LIM homeobox 2 promotes interaction between human iPS-derived hepatic progenitors and iPS-derived hepatic stellate-like cells

Human induced pluripotent stem (iPS) cells can differentiate into hepatocyte lineages, although the phenotype of the differentiated cells is immature compared to adult hepatocytes. Improvement of cell-cell interactions between epithelium and mesenchyme is a potential approach to address this phenotype issue. In this study, we developed a model system for improving interactions between human iPS-derived hepatic progenitor cells (iPS-HPCs) and human iPS-derived hepatic stellate cell-like cells (iPS-HSCs). The phenotype of iPS-HSCs, including gene and protein expression profiles and vitamin A storage, resembled that of hepatic stellate cells. Direct co-culture of iPS-HSCs with iPS-HPCs significantly improved hepatocytic maturation in iPS-HPCs, such as their capacity for albumin production. Next, we generated iPS cell lines overexpressing LIM homeobox 2 (LHX2), which suppresses myofibroblastic changes in HSCs in mice. Hepatocytic maturation in iPS-HPCs was significantly increased in direct co-culture with iPS-HSCs overexpressing LHX2, but not in co-culture with a human hepatic stellate cell line (LX-2) overexpressing LHX2. LHX2 regulated the expression of extracellular matrices, such as laminin and collagen, in iPS-HSCs. In conclusion, this study provides an evidence that LHX2 upregulation in iPS-HSCs promotes hepatocytic maturation of iPS-HPCs, and indicates that genetically modified iPS-HSCs will be of value for research into cell-cell interactions.


Differentiation of human iPS cells into mesenchymal cells exhibiting a hepatic stellate cell (iPS-HSC) phenotype.
Hepatic stellate cells develop from mesoderm, STM and mesothelium 16,17 . To obtain iPS-derived mesenchymal cells exhibiting a hepatic stellate cell phenotype (iPS-derived hepatic stellate cell-like cells, iPS-HSCs), we modified previously described protocols for differentiation of human iPS cells into mesoderm or lateral plate mesoderm (LPM) [28][29][30] . The protocol for iPS-HSC differentiation is shown in Fig. 1a. In the "bABC protocol", human iPS cells (RIKEN-2F) were differentiated into mesodermal progenitor cells (iPS-MP) by treatment with basic fibroblast growth factor (bFGF), activin A, bone morphologic protein-4 (BMP-4), and GSK-3β inhibitor (CHIR99021) for 3 days. The cells were then cultured with bFGF and BMP-4 from days 4 to 6. The cultured cells actively proliferated under this differentiation protocol, and exhibited a spindle shape similar to that of mesenchymal cells (Fig. 1a).
After differentiation for 3 days, NCAM + and EpCAM − cells constituted 76% of the iPS-derived cells, and most of these cells had differentiated into mesodermal progenitor cells (iPS-MP cells, Supplementary Fig. 1a) 28 . By comparison, the NCAM + and EpCAM − fraction among differentiated cells was 45% and 19% using the "bB protocol" (bFGF and CHIR99021 for the first 3 days) and "bC-bAB protocol" (bFGF and CHIR99021 for a day, and bFGF, activin A, and BMP-4 on days 2 and 3), respectively ( Supplementary Fig. 1a). These data show the higher efficiency of the bABC protocol for inducing differentiation of human iPS cells into iPS-MP cells compared to the bB and bC-bAB protocols.
An expression analysis of differentiated (iPS-MP) cells was undertaken at days 0 to 3 for Oct3/4 (a pluripotent marker), Brachyury (a marker of primitive streak), and mesoderm posterior basic helix-loop-helix transcription factor 1 (MESP1, a mesoderm marker). On day 3, expression of these genes was significantly decreased compared to peak expression level found at day 0 for Oct3/4, at day 1 for Brachyury, and at day 2 for MESP1 ( Supplementary  Fig. 1b). By contrast, expression of forkhead box F1 (FOXF1, a LPM marker) significantly increased in iPS-MP cells using the bABC differentiation protocol (Supplementary Fig. 1b). These data indicated that the iPS-derived cells had differentiated into the LPM lineage. After differentiation for 6 days, Oct3/4 expression in the iPS-derived cells markedly decreased, and expression of Brachyury and MESP1 was significantly reduced compared to iPS-MP cells (Fig. 1b). FOXF1 expression in iPS-derived cells at day 6 remained significantly higher than at day 0 (Fig. 1b).
