Author contributions: H.T.-W.: conception and design, collection/assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; J.E.S.: conception and design, financial support, data analysis and interpretation, final approval of manuscript, A. J. E.: conception and design, financial support, collection/assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript.
Differentiation methods often rely exclusively on growth factors to direct mouse embryonic stem cell (ESC) fate, but the niche also contains fibrillar extracellular matrix (ECM) proteins, including fibronectin (FN) and laminin, which could also direct cell fate. Soluble differentiation factors are known to increase ECM expression, yet ECM's ability to direct ESC fate is not well understood. To address the extent to which these proteins regulate differentiation when assembled into a matrix, we examined mouse ESC embryoid bodies (EBs) and found that their ability to maintain