Synthesized basement membranes direct the differentiation of mouse embryonic stem cells into pancreatic lineages

Embryonic stem (ES) cells have an unlimited replicative ability and the potential to differentiate into most cell types in an organism, including pancreatic cell lineages. Recently, the forced expression of transcription factors in mouse or human somatic cells was demonstrated to induce the reprogramming of cell fate and the generation of so-called induced pluripotent stem (iPS) cells (Nakagawa et al., 2008; Takahashi et al., 2007). Many researchers have reported differentiation procedures for pancreatic cell lineages from ES cells (Jiang et al., 2007; Kroon et al., 2008). Therefore, ES and iPS cells are proposed as a source of surrogate cells for regenerative medicine.

We previously reported that the M15 mesonephric cell line had the ability to induce mouse ES cells to differentiate into pancreatic progenitor cells in vitro (Shiraki et al., 2008b). Using this M15 cell procedure, ES cells are sequentially induced into mesendoderm, definitive endoderm and finally regional-specific definitive endoderm-derived cells in vitro, in a manner that mimics early embryonic inductive events in vivo. The undifferentiated ES cells receive fibroblast growth factor (FGF) stimulation, which activates the ERK signaling cascade and this triggers transition from self-renewal to lineage commitment. Then, activin and/or p38 MAPK, induces divergence into the mesendoderm lineage and activin induces the divergence of mesendoderm into the definitive endoderm. Once definitive endoderm is established, regional specification of the pancreatic cells can be manipulated by the activation of activin and bFGF signaling via modification of the culture conditions. We also examined the effect of feeder-free conditions on the differentiation of mouse ES cells into pancreatic cell lineages (Shiraki et al., 2008b). We found that only the final stage of differentiation into the regional-specific definitive endoderm required direct contact with M15 cells. Because even the fixed M15 cell layer still retained the ability to induce pancreatic differentiation, the extracellular matrices secreted from M15 cells were supposed to integrate into a solid environment in the surrounding of cell surface and to have a key role in guiding the differentiation into regional-specific lineages of the definitive endoderm. M15 cells secreted major components of basement membrane (BM), including laminin (LN)-511.

The BM is known to regulate various kinds of cellular functions such as adhesion, migration, proliferation and cell differentiation (Taipale and Keski-Oja, 1997; Yurchenco and Schittny, 1990). The BM has a highly integrated structure composed of extracellular matrix (ECM) molecules. The major components of most BM are type IV collagen, laminins, entactin (nidogen) and heparan sulfate proteoglycans such as perlecan. Among these components, LNs serve as the major adhesive proteins and mediate cell adhesion to BMs (Kleinman et al., 2003). These molecules are either provided from the epithelial cells or the surrounding mesenchymal cells (Furuyama et al., 1997; Furuyama and Mochitate, 2000). The secreted cytokines from epithelium or mesenchyme are integrated to the structure of the BM (Taipale and Keski-Oja, 1997; Yurchenco and Schittny, 1990). These molecules are assembled to form an optimal extracellular environment for developing, regenerating or maturing cells. Previously, we reported that immortalized type 2 alveolar epithelial (SV40-T2) cells on fibrillar collagen matrix can assemble a BM of LN-111 isoform in vitro, when co-cultured with Matrigel (Furuyama and Mochitate, 2000). The synthesized BM (sBM) could become available as a novel substratum only by removing the SV40-T2 cell layer without impairment on the substructure. Tracheal basal cells, which are airway epithelial progenitors, succeeded to terminally differentiate to ciliated cells on the novel substratum of sBM, but failed to do so when grown on an LN-111-coated or on type I collagen substrate (Hosokawa et al., 2007).

Here, we propose a novel approach to direct ES cells to differentiate on a solid environment of BM without a layer of M15 feeder cells. We show that mouse ES or iPS cells grown on a sBM substratum composed of the LN-511 isoform, could differentiate into definitive endoderm and further into pancreatic lineages. We reveal that one of the components of BM, laminin α5 (Lama5), mediates production of the pancreatic differentiation signal, and this signal is transduced through integrin β1. Moreover, we indicate that heparan sulfate proteoglycan secreted from the differentiated ES cells is also important for regional specification of definitive endoderm fate. Our results indicate that the sBM substratum functions as a guidance cue to support differentiation of ES cells into an endodermal lineage and is useful for the elucidation of the molecular mechanisms underlying the determination of pancreatic fate.

We have previously shown that M15 cells are potent in guiding differentiation of ES cells into the endodermal and then pancreatic direction (Shiraki et al., 2008b). Since even fixed M15 cells retained the ability to induce pancreatic differentiation, solid environments of extracellular matrices produced by M15 cells appeared to have a role in guiding this differentiation. LN families are heterotrimers of glycoprotein that consists of α-, β- and γ-chains and are one of the major components of BMs. The multiplicity of LN isoforms (α1-5, β1-3 and γ1-3 chains) endows BM diversity (Colognato and Yurchenco, 2000). Microarray analysis of M15 cells showed that Lama5 was highly expressed in M15 cells (Fig. 1A). Compared with OP9 or PA6 cells, which did not show endodermal inducing activities, M15 cells expressed Lama5 transcripts at an elevated level (Fig. 1B).

