in cells attached to L111‐, LN511‐, and Matrigel‐coated dishes. No cells expressed LNα1 (Supporting Fig. S6A). However, with the exception of MHs, the cells could produce the LNα5‐chain. Comprehensive analysis of ECMs produced by cells (Supporting Fig. S6B) illustrated that, in addition to many kinds of ECM‐related genes, including those encoding many more isoforms of collagens, nidogens, and tenascin than those attached to LN511, cells attached to both Matrigel and LN111 expressed LNα5, β1, and γ1, suggesting that they produce LN511. These results indicate that the LN511 produced by HPPCs on Matrigel and LN111 may support the survival and proliferation of their daughter cells that are typically found on LN511.

We then examined whether inducing maturation differed between cells on LN111 and those on LN511 by overlaying cell cultures for 7 days with Matrigel. Cells without MGOL (control) retained flat‐shaped colonies (Fig. 8A). As a result, cells on both Matrigel and LN111 formed 3D structures, whereas colonies on LN511 shrank (Fig. 8A). When FD was added to the culture medium, cells retained fluorescence in their cytoplasm in the colonies without MGOL, because BC between the cells were not well assembled. However, in colonies formed on both Matrigel and LN111, fluorescence was secreted into BC, illustrating the presence of well‐developed BC networks. After MGOL, cell colonies on LN111 and Matrigel increased the expression of genes related to hepatic functions, such as Alb, Tat, Tdo2, and Cyp1a2 (Fig. 8B). Alb and Tat expression, in particular, increased significantly in cells on both LN111 and Matrigel, and Tdo2 expression increased significantly in cells on Matrigel. In contrast, MGOL did not induce hepatic differentiation in cells on LN511. Because Cd44 and Krt19 expression was significantly suppressed in cells on both LN511 and Matrigel, inefficient induction of hepatocyte differentiation was not caused by maintaining an immature state or committing to the cholangiocyte lineage (Supporting Fig. S7).

Figure 8.

Maturation of colonies of second‐passage cells on each ECM after MGOL. The second‐passage cells were cultured on Matrigel‐coated, LN111‐coated, or LN511 (E8 fragment)‐coated dishes for 14 days and treated with medium containing or lacking (control) 5% Matrigel for 7 days. (A) To examine the formation of BC, the cells were incubated with FD. After incubation with FD, the phase contrast and fluorescent images were obtained. BC were formed after MGOL by the colonies of cells cultured on Matrigel or LN111. No colonies without MGOL formed BC. Scale bars: 200 µm. (B) Gene expression related to hepatic function in the cells, with or without MGOL, was analyzed by qRT‐PCR. All data were analyzed using Kruskul–Wallis ANOVA with Dunn’s multiple comparison test. *P < 0.05). Bars represent the mean ± SEM of three dishes.

Discussion

Our previous study revealed that HPPCs are present among CD44⁺ SHs.10 These cells retain their capability to proliferate and redifferentiate but gradually lose their ability to self‐renew. Furthermore, we found that two types of morphologically distinguishable colonies emerged during the culture of HPPCs: One type is round and consists of homogeneously small‐sized mononuclear cells, and the other consists of cells with large cytoplasm. Therefore, we speculated that the small cells forming round colonies were HPPCs, whereas the other large cells were daughter cells derived from HPPCs. In the present study, we clarified the mechanism behind the maintenance of self‐renewal of HPPCs, to analyze their characteristics and those of their daughter cells.

We have reported that HPPCs can be passaged only in dishes coated with Matrigel.10 In addition, fewer than 20% of passaged cells could attach to Matrigel‐coated dishes. When passaged cells were plated on a dish coated with LN111, their attachment rate was slightly lower than those on Matrigel. Whenever colonies grown on LN111 were passaged, less than 20% of the cells could attach, and this rate gradually decreased with increasing passages. These results suggest that the number of HPPCs decreases with increasing passages. In addition, many cells derived from HPPCs can attach not to LN111 but to LN511. In fact, when first‐passage cells are separated and plated on LN111, the unattached cells can still attach to LN511. Furthermore, HPPCs possessing the ability to self‐renew have never emerged on LN511. These results also suggest that HPPCs generate two distinctive cell populations and that many HPPCs perform asymmetric cell division. One division is LNα1‐dependent, and the other is LNα5‐dependent. In addition, the fact that LN111‐neutralizing Ab suppresses colony formation of HPPCs on LN111 also supports the idea that HPPCs’ self‐renewability is LN111‐dependent. The survival and proliferation of LNα5‐dependent daughter cells may be supported by LNα5 produced by passaged cells.

Maintenance of the ability of stem cells to self‐renew on LN111 has been reported in hepatoblast‐like cells derived from human‐induced PSCs (HiPSCs).15 Undifferentiated HiPSCs can be maintained on LN511 but not on LN111.16 These results suggest that LN111 could not only selectively maintain hepatoblast‐like cells but also eliminate residual undifferentiated cells. It has been reported that, in human liver development, LN is present in both the perisinusoidal space and portal tract.17 In fetal mouse liver, LNsα1 and α5 have been detected at the basal side of cholangiocytes, whereas only LNα5 was detected in the adult liver.18 Interestingly, EpCAM⁺ cells are bipotential during both the fetal and neonatal stages but later lose this ability. It is possible that EpCAM⁺ cells can keep their bipotency only in the presence of LN111. On the other hand, hepatoblasts isolated from mouse E14 fetal liver kept their bipotency when grown on LN111.19 Intercellular cell adhesion molecule‐1 (ICAM‐1)⁺ liver progenitor cells separated from late‐fetal and postnatal mouse livers could also maintain their ability to self‐renew on LN111 after passages.20 Thus, hepatic cells with fetal characteristics or less differentiated hepatocytes may acquire the ability to proliferate in an LNα1‐dependent manner.