Next, the expression of HSC marker genes in iPS-derived cells was assessed. Expression of activated leukocyte cell adhesion molecule (ALCAM, a mesothelial cell marker), nerve growth factor receptor (NGFR, a typical HSC marker), Wilms tumor 1 (WT1, a transcription factor related to HSCs), PPARγ, and LHX2 were significantly increased in iPS-derived cells on day 6 compared to day 0 (Fig. 1c). Hepatocyte growth factor (HGF) and cytoglobin (CYGB) were also up-regulated in iPS-derived cells on day 6 ( Fig. 1c). A flow cytometric analysis demonstrated that most iPS-derived cells on day 6 produced platelet-derived growth factor receptor (PDGFRα, a marker Culture medium was supplemented with the following factors: Days 1-3, 10 ng/ml basic fibroblast growth factor (bFGF), 10 μM CHIR99021, 30 ng/ ml bone morphogenetic protein-4 (BMP-4), and 10 ng/ml activin A; Days 4-6, 100 ng/ml bFGF and 50 mg/ ml BMP-4. Right panel: phase contrast views showing cells at day 0, day 3, and day 6 before passage, and at day 6 after passage. Scale bars: 100 μm. (b) Expression of OCT3/4, Brachyury, mesoderm posterior basic helixloop-helix transcription factor 1 (MESP1), and forkhead box F1 (FOXF1) in human iPS-derived cells at day 0 (iPS cells), day 3, and day 6. The y-axis represents the ratio of expression relative to the means of day 0 (iPS cells). (c) Expression of activated leukocyte cell adhesion molecule (ALCAM), nerve growth factor receptor (NGFR), cytoglobin (CYGB), hepatocyte growth factor (HGF), LIM homeobox 2 (LHX2), Wilms tumor 1 (WT1), and peroxisome proliferator-activated receptor gamma (PPARγ) in human iPS-derived cells, which are related to differentiation of septum transversum mesenchyme (STM) and hepatic stellate cells (HSCs for the mesenchymal cell lineage) and that ~20% of cells produced NGFR; the PDGFRα + or NGFR + fraction among iPS-MP cells was very low (Fig. 1d). An immunostaining analysis showed that iPS-derived cells on day 6 produced desmin (a HSC marker protein, Fig. 1e). Thus, we term the day 6 iPS-derived cells as "iPS-HSCs". To evaluate whether our protocol (Fig. 1a) can induce other iPS cell lines to differentiate into iPS-HSCs, we cultured the iPS cell line PB001 using the bABC protocol. Analyses of gene expression showed that day 6 PB001-derived cells had a similar profile to that of RIKEN-2F iPS-derived cells ( Supplementary Fig. 1c). Thus, our protocol (Fig. 1a) was able to induce human iPS-HSC differentiation in other cell lines. Our data indicated that the differentiated iPS-derived cells are mesodermal lineage cells that exhibit a phenotype consistent with that of HSCs.
Vitamin A storage and myofibroblastic change in iPS-HSCs. Human HSCs display characteristic features of vitamin A storage and myofibroblastic change (HSC-activation). To confirm that human iPS-HSCs have the same phenotype as human HSCs, we evaluated these features in iPS-HSCs.
To investigate the ability of iPS-HSCs to store vitamin A, iPS-MP cells (at day 3) and iPS-HSCs (at day 6) were incubated with retinol and palmitic acid for 4 days (Fig. 2a). Expression of lecithin retinol acyltransferase (LRAT, a molecule related to vitamin A metabolism and known as a marker of quiescence in HSCs) was significantly increased in iPS-HSCs compared with iPS-MP cells (Fig. 2b). Lipid droplets in iPS-HSCs after incubation with retinol and palmitic acid were detected, whereas they were infrequent in iPS-MP cells (Fig. 2c). A flow cytometric analysis showed that ~20% of iPS-HSCs had increased autofluorescence after incubation with retinol and palmitic acid compared to control cells treated with vehicle ( Fig. 2d), indicating that iPS-HSCs can store vitamin A in the cytoplasm.