To examine whether Lama5 has an important role in guiding the differentiation into pancreatic progenitor cells, we tested the effects of Lama5 knockdown in M15 cells. In M15 cells transduced with Lama5 shRNA (Lama5 KD), but not the non-silencing (NS) control, the expression level of Lama5 protein is substantially lowered (Fig. 1C). Using NS or Lama5 KD M15 cells, we performed a pancreatic differentiation assay. We used SK7 ES cells, which express GFP driven by the Pdx1 promoter (Shiraki et al., 2008b). At differentiation day 8 (d8), Pdx1-positive pancreatic progenitor cells were significantly decreased in ES cells grown on Lama5 KD M15 cells (Fig. 1D), without a significant difference in the proportion of E-cadherin-positive CXCR4-positive definitive endoderm (DE) cells (Higuchi, data not shown). Transient shRNA transfection experiments gave similar results as the above stable shRNA transfected M15 line. These results indicate that Lama5 has an important role in guiding ES-cell differentiation into lineage-specific DE cells.

Our results suggest the importance of the BM components in the process of pancreatic differentiation, and the potential of using extracellular matrix as an inducer of pancreatic differentiation. Previously, we reported a culture substratum, the sBM, which had a lamina densa of the mouse LN-111 (also known as LN-1) isoform derived from Matrigel® on a fibrillar collagen matrix (Furuyama and Mochitate, 2000; Hosokawa et al., 2007). Here, we prepared an sBM substratum of human recombinant LN511 (rLN511 sBM), in which rLN-10 cells (Doi et al., 2002) were used, according to the schema in Fig. 2A. First, a stiff matrix of fibrillar type I collagen (fib) was adsorbed by maleic anhydride-styrene copolymer (MAST) with side chains of oligo-GlcNAc. After the coating of fib with GlcNAc ligands, rLN-10 cells (Doi et al., 2002) were seeded on it (Fig. 2Aa). After 2 weeks of culture, rLN-10 cells formed a lamina densa beneath the basal surface (Fig. 2Ab,Ba,b,c) and they were connected through anchoring filaments (Fig. 2Ba). LN isoforms of major BM components were integrated into the lamina densa (Fig. 2Bd,e). As rLN-10 cells synthesize only hLN511 as a LN source, the LN isoform of the assembled lamina densa should be α5β1γ1. rLN-10 cells also secreted other BM components, such as type IV collagen, nidogen-1/entactin-1 and perlecan of heparan sulfate proteoglycan, which were integrated beneath the basal surface. Fig. 2Bd,e shows that LN (2B-d1 and e1) and type IV collage (2B-d2 and e2) are integrated in the lamina densa. Unless fib matrix was coated with oligo-GlcNAc ligands, no lamina densa could be assembled, nor were BM components integrated. Since HEK293 cells scarcely secreted LN, they failed to form a lamina densa structure, nor did they integrate type IV collagen, even on oligo-GlcNAc-coated fib (data not shown). After culture, rLN-10 cells were removed by treatment with alkali and detergent (Fig. 2Ac). The de novo sBM substratum was thus prepared in vitro, immersed in the preservation solution, and then stored at −75°C until use.

To start differentiation of mouse ES or iPS cells, the cells were seeded on the sBM substratum (Fig. 2Ad). As we have previously reported, the addition of activin and bFGF potentiates DE differentiation, and these factors were therefore also added to the ES cell culture on the rLN511 sBM substrata. The differentiation of ES cells into E-cadherin-positive CXCR4-positive DE cells was quantified by flow cytometry (Fig. 3A). When assayed on day 8 (d8), 27.61% of the cells grown on the sBM, and 50.89% of those on M15 were E-cadherin-positive CXCR4-positive DE cells. Quantitative RT-PCR analysis revealed that the expression of other DE markers, Sox17 and Foxa2, was detected on d8, and decreased on d15 in sBM samples, whereas a regional-specific DE marker, Pdx1, was detected on d8 and increased by d15. By contrast, the expression of an undifferentiated ES marker, Pou5f1 (also known as Oct3/4), was significantly decreased (Fig. 3B). Taken together, these findings indicate that rLN511 sBM substratum is effective in directing ES cells into a DE fate.