In liver regeneration after partial hepatectomy, studies have indicated that regeneration may be achieved by the simple duplication of MHs, and hepatic stem/progenitor cells do not contribute to this type of liver regeneration.21, 22, 23, 24, 25 However, recent studies have revealed that MHs maintain the potential to dedifferentiate into immature hepatocytes and transdifferentiate into cholangiocytes, suggesting that they possess plasticity.26, 27, 28, 29 More recently, Axin2⁺ pericentral hepatocytes were reported to be the physiological progenitors of hepatocytes during cellular turnover in normal mouse liver26; however, there is still controversy about this.27 Conversely, we have reported that Sox9⁺EpCAM⁻ biphenotypic hepatocytes derived from MHs emerge adjacent to expanding ductal structures in 3,5‐diethoxycarbonyl‐1,4‐dihydrocollidine‐fed mouse livers.28 Sox9⁺ periportal hepatocytes have also been identified as hepatocyte progenitors during regeneration from chronic liver injury. Furthermore, we have shown that small, mononuclear hepatocyte progenitors are present in ICAM‐1⁺ fractions.20 The ICAM‐1⁺HNF4α⁺ mononuclear cells, isolated from healthy adult mouse livers, did not specifically localize to liver lobules and exhibited low expression of both Axin2 and Lgr5 genes. On the other hand, HPPCs express Lgr5, whereas their Axin2 gene expression is much lower than that of MHs. These accumulated findings suggest that MHs contain a subpopulation of cells possessing the potential to become self‐renewable progenitor cells with the ability to redifferentiate. Whether or not a single subpopulation of MHs possesses progenitor potential, and where these cells localize to in a liver lobule, remains to be determined.

It is important to elucidate the differences between ECM‐based ligands and their receptors in HPPCs and their progeny. Although the progeny proliferating on LN511 can still grow while retaining hepatocytic characteristics, this potential varies among cells and is lower than in HPPCs. It is known that LN511, LN332, and LN111 share a high affinity for integrin α6β1.16, 30 Using function‐blocking Abs, integrin α6β1 has been identified as the major cell adhesion receptor for cells on both Matrigel and LN111. It is also known that integrin α6β1 is abundantly expressed by PSCs31 and hepatoblast‐like cells derived from PSCs.15 On the other hand, integrin α6β1⁺ cells isolated from fetal mouse livers have been shown to form colonies consisting of cells expressing both hepatocyte and biliary markers.32 Of those cells, c‐Met⁺CD49f (integrin α6)^+/low c‐Kit⁻ cells were the most potent with respect to self‐renewal and differentiation.33

Tissue homeostasis in skeletal muscles relies on the activities of muscle‐specific stem cells, called satellite cells (SCs).34 When SCs are activated, a remodeling event, mediated by matrix metalloproteinases, leads to the deposition of LNα1 and LNα5 at the SC niche during muscle regeneration.35 Loss of LNα1 function impairs SC proliferation and self‐renewal, resulting in decreased long‐term regenerative ability. LN111 mediates its effect through integrin‐α6β1 signaling. The LN111/integrin‐α6β1 axis has been associated with the long‐term self‐renewal of hepatoblast‐like cells derived from HiPSCs,15 with the sphere‐forming capacity of human prostate cancer stem cells and neural stem cells36, 37 and with asymmetric cell division of Drosophila ovarian follicle stem cells.38 Therefore, it is likely that the interaction between LN111 and integrin‐α6β1 represents an ancient mechanism to maintain stem cell self‐renewal.35 In this study, only integrin‐α3^highα6β1^high cells could form HPPC colonies, and both anti‐integrin β1 and anti‐LN111‐neutralizing Ab suppressed the attachment of, and colony formation by, these HPPCs. Thus, HPPCs’ self‐renewal capabilities depend on LN111 by integrin β1 signaling.

While Matrigel‐containing LN111 plays an essential role in the expansion of HPPCs, it also induces hepatocytic maturation, which is accompanied by hepatic organoid formation. This phenomenon was also supported by the results from differentiation‐related gene expression. Considering that the administration of a single component of Matrigel, such as LN111, type IV collagen, or nidogen, did not induce 3D structural change of a colony by HPPCs (data not shown), the 3D structure of a combination of some constituents present in Matrigel may be important for inducing the conformational changes of HPPCs, which results in hepatocytic maturation.

In the present study, we have shown that HPPCs’ self‐renewal capability depends on LN‐α1 and that the signal is transmitted through integrin β1, and HPPCs can easily expand on LN111‐coated culture devices in serum‐free medium. Although some improvements to the procedure may be required to isolate human HPPCs, this method might generate large numbers of healthy hepatocytes from a limited number of donors or surgically resected liver.

Supporting information