To investigate whether iPS-HSCs are activated by inflammatory stimulation and are transformed into myofibroblast-like cells, we examined the response of iPS-HPCs to TGF-β stimulation as this is known to activate HSCs (Fig. 2e) 18 . In TGF-β1 treated cells, the expression of lysyl oxidase (LOX), type I collagen alpha 1 chain (COL1A1), and α-smooth muscle actin (ACTA2) in was significantly increased compared to vehicle-treated cells (Fig. 2e). By contrast, the expression of LHX2 was significantly decreased after TGF-β1 stimulation (Fig. 2e). These results demonstrated that the phenotype of iPS-HSCs, with regard to vitamin A storage and myofibroblastic change, resembles that of human HSCs.
iPS-HSCs promote hepatic maturation of iPS-HPCs in co-culture. To investigate whether iPS-HSCs promoted hepatic maturation of iPS-HPCs, we co-cultured iPS-HSCs with iPS-HPCs in transwell or 2-dimensional (2D) direct cultures. In transwell co-culture, expression of α-fetoprotein (AFP) and albumin (ALB) was unchanged in iPS-HPCs co-cultured with either iPS-HSCs or the human hepatic stellate cell line LX-2 compared to cells not in co-cultures ( Supplementary Fig. 2). In 2D direct co-cultures, ALB expression was significantly increased in iPS-HPCs co-cultured with iPS-HSCs or LX-2 cells (Fig. 2f). These data indicated that iPS-HSCs can induce hepatic maturation in iPS-HPCs, and that the cell-cell interaction effect is not due to humoral factors secreted by the iPS-HSCs.

Generation of doxycycline (Dox)-inducible LHX2-overexpressing human iPS cell lines and differentiation into iPS-HSCs.
Next, we examined the ability of the transcription factor LHX2 to enhance the iPS-HSC-induced hepatic maturation of iPS-HPCs. LHX2 expression was increased in iPS-HSCs compared to iPS cells and iPS-MP cells (Fig. 1b); however, expression in iPS-HSCs was lower than in the HSC cell line LX-2 ( Supplementary Fig. 3a). Thus, we generated human iPS cell lines overexpressing LHX2. We constructed a self-contained, tetracycline-inducible expression vector based on the PiggyBac transposon as previously described (Fig. 3a) 6,31 . Activation of gene expression in response to Dox was indirectly monitored by coincident green fluorescent protein (GFP) activation (Fig. 3b). We generated and selected >10 different iPS cell lines overexpressing LHX2 (and termed these iLHX2-iPS cells); in these cells, LHX2 expression in undifferentiated iPS cells was induced by Dox (Fig. 3c,d). Oct3/4 expression was slightly increased in Dox-treated iLHX2-iPS cells (Fig. 3c), consistent with a previous report that LHX2 interferes with an initial step of iPS cell differentiation 32 . Immunoblot analysis confirmed that the production of LHX2 protein was up-regulated in Dox-treated iLHX2 cells (Fig. 3d).