We then examined differentiation from the DE into the pancreatic lineages. In our previous study, we demonstrated that M15 expressed a high level of the retinoic acid (RA) synthesizing enzyme gene Aldh1a1 (Raldh1), and that inhibition of RA signaling strongly inhibited the differentiation of pancreatic progenitor cells (Shiraki et al., 2008b). Thus, RA was added to the culture medium from d10 to d13 (Fig. 4A). When SK7 ES cells were grown on rLN511 sBM, expression of Pdx1 became detectable at d10 and reached a maximum at d15 (Fig. 3B). The co-staining of anti-GFP antibody and the nuclear dye DAPI reveals the nuclear localization of the Pdx1-positive signals (supplementary material Fig. S1). The Pdx1 transcript was observed first on d8 (Fig. 3B), and Pdx1-positive cells accounted for 20.7% of the total differentiated ES cells (Fig. 4C). Thereafter, Pdx1-expressing cells started to aggregate into three-dimensional clusters. On d28, GFP expression was still observed in the clusters of differentiated cells. To trace the expression of insulin, an Insulin1 (Ins1) GFP ES cell line, ING112, was established from a transgenic mouse line bearing the insulin promoter driving GFP expression (Hara et al., 2003). We confirmed that ING112 cells differentiate into DE and Pdx1-expressing pancreatic progenitor cells on M15 (supplementary material Fig. S2), and then tested the differentiation on rLN511 sBM. As a result, Ins1-positive cells were detected on d26 and increased at d28 (Fig. 4D).

The expression of pancreatic marker genes was analyzed by RT-PCR. Correlating with GFP expression, the transcripts of Pdx1 and Ins1 were detected at d15 and d28, respectively. The expression of Ins1 was only detectable in the cells grown on sBM, not those grown on M15 (Fig. 5A). By contrast, other endocrine (Gcg, glucagon; Sst, somatostatin) or exocrine markers (Amy, amylase; Ptf1a, pancreas specific transcription factor 1a) were detected in ES cells differentiated either on sBM or on M15 cells. These results indicate that the sBM procedure potentiates not only the DE or pancreatic progenitor, but also pancreatic β-cell differentiation. Molecular markers associated with endocrine differentiation (Neurod1, Nkx2-2, Pax6, Nkx6-1, Isl1, Glut2 and Iapp) were expressed in cells differentiated by both procedures (Fig. 5B). Taken together, these results indicate that rLN511 sBM substratum induces ES cells to differentiate into the DE lineages of the pancreas including insulin-expressing β-cells. A mouse iPS cell line, 20D-17, which was confirmed to differentiate into DE cells and Pdx1-expressing cells on M15 (supplementary material Fig. S2), was also tested on sBM. In the 20D-17 iPS cell line, Pdx1 expression was detected on d16 and Ins1 expression on d28 of differentiation, which is similar to the results obtained in ES cells (Fig. 5C). These results suggest that sBM is a useful tool for guiding mouse ES cells and iPS cells to differentiate into regional-specific definitive endoderm lineages.

We then focused on the mechanism by which sBM directs pancreatic differentiation. We demonstrate that Lama5 knockdown in M15 cells showed a decreased potential for guiding ES cell differentiation into Pdx1-expressing cells. Moreover, rLN511 sBM was mainly assembled by LN511, which consists of Lama5, β1 and γ1 chains. From this, we speculate that the key component of the guiding signal from sBM is Lama5. We then tested the role of the LN receptor, integrin. Integrin is known to interact with LNs, and one of the integrin components, integrin β1(Itgb1), is expressed in differentiated ES cells (N.S. and Y.H., unpublished results). Since integrin β1 is also expressed on the cell surface of the adult islet (Nikolova et al., 2006), we then performed knockdown of Itgb1 in differentiating ES cells to test its role in the differentiation of Pdx1-expressing cells. NS control or Itgb1 knockdown (Itgb1 KD) lentivirus was added on d10 (Fig. 6A). On d15, the expression level of Itgb1 transcripts was analyzed and Pdx1-positive cells were evaluated. As a result, Itgb1 KD samples expressed a lower level of Itgb1 transcripts compared with NS samples (Fig. 6B) and yielded a significantly reduced proportion of Pdx1-positive cells (Fig. 6C). These results suggest that the guiding signal from sBM is transduced through Itgb1 in ES cells.

Knockdown of Lama5 and Itgb1 indicates the importance of LN-integrin signaling in pancreatic differentiation. Since knockdown of LN-integrin signaling partially decreased the number of Pdx1-expressing cells, we speculated that other molecules might also be involved in this process. We then focused on the heparan sulfate proteoglycans (HSPGs), because they are known to act as a reservoir for many kinds of soluble factors. HSPG2 (also known as perlecan), is a major component of the BM. Hspg2 transcripts are detected in undifferentiated or differentiated ES cells (supplementary material Fig. S3). Moreover, it is also reported that HSPGs are involved in the postnatal β-cell proliferation or β-cell function (Takahashi et al., 2009). To test whether the HSPGs have a role in pancreatic differentiation, we generated Hspg2-knockdown SK7 ES cells. To exclude the effect of heparan sulfate chains in the sBM, pre-treatment of the sBM with heparitinase was carried out at d0. Moreover, to degrade heparan sulfate side chains in other HSPGs expressed in ES-cell-derived differentiated cells, heparitinase treatment at d10-d15 was performed (Fig. 7A,B). As a result, Hspg2 knockdown or heparitinase treatment caused a reduction of Pdx1-positive cells. Moreover, the effect yielded by a combination of Hspg2 knockdown and heparitinase treatment showed a synergistic effect (Fig. 7C). These results strongly suggested that Hspg2 and the heparan sulfate chains of HSPGs have an important role in pancreatic differentiation.

RT-PCR analysis revealed that mouse ES cells grown on the sBM substratum differentiated into pancreatic cell lineages (Fig. 4). Previously, our group and others have demonstrated that engraftment of the differentiated ES cells into an in vivo environment improved the maturation of ES-cell-derived pancreatic cells (Kroon et al., 2008; Shiraki et al., 2008b). We performed transplantation of cells grown on sBM to evaluate differentiation into pancreatic β-cells in the ING112 ES cell line. ING112 cells grown on the sBM were collected at d28 and transplanted under the kidney capsules of severe combined immunodeficient (SCID) mice. Four weeks after transplantation, the grafts were recovered and analyzed. Many Ins1-positive cell clusters were found in the recovered graft (Fig. 8A). Quantitative RT-PCR analysis revealed that pancreatic endocrine (Ins1, MafA, Ppy, Sst) and exocrine (Amy) markers were highly upregulated (Fig. 8B). Ins1 and Amy were upregulated most significantly, to a level higher than that in the fetal pancreas (Fig. 8C). Immunostaining analysis revealed that GFP expression overlapped with insulin or C-peptide staining. Ins1 signals are detected in the nuclei and the cytoplasm, and almost completely overlapped with insulin signals (supplementary material Fig. S4). These insulin-positive cells also expressed the mature β-cell markers MafA and Nkx6-1 (Fig. 8Da-d). The Ins1-positive cells resembled normal islet β-cells in that they formed clusters and were surrounded by Amy-positive exocrine cells (Fig. 8De). The expression of other pancreatic markers (Gcg, Sst, Ppy and DBA) was also observed (Fig. 8De-h). Measurement of insulin contents of the recovered grafts revealed that up to 113 ng of insulin was detected in the Ins1-positive graft, and this approximately correlated with the insulin content of one islet, which is 1000 cell-equivalents. However, glucose-responsive insulin secretion in the Ins1-positive graft was not observed (supplementary material Fig. S5).

Taken together, our results show that the BM increases the efficacy of ES-cell differentiation into pancreatic endocrine β-cells and that cells grown on the sBM substratum can differentiate into all pancreatic cell lineages and form islet-like structures in vivo.

Previous studies indicated that the addition of soluble growth factors is sufficient to support differentiation from ES cells into DE cells. In our study, however, the differentiation from DE cells into pancreatic lineages required a direct interaction with M15 cells (Shiraki et al., 2008b). These findings indicated the possible participation of the ECM components in guiding the differentiation of ES cells.

Microarray analysis of M15 cells revealed high-level expression of collagen type IV and Lama5, which are major components of the BM, and type V collagen of lamina fibroreticulas. In the current model, BM assembly occurs via linking of a type IV collagen network with an LN network by entactin (Yurchenco and Cheng, 1994). To test whether the BMs produced from M15 cells have an important role in the regional specification of DE cells, we inhibited Lama5 expression in M15 cells, which resulted in reduced differentiation of ES cells into the Pdx1-expressing cells (Fig. 1D). These results suggested that the Lama5 chain has an important role in guiding the differentiation of DE cells into pancreatic cells.

Dishes coated with type IV collagen have been used as a substratum for DE differentiation (Tada et al., 2005; Yasunaga et al., 2005). Type I collagen or LN-coated dishes are reported to be useful tools for hepatic differentiation of mouse ES cells (Teratani et al., 2005). These molecules are candidate ECM components, which have important roles in the regionalization of DE cells. We have previously investigated BM formation by culturing epithelial cells in vitro (Furuyama and Mochitate, 2000) and reported the terminal differentiation of tracheal basal cells into ciliated cells on a novel substratum of sBM (Hosokawa et al., 2007). Since Lama5 mediates the guiding signal from M15 cells, we reconstituted the ECM environment using the sBM prepared from rLN-10 cells. As expected, mouse ES or iPS cells were sequentially induced into endodermal and pancreatic lineages (Figs 3, 4 and 6). When transplanted under the kidney capsules of SCID mice, the ES cells further differentiated into mature pancreatic cell lineages, and formed islet-like clusters in vivo (Fig. 8). Although the recovered grafts showed a detectable level of insulin, we did not observe secretion of glucose-responsive insulin (supplementary material Fig. S5). The ability of the endocrine β-cell to acquire responsiveness to glucose is established after birth. It is reported that EphA-ephrin-A-mediated β-cell communication regulates insulin secretion from pancreatic islets (Konstantinova et al., 2007). It is possible that Eph-ephrin-dependent signaling is not active in this system to obtain an optimized glucose response. Therefore, further improvement of the pancreatic maturation in vivo or in vivo is required for the cells to acquire the ability to secrete insulin in a glucose-dependent manner.

Since the rLN511 sBM contains a high level of LN511, we speculated that Lama5 mediates the guiding signal from sBM. The Itgb1-knockdown experiment clearly demonstrated that the guiding signal from sBM was transduced through integrin β1. It is reported that pancreatic β-cells express Itgb1, and vascular endothelial cells in the islet provide the BMs containing LN411 and LN511 (Nikolova et al., 2006). In the adult islet, the interaction of LNs with integrin regulates insulin synthesis. Taken together, these findings indicate that LN-integrin signaling affects not only the adult β-cell function, but also early pancreatic differentiation.

Since it is known that heparan sulfate chains act as a reservoir or modulator for various kinds of growth factors and signaling molecules (Izvolsky et al., 2003; Wu et al., 2004), we also examined the role of HSPGs in the differentiation of Pdx1-expressing cells. HSPGs are divided into two groups: those that exist in the cell surface (such as syndecan and glypican) and those that exist in the BM. In particular, HSPG2 is known to be the main BM component. Hspg2 knockdown and degradation of heparan sulfate chains did not affect the differentiation of ES cells into the DE (Y.H., unpublished data), but synergistically decreased the differentiation into Pdx1-expressing cells (Fig. 7). FGFs, FGF receptors and HSPGs are reported to form a ternary complex, which confers the ligand-receptor specificity of FGF signaling (Rodgers et al., 2008). Moreover, selective binding of VEGFs to HSPGs is shown to control the growth and shaping of vascular trees (Ruhrberg, 2003). FGFs and VEGFs are pancreatic-inducing signals secreted from the notochord, or β-cell-inducing signals secreted from the dorsal aorta, respectively (Lammert et al., 2001; Wells and Melton, 2000; Yoshitomi and Zaret, 2004). Therefore, our results strongly indicate that the heparan sulfate chain participates in mediating signal transduction through growth factors, such as FGF and VEGF. Taken together, these findings suggest that HSPG2 and other HSPGs directly or indirectly guide the differentiation of pancreatic cell lineages (Fig. 9).

In conclusion, we found that Lama5 mediates, in part, the activity of M15 cells in guiding the differentiation of ES cells into endoderm and then Pdx1-expressing cells. Synthesized BM prepared from recombinant rLN-10 cells, could be used as a substitute for M15 cells in guiding ES and iPS cells to undergo differentiation into pancreatic endocrine cells. The differentiation signals from the BM were transduced, in part, through LN-integrin, and in part through growth factor receptors, which function in the presence of HSPG2 or other HSPGs in the BM. This novel procedure using sBM for pancreatic differentiation is useful for the elucidation of the molecular mechanisms determining lineage-specific fate during gut regionalization, without having to take into account the effects of feeder cells. Differentiation of ES cells using sBM represents an attractive approach for generating surrogate cells for regenerative medicine.

The mouse ES cell line, SK7, containing a Pdx1-promoter-driven GFP reporter transgene, was established by culturing blastocysts obtained from transgenic mice homozygous for the Pdx1-GFP gene (Gu et al., 2004; Shiraki et al., 2008b). The SK7 ES cell line was maintained on mouse embryonic fibroblast (MEF) feeder cells in Glasgow minimum essential medium (Invitrogen) supplemented with 1000 U/ml leukemia inhibitory factor (LIF; Chemicon), 15% Knocked-out Serum Replacement (KSR; Gibco), 1% fetal bovine serum (FBS; Hyclone), 100 μM nonessential amino acids (NEAA; Invitrogen), 2 mM L-glutamine (L-Gln; Nacalai tesque, Kyoto, Japan), 1 mM sodium pyruvate (Invitrogen), 50 U/ml penicillin and 50 μg/ml streptomycin (PS; Nacalai tesque) and 100 μM β-mercaptoethanol (β-ME; Sigma). The mouse ES cell line, ING112, containing a Ins1-promoter-driven GFP reporter transgene, was established by culturing the blastocysts obtained from transgenic mice homozygous for the Ins1-GFP gene (Hara et al., 2003). The ING112 ES cell line was maintained in a similar way to the SK7 ES cell line. The mouse iPS cell line, 20D-17 (Okita et al., 2007), was maintained on MEF feeder cells in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with LIF, 15% FBS, NEAA, L-Gln, PS and β-ME.

The mesonephric cell line M15 (Larsson et al., 1995) was kindly provided by Toshiaki Noce (Mitsubishi Kagaku Institute of Life Science, Tokyo, Japan) and Minoo Rassoulzadegan (University of Nice-Sophia Antipolis, Antipolis, France) (Shiraki et al., 2009; Shiraki et al., 2008a; Shiraki et al., 2008b). M15 cells were cultured in DMEM supplemented with 10% FBS, L-Gln and PS and treated with mitomycin C at 10 mg/ml for 2.5 hours before use. Mitomycin-C-treated M15 cells were seeded on gelatin-coated six-well or 24-well plates at a concentration of 8×10⁵ cells or 2×10⁵ cells per well, respectively. OP9 cells (Kodama et al., 1994) and PA6 cells (Kleinman et al., 2003) were kindly provided by Shinichi Nishikawa and Yoshiki Sasai (Center for Development Biology, Riken, Japan), respectively. Recombinant human laminin 511 derived from HEK293 cells (rLN-10) was a kind gift from Masayuki Doi and Karl Tryggvason at the Karolinska Institute, Sweden (Doi et al., 2002).

Lama5 gene silencing was carried out by introducing M15 cells with NS or Lama5 shRNA. Expression Arrest™ non-silencing control shRNA (Open Biosystems, RHS1707) or Lama5 shRNA (Open Biosystems, RMM1766-96742027) retroviral vectors were used. According to the manufacturer's protocol, LinX cells (Open Biosystems) for packaging ectopic retroviruses were plated the day before transfection. After overnight culture, the cells were transfected with plasmid DNA using Arrest-In Transfection Reagent (Open Biosystems) according to the manufacturer's protocol. Cells were incubated in M15 culture medium for 24 hours, and viral supernatants were collected. M15 cells were infected with the viral supernatants supplemented with 4 μg/ml polybrene (Sigma). After incubation for 24 hours, the virus-containing medium was replaced with M15 culture medium. Infected M15 cells were selected using 1.5 μg/ml puromycin. Then infected cells were re-plated and cloned. The reduction of Lama5 was checked by western blot analysis (described below).

In the Itgb1-knockdown assay, Expression Arrest non-silencing control shRNA (Open Biosystems, #RHS4080) or Itgb1 shRNA (Open Biosystems, #RMM3981-97055034) lentiviral vectors were used. For virus preparation, HEK293-FT cells (Invitrogen) were plated the day before transfection. After overnight culture, the cells were transfected with lentiviral vectors and ViraPower™ Lentiviral Packaging Mix (Invitrogen) using FuGENE6 Transfection Reagent (Roche) according to the manufacturer's protocol. Cells were incubated in the differentiation media used on d10-d13 in the sBM substratum method (described below) for 24 hours, and viral supernatants were collected. On d10, SK7 cells growing on sBM were infected with the viral supernatants. After incubation for 24 hours, the virus-containing medium was replaced with differentiation medium. From d13 to d15, infected cells were selected using 1.5 μg/ml puromycin, and then infected cells were analyzed on d15.

Hspg2 gene silencing was carried out by introducing SK7 cells with NS or Hspg2 shRNA. Expression Arrest™ non-silencing control shRNA (Open Biosystems, RHS1707) or Hspg2 shRNA (Open Biosystems, RMM1766-98467532) retroviral vectors were used. Virus preparation and infection were performed in a similar way to the Lama5 knockdown. After puromycin selection, infected cells were re-seeded and cloned.

Reagents were purchased and used at the designated concentrations as follows: recombinant human activin A (R&D Systems; 20 ng/ml), recombinant human bFGF (Peprotech; 50 ng/ml), retinoic acid (RA, Sigma; 1 μM), nicotinamide (NA, Sigma; 10 mM), glucagon-like peptide 1 (GLP-1, Sigma; 10 nM) and heparitinase (Seikagaku Biobusiness; 3.5 mU/ml).

The procedure for sBM preparation is summarized in Fig. 2A (Mochitate, K. Method of preparing BM, method of constructing BM specimen, reconstituted artificial tissue using the BM specimen and process for producing the same. US Patent number 7,399,634). The sBM substrata were prepared on six-well culture inserts with PET porous membrane with a 3 μm diameter pore size (BD, 353091). Fibrillar collagen substratum designated ‘fib’, a stiff matrix of type I collagen gel, was prepared on a porous membrane (Furuyama and Mochitate, 2000). Then the fib substratum was coated with MAST-(GlcNAc)_n, oligo-N-acetylglucosamine (GlcNAc)_n ligands that were covalently bound to styrene-maleic anhydride copolymer (MAST), at a concentration of 10-20 μg/ml in DMEM for at least 1 day. After coating the fib substratum with MAST-GlcNAc, excess MAST-(GlcNAc)_n ligands, temporally adsorbed on collagen fibrils or still existing as free molecules, were rinsed with fresh DMEM for several hours. Then, rLN-10 cells were seeded on the collagen fibrils coated with MAST-(GlcNAc)_n ligands at the dose of 9.6×10⁵ cells per fib substratum, and then cultured in DMEM containing less than 1% FBS and 0.2 mM ascorbate-2-phosphate. During 2 weeks of culture, the secreted rhLN511 molecule as well as endogenous collagen type IV, nidogen-entactin and HSPG of perlecan molecules were integrated to form a lamina densa structure beneath rLN-10 cells with the aid of oligo-GlcNAc ligands (Hosokawa et al., 2007). After culturing, the rLN-10 cells were thoroughly removed with D-PBS⁻ containing 50 mM NH₄OH, 0.1% Triton X-100 and protease inhibitor cocktail (Hosokawa et al., 2007), so that the assembled lamina densa structure of the hLN511 isoform was bared to an intact surface. The de novo synthesized sBM substratum thus prepared in vitro was immersed in the preservation solution, then frozen and stored at −75°C until needed to serve as a novel substratum for ES cell culture.

Differentiation studies of M15 procedures were carried out as described previously (Shiraki et al., 2009; Shiraki et al., 2008a; Shiraki et al., 2008b). Briefly, ES cells were plated at 5000 or 20,000 cells per well in 24-well or six-well plates (Nunc), respectively, that had been previously coated with M15 cells. The cells were cultured in differentiation medium supplemented with activin A (20 ng/ml), bFGF (50 ng/ml), 10% FBS and 4500 mg/l glucose. For long-term culture, as shown in Fig. 4, differentiation medium (described above) was used from d0 to d13. The medium was then changed to low-glucose serum-free medium, which was similar to the differentiation medium used between d13 and d28 in the sBM substratum method (described below).

The sBM substrata stored at −75°C was gently thawed at 4°C overnight prior before use. The procedure of ES cell culture on sBM substratum is summarized in Fig. 2A. From d0 to d10, ES or iPS cells were cultured in high-glucose (4500 mg/l) DMEM supplemented with 10 mg/l insulin, 5.5 mg/l transferrin, 6.7 μg/ml sodium selenite (Insulin-Transferrin-Selenium-G Supplement, ITS, Invitrogen), 2.5 mg/ml ALBUMAX II (Invitrogen), NEAA, L-Gln, PS, β-ME, 20 ng/ml activin A and 50 ng/ml bFGF. From d10 to d13, 1 μM all trans-retinoic acid (RA) was added to the culture medium. After d13, the culture media was switched to low-glucose (1000 mg/l) DMEM (Invitrogen) supplemented with ITS, ALBUMAX II, NEAA, L-Gln, PS, β-ME, 10 mM nicotinamide and 10 nM glucagon-like peptide (GLP1). For heparitinase treatment, sBM was treated with 3.5 mU/ml heparitinase before seeding of the ES cells, and then heparitinase treatment was carried out every day from d10 to d15.

The following antibodies were used for flow cytometry: biotin-conjugated anti-E-cadherin monoclonal antibody (mAb) ECCD2 (Shirayoshi et al., 1986) and phycoerythrin (PE)-conjugated anti-Cxcr4 mAb 2B11 (BD Biosciences Pharmingen). The stained cells were analyzed using a FACS Canto analyzer (BD). Data were recorded with the BD FACS Diva Software program (BD) and analyzed using the Flowjo program (Tree Star).

NS or Lama5 KD M15 cells were homogenized in 400 μl SDS sample buffer (62.5 mM Tris-HCl, 10% glycerol, 2% SDS, pH 6.8). After centrifugation, the supernatants were collected and used as total protein extracts. Total proteins were subjected to SDS-PAGE (5% for Lama5 detection, 12.5% for actin detection) and transferred to polyvinylidene difluoride membranes. The membranes were incubated with rabbit anti-Lama5 antibody (Santa Cruz) at 1:200 dilution or mouse anti-actin antibody (Millipore) at 1:1000 dilution. Horseradish-peroxidase-conjugated anti-mouse or anti-rabbit Ig (Jackson) was used as a secondary antibody at 1:1000 dilution, and chemiluminescent signals were detected with ECL Plus western blotting detection reagents (GE Healthcare).

RNA extraction, reverse-transcription reactions, PCR analyses and real-time PCR analysis were carried out as previously described (Shiraki et al., 2009; Shiraki et al., 2008a; Shiraki et al., 2008b). The primer sequences for each primer set are shown in supplementary material Table S1. The PCR conditions for each cycle were denaturation at 96°C for 30 seconds, annealing at 60°C for 2 seconds, and extension at 72°C for 45 seconds. RT-PCR products were separated by 5% non-denaturing polyacrylamide gel electrophoresis, stained with SYBR Green I (Molecular Probes), and visualized using a Gel Logic 200 Imaging System (Kodak). The real-time PCR conditions were as follows: denaturation at 95°C for 3 seconds and annealing and extension at 60°C for 30 seconds, for up to 40 cycles. Target mRNA levels, expressed as arbitrary units, were determined using a standard curve.

On d28, differentiated cells on the sBM substratum were dissociated with 0.25% trypsin, and the recovered cells were additionally cultured for 1 day in 2-methacryloyloxyethyl phosphorylcholine (MPC)-coated 24-well dishes (Nunc). The concentration of the cells was 5×10⁵ cells/well. The following day, floating cells were collected by centrifugation and suspended in 2.1 mg/ml collagen gel at the concentration of 5×10⁵ cells per 10 μl. The collagen gel was prepared from the Cellmatrix® Type I-A collagen kit (Nitta Gelatin, Japan) according to the manufacturer's protocol. Then, 5×10⁵ cells, suspended in 10 μl gel, were injected under the kidney capsules of C.B-17/Icr-SCID/SCID Jcl mice (CLEA-Japan, Japan), with a syringe (29G, BD). Four weeks after injection, grafted mice were sacrificed. The grafts were recovered from the kidney, and total RNA was extracted and subjected to RT-PCR analysis, or fixed with 4% paraformaldehyde (PFA) and processed for immunohistochemistry

The following antibodies were used for immunohistochemistry: goat anti-amylase (×100, Santa Cruz Biotechnology), guinea pig anti-C-peptide (×1000, Linco), biotin-conjugated Dolichos biflorus agglutinin (DBA) lectin (×100, Sigma), rabbit anti-GFP (×1000, Medical and Biological Laboratories), mouse anti-glucagon (×1000, Sigma), mouse anti-insulin (×1000, Sigma), rabbit-MafA (×1000, Abcam), mouse anti-Nkx6-1 (×1000, Hybridoma bank), rabbit anti-pancreatic polypeptide (×100, Dako) and goat anti-somatostatin (×100, Santa Cruz Biotechnology). Secondary antibodies used for immunohistochemistry were conjugated to Alexa Fluor 488, Alexa Fluor 568 or Alexa Fluor 633 (Molecular Probes). Cells were counterstained with DAPI (Roche).

All fixatives and dyes required for transmission electron microscopy were obtained from TAAB (Reading, UK), except for Quetol resin (Nissin EM, Tokyo, Japan). The rLN-10 cell culture and the bared extracellular matrix (ECM) after cell removal were fixed at 4°C with 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) supplemented with 0.2 M sucrose and 0.1% tannic acid, then post-fixed with 1% osmium tetroxide. The tissues were dehydrated step-wise with graded ethanol, replaced with n-butyl glycidyl ether and embedded in Quetol resin. The resin was processed to ultrathin sections with a Leica Ultrotome Nova. Specimens stained with lead citrate and uranyl acetate were observed with a JEOL JEM-2010 transmission electron microscope. For scanning microscopy, the bared ECM beneath rLN-10 cells was also fixed, dehydrated, and then replaced with t-butyl alcohol. After critical-point drying (with a JEOL JFD-310) and spatter-coating with a gold-palladium mixture (by a JEOL JFC-1600), the surface structure of the ECM was observed with a JEOL JSM-6510LA scanning electron microscope.

The rLN-10 cell culture and the bared ECM after cell removal were fixed with cold 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) containing 0.2 M sucrose. For vertical observation the cell culture was embedded in Technovit resin 8100 (Heraeus Kulzer) and processed to 2 μm transverse sections with a Leica rotary microtome RM2165. The diced sections of the bared ECM were served for horizontal viewing of the surface. The thin or diced sections were incubated with rabbit anti-human laminin (Millipore, AB19012) and anti-human type IV collagen (Progen Biotechnik, 10706) polyclonal antibodies in a humidified chamber. The nuclei of the rLN-10 cells were stained with propidium iodide (PI) (Wako Pure Chemical Industries, Osaka, Japan). FITC and PI fluorescence was observed with a Leica confocal laser-scanning microscope TCS SPE.

We thank members of the Gene Technology Center and the Center for Animal Resources and Development at Kumamoto University for their technical assistance. We also thank K. Ohgomori, S. Yamako and K. Katagiri at the BM Matrix laboratory in NIES for their technical support and M. Nakamura for her SEM operation. This work was supported by a grant (to S.K. and K.M.) from the New Energy and Industrial Technology Development Organization (NEDO), and in part by a global COE grant (Cell Fate Regulation Research and Education Unit) and a Grant-in-Aid (21390280) for the Realization of Regenerative Medicine from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (to S.K.). S.K. is a member of the Global COE Program (Cell Fate Regulation Research and Education Unit), MEXT, Japan.