iLHX2-iPS cells were cultured using the protocol for iPS-HSC differentiation (shown in Fig. 1a) without Dox to determine whether they could differentiate into HSC-lineage cells. After differentiation, expression of NGFR, CYGB, HGF was significantly increased compared to iLHX2-iPS cells. This result indicated that iLHX2-iPS cells differentiated into HSC-lineage cells in a similar manner to normal iPS cells using this culture protocol; we termed these differentiated iLHX2 cells as iLHX2-HSCs (Fig. 3e). Next, to determine whether LHX2 overexpression affects iPS-HSC differentiation, iLHX2-iPS cells were differentiated and treated with Dox from day 4 to 6 ( Fig. 3f). Expression of LHX2 in Dox-treated iLHX2-HSCs was ~60 times higher than in vehicle-treated cells (Fig. 3f). Immunoblots showed that the LHX2 protein was increased in Dox-treated iLHX2-HSCs compared with vehicle-treated cells (Fig. 3d). NGFR expression in Dox-treated iLHX2-HSCs was significantly y-axis represents the ratio of expression relative to means of day 0 (iPS cells). (d) Flow cytometric analysis of human iPS-derived cells at day 3 and day 6. The x-and y-axis in the upper panels shows platelet derived growth factor receptor alpha (PDGFRα) and fetal liver kinase 1 (FLK1), respectively. The x-and y-axis in the lower panels shows NGFR and FLK1, respectively. (e) Immunostaining of desmin (red). Nuclei were stained with 4′, 6-diamidine-2′-phenylindole dihydrochloride (DAPI, blue). Scale bars: 50 µm. Based on these data, human iPS-derived cells at day 3 and day 6 were termed iPS-mesodermal progenitor (iPS-MP) cells and iPS-HSCs, respectively. Results represent the mean ± SD of three separate experiments. *P < 0.05. increased compared to vehicle-treated cells, however CYGB expression was decreased by LHX2 overexpression (Fig. 3f). These data indicated that LHX2 overexpression in differentiating iPS cells has no influence on iPS-HSC differentiation.
LHX2 overexpression in iPS-HSCs facilitates hepatic maturation of iPS-HPCs in co-culture. Functional changes in iPS-HSCs due to LHX2 overexpression were analyzed in 2D direct co-cultures of iLHX2-HSCs and iPS-HPCs. After co-culture for 6 days, cells were harvested and sorted using FACS; tdTOMATO-labelled iPS-HPCs and GFP-labelled iLHX2-HSCs were collected (Fig. 4a). An expression analysis showed that AFP expression was significantly increased in iPS-HPCs co-cultured with Dox-treated iLHX2-HSCs compared to other groups (Fig. 4b). Expression and production of ALB was significantly increased in iPS-HPCs co-cultured with vehicle-treated iLHX2-HSCs compared to cells not subjected to co-culture. ALB expression and production was also significantly increased in iPS-HPCs co-cultured with Dox-treated iLHX2-HSCs compared to vehicle-treated iLHX2-HSCs (Fig. 4b,c). These data clearly indicated that LHX2 expression in iPS-HSCs facilitated ALB production in iPS-HPCs in co-culture. Expression of apolipoprotein B (ApoB) was significantly increased in iPS-HPCs co-cultured with Dox-treated iLHX2-HSCs compared to vehicle-treated iLHX2-HSCs (Fig. 4b). CYP7A1 expression in HPCs co-cultured with Dox-treated or vehicle-treated iLHX2-HSCs was increased (Fig. 4b). These data indicated that LHX2 expression in iPS-HSCs facilitated hepatic maturation of iPS-HPCs via direct cell-cell interactions.

LHX2 overexpression in LX-2 cells does not facilitate hepatic maturation of iPS-HPCs in co-culture.
To determine whether LHX2 overexpression in the HSC cell line LX-2 induced hepatic maturation of iPS-HPCs to the same extent as iPS-HSCs, we constructed LX-2 cell lines overexpressing LHX2 (LX-2-LHX2) as described for iPS cells ( Supplementary Fig. 3b). A previous study showed that Lhx2 overexpression in murine HSCs cells does not contribute to maintenance of the quiescent phenotype 33 . Expression analysis showed that expression of ACTA2 was decreased in Dox-treated LX-2-LHX2 cells; however, expression of COL1A1 in these cells was similar to that in vehicle-treated LX-2-LHX2 cells and expression of LOX was slightly increased (Supplementary Fig. 3c). There was no difference in the numbers of desmin-producing cells between vehicle-and Dox-treated LX-2-LHX2 ( Supplementary Fig. 3d). These results indicated that LX-2 cells overexpressing LHX2 did not switch to a quiescent phenotype. Overexpression of LHX2 in LX-2 cells did not change expression of AFP, ALB, CYP7A1, and ApoB in iPS-HPCs in both transwell and 2D direct co-culture ( Supplementary Fig. 3e,f). Down-regulation of LHX2 in LX-2 cells did not change expression of AFP and ALB in iPS-HPCs in 2D direct co-culture ( Supplementary Fig. 3g,h). These data indicated that the change of LHX2 expression in LX-2 cells has no effect on hepatic maturation of iPS-HPCs in direct co-culture, in contrast to that in iPS-HSCs.
Change of ECM expression profile in LHX2-overexpressed iPS-HSCs. The mechanism through which overexpression of LHX2 in iPS-HSCs facilitates hepatic maturation of iPS-HPCs was investigated using a microarray assay. Vehicle-treated or Dox-treated iLHX2-HSCs were co-cultured with iPS-HPCs prior to analysis. A gene ontogeny analysis showed that LHX2 overexpression up-regulated genes related to development of the neural crest that differentiates into HSCs in zebrafish 34 . Genes related to development of neurons and brain were also up-regulated by LHX2 overexpression (Supplementary Fig. 4a) 23,35 . These results confirm that the overexpressed LHX2 in iLHX2-HSCs functioned as a transcription factor against LHX2-target genes. The microarray assay also showed that genes related to cell adhesion and collagen fibril organization were down-regulated by LHX2 overexpression; these down-regulated genes were associated with ECM production ( Supplementary  Fig. 4b).
Expressions of PDGFRα, WT1, NGFR, LRAT, and Neurotrimin (a marker for quiescent HSCs) 36 were significantly up-regulated in Dox-treated iLHX2-HSCs as well as that of LHX2, whereas the expressions of ALCAM and PPARγ were not changed by LHX2 overexpression (Fig. 5a-c). Expression of LOX was down-regulated and that of COL1A1 was up-regulated in Dox-treated iLHX2-HSCs; expression of ACTA2 and Desmin was not significantly changed (Fig. 5d). The numbers of desmin-producing cells in Dox-treated iLHX2-HSCs were equal to those in vehicle-treated cells ( Supplementary Fig. 5a,b). Taken together, these results suggested that LHX2 overexpression in iPS-HSCs partially maintains the quiescence phenotype of HSCs.
The microarray analysis showed that humoral factors and expression of FGF7 and HGF were increased in Dox-treated iLHX2-HSCs compared with vehicle-treated iLHX2-HSCs. A quantitative RT-PCR assay confirmed that expression of FGF7 and HGF was up-regulated ( Supplementary Fig. 5c). The function of such humoral factors in hepatic maturation of iPS-HPCs induced by Dox-treated iLHX2-HSCs was analyzed using iPS-HPCs cultured in media supplemented with HGF, FGF7, and HGF plus FGF7. Expression of AFP and ALB in iPS-HPCs did not change significantly after treatment with the humoral factors ( Supplementary Fig. 5d). Furthermore, expression of AFP and ALB in iPS-HPCs did not change significantly in transwell co-cultures with Dox-treated iLHX2-HSCs (Supplementary Fig. 5e). These results showed that the humoral factors derived from Dox-treated iLHX2-HSCs were insufficient to induce hepatic maturation in iPS-HPCs.
The microarray analysis also demonstrated that expression of collagen and laminin (major components of the liver ECM) was clearly changed in Dox-treated iLHX2-HSCs compared with vehicle-treated iLHX2-HSCs in iPS-HPCs co-cultured with iPS-HSCs or LX-2 cells (a human hepatic stellate cell line). The y-axis represents the ratio of expression relative to the means of iPS-HPCs. Results represent the mean ± SD of three separate experiments. *P < 0.05.  ( Fig. 5e). Expression of laminin-family genes was up-regulated by LHX2 overexpression, whereas expression of collagen-family genes, except for type IV collagen, was down-regulated. A quantitative RT-PCR analysis confirmed that expression of LAMA4, LAMA2, LAMA5, and COL4A5 (type IV collagen) was significantly increased in Dox-treated iLHX2-HSCs compared to controls (Fig. 5f). Expression of COL3A1 and COL5A1 was significantly decreased in Dox-treated iLHX2-HSCs compared to controls (Fig. 5f). In contrast to iLHX2-HSCs, expressions of COL3A1 and COL5A1 were significantly increased in Dox-treated LX-2-LHX2 cells and LAMA5 expression in such cells was significantly decreased relative to controls (Supplementary Fig. 6). These data showed that LHX2 overexpression in iLHX2-HSCs changed the expression profile of the ECM, and suggested that changes in the ECM were responsible for hepatic maturation of iPS-HPCs after co-culture with iLHX2-HSCs. Previous studies showed that culture of iPS-derived cells on laminin-coated dishes supports their differentiation into the hepatic lineage and their maintenance as hepatoblast-like cells 37,38 . Thus, we analyzed expression of AFP and ALB in iPS-HPCs cultured on laminin-511-coated, type I collagen-coated, and fibronectin-coated dishes (the same dishes mentioned above in the 2D direct co-culture experiments). AFP expression in iPS-HPCs cultured on laminin-511-coated dishes was higher than on fibronectin-coated or type I collagen-coated dishes, whereas ALB expression in such cells was similar to that on fibronectin-coated or type I collagen-coated dishes ( Supplementary Fig. 5d). These results indicated that the simple addition of a particular laminin is not enough to induce hepatic maturation in iPS-HPCs, and suggested that changes in the ECM expression profile, including production of various laminin and collagen proteins in LHX2-overexpressing iPS-HSCs, were responsible for the additive effects on hepatic maturation of iPS-HPCs.

Discussion
This study demonstrated that LHX2 overexpression in iPS-HSCs promotes hepatic maturation in iPS-HPCs; this effect is mediated through modification of the expression profile of iPS-HSC-derived ECM and is not influenced by humoral factors (Figs 4 and 5). This is the first evidence that genetic modification of human hepatic non-parenchymal cells can facilitate hepatic maturation in iPS-derived hepatic cells. LHX2 overexpression in human iPS-HSCs significantly enhanced the expression of laminin family genes, such as LAMA4, LAMA2, and LAMA5, and decreased expression of collagen family genes (Fig. 5). A previous study reported that LHX2 overexpression in the LX-2 cell line decreased the expression of ACTA2; however, changes in the ECM expression profile induced by LHX2 were unclear 21 . Our analyses here revealed that the ECM expression profile, including laminin and collagen genes, is regulated by LHX2 expression in human iPS-HSCs. Consistent with our data, laminin-411, which is composed of LAMA4, LAMB1, and LAMC1, plays an essential role in the maturation and proliferation of hepatocytes and iPS-derived hepatic cells 39 . Other studies have shown that culture of iPS-derived cells on laminin-coated dishes, such as on laminin-511 and laminin-111, supports cell differentiation into the hepatic lineage and the maintenance of hepatoblast-like cells 37,38 . Our results here showed that culture on defined laminin-coated dishes induced a lower rate of hepatic maturation in iPS-HPCs compared to co-culture with iPS-HSCs ( Fig. 4 and Supplementary Fig. 5f). This finding supported our conclusion on the significance of the ECM profile, which is regulated by LHX2 expression, of human iPS-HSCs for maturation of human iPS-HPCs.
In this study, iPS-HSCs were differentiated via mesodermal cells from pluripotent stem cells (Fig. 1), in a similar manner as previously reported for iPS-hepatic stellate cell-like cells [13][14][15] . On the other hand, several supplemented factors for induction of HSCs from iPS cells were different from those in previous reports, and we did not purify a defined fraction in differentiated cells using FACS. Vitamin A storage (judged from autofluorescence) was detected in 20% of iPS-HSCs (Fig. 2d). The expression of ACTA2, COL1A1, and LOX in iPS-HSCs was increased by TGF-β stimulation (Fig. 2e). These data showed that the phenotype of the iPS-HSCs used in this study was similar to that of iPS-derived hepatic stellate cell-like cells reported previously 13,14 . Furthermore, iPS-HSCs facilitates the maturation of iPS-HPCs in 2D direct co-culture, but not in transwell co-culture ( Fig. 2 and Supplementary Fig. 2). Consistent with our data, murine integrin α8 + fetal HSCs support hepatic maturation of hepatoblast in direct co-culture, but not in transwell co-cutlure 40 , indicating that a direct cell-cell interaction between HSCs and HPCs is important for hepatic maturation of progenitors. Previous transcriptome analyses showed that iPS-derived hepatic stellate cell-like cells resembled human primary HSCs 14 . Our data and those from other studies suggest that the phenotypes of iPS-HSCs partially overlap those of primary HSCs.
The results obtained here showed that LHX2 overexpression in iPS-HSCs increased the level of expression of genes related to HSC quiescence, such as NGFR, Neurotrimin, and LRAT, whereas expression of genes related to HSC activation (COL1A1, ACTA2, and LOX) was not down-regulated by LHX2 overexpression. This suggests that LHX2 overexpression in iPS-HSCs partly contributes to the maintenance of quiescence. Lhx2-deficient mice develop progressive liver fibrosis during fetal development, indicating that LHX2 expression inhibits myofibroblastic changes in HSCs in mice 21 . However, Lhx2 overexpression in mouse primary HSCs and human LX-2 cell line did not maintain the quiescence of HSCs in vitro 33 . One possibility for the discrepancy between our data and previous studies is that the overexpression approach influenced the final expression levels of LHX2 in host cells. Another possibility is that there are differences in the basal expression level of the LHX2 gene in host cells. LHX2 expression in iPS-HSCs was higher than in iPS cells, but lower than in LX-2 cells (Supplementary Fig. 3a). Thus, LHX2 overexpression in iPS-HSCs may be effective for the maintenance of the quiescence phenotype of these cells.
The genes targeted by the transcription factor activity of LHX2 in primary HSCs remain unclear. LHX2 expression plays an important role in the differentiation of human iPS cells into hematopoietic progenitor cells; however, the mechanism of LHX2 activity during hematopoietic differentiation has not been fully elucidated 27,32 . Co-culture of a bone marrow stroma cell line overexpressing LHX2 with iPS cells enhances the hematopoietic differentiation of the iPS cells via the upregulation of the humoral factor apelin 26 . However, expression of apelin was not increased in Dox-treated iLHX2-HSCs compared to control cells (data not shown), indicating that apelin is not a target molecule of LHX2 in iPS-HSCs. LHX2 regulates the Wnt signaling pathway in neural development and in the maintenance of hair follicle stem cells 24,25,41 . We examined the potential role of the Wnt signaling pathway in iLHX2-HSCs; a TOP/FOP flash reporter assay revealed that LHX2-overexpression did not stimulate Wnt signaling pathway in iPS-HSCs (data not shown). The molecular mechanism regulating the expression of genes related to quiescence and to ECM production in iPS-HSCs by LHX2 expression remains unclear, and further study will be needed to address this question.
Our data demonstrated that genetically modified iPS-HSCs are a useful tool for improvement of hepatic maturation of iPS-derived hepatic cells. Several reports showed that co-culture of iPS-derived hepatic cells, stellate cells, and sinusoidal endothelial cells can recapitulate organ development and the pathological development observed in genetic and non-genetic diseases 7,14,15 . The use of patient-derived iPS-HSCs for the co-culture may offer new insights into the pathophysiology of genetic and non-genetic liver diseases whose molecular mechanism is unknown. Genetically modified iPS-HSCs cells will be helpful for validation analysis of potential mechanisms, and be valuable for further screening of novel molecular targets. Our study into the use of genetically modified iPS-HSCs will contribute to strategies using iPS cells and to future developments of new treatments for liver fibrosis as well as improvement of hepatic maturation of iPS-derived hepatic cells.

Materials and Methods
Materials are listed in the supplementary information. Establishment of human iPS cell lines overexpressing LHX2 (iLHX2-iPS cells). As previously described, we constructed a self-contained, tetracycline-inducible expression vector based on the PiggyBac transposon 6,31,43 . Activation of gene expression in response to Dox can be indirectly monitored by co-incident GFP production. Briefly, we produced a derivative vector expressing human LHX2 using the Gateway cloning technique. The vector, PB-T-LHX2-G-ERN (Fig. 3a), was transfected together with PiggyBac transposase into the RIKEN-2F iPS cell line and was selected using media containing G418 for 3 days to generate pooled iPS cell lines containing genomic transposon integrations (iLHX2-iPS cells). We obtained 11 iLHX2-iPS cell lines and 4 cell lines were analyzed in this study after the evaluation of LHX2 expression levels. The resultant 2F iLHX2-iPS cell lines were differentiated into the HSC-lineage as described below.

Myofibroblast-like change of iPS-HSCs.
Myofibroblast-like changes in iPS-HSCs were investigated using cells passaged on fibronectin-coated 24-well plates at a density of 2.5 × 10 4 cells/cm 2 . The cells were incubated in RPMI 1640 supplemented with 2% B27 and 3 ng/ml recombinant human TGF-β1 for 4 days and then analyzed.
Retinol storage in iPS-HSCs. Vitamin A (retinol) storage in iPS-HSCs was analyzed in cells passaged on fibronectin-coated plates or fibronectin-coated microscope slides. The cells were cultured in RPMI 1640 supplemented with 2% B27, 5 μM retinol, and 100 μM palmitic acid for 4 days. Harvested cells were stained with BODIPY493/503, and were analyzed using an FV10i confocal laser microscope (Olympus, Tokyo, Japan). For flow cytometric analysis, cells were dissociated using 0.05% trypsin/0.5 mM ethylenediaminetetraacetic acid (EDTA). The autofluorescence of harvested cells was analyzed by nUV laser (375 nm) and 440/50 filter using a FACS Aria2 cell sorter (Becton Dickinson, Franklin Lakes, NJ, USA) 44 .
Co-culture of iPS-HPCs and iPS-HSCs. For transwell co-culture of iPS-HPCs and iPS-HSCs or iLHX2-derived HSCs (iLHX2-HSCs), or LX-2 as a control, iPS-HPCs were seeded onto fibronectin-coated culture inserts (Millipore) at a density of 5.0 × 10 4 cells/cm 2 , and iPS-HSCs, iLHX2-HSCs, or LX-2 cells were seeded on fibronectin-coated plates at a density of 5.0 × 10 4 cells/cm 2 . Cells were incubated in HPC medium for 1 day, and incubated in Hepatocyte Culture Medium without EGF (Lonza, Basel, Switzerland) for 6 days and then harvested. The medium was changed every other day.
For 2D direct co-culture, the CytoSelect TM 24-well Cell Co-culture System (Cell Biolabs, San Diego, CA, USA) was used following the manufacturer's protocol with slight modifications. Briefly, iPS-HSCs or iLHX2-HSCs were dissociated using 0.05% trypsin/0.5 mM EDTA, and 2.0 × 10 5 cells were seeded into fibronectin-coated wells. The next day, 1.0 × 10 5 tdTOMATO-labeled iPS-HPCs derived from the PB001 cell line were seeded into the iPS-HSC-seeded wells, and incubated with HPC medium for 1 day. Both cells were cultured in Hepatocyte Culture Medium without EGF for 6 days, and the medium was changed every other day. Co-cultured cells were dissociated using 0.05% trypsin/0.5 mM EDTA and were sorted using FACS. Sorted cells were then analyzed.
Statistics. GraphPad Prism software (GraphPad Software, San Diego, CA, USA) was used to calculate the standard deviation (SD) and statistical significance by Student's two-tailed t test between two groups or by one-way ANOVA followed by Tukey's test between more than three groups; P values < 0.05 were considered statistically significant. In all graphs, bars represent the mean ± SD of three or four separate experiments.
Materials, methods for establishment of human iPS-HPCs, expression analysis using quantitative RT-PCR and microarray, flow cytometric analysis and cell sorting, immunohistochemistry, immunoblot analysis, enzyme-linked immunosorbent assay (ELISA), and knockdown assay of LHX2 using lentiviral overexpression of shRNA are shown in the Supplementary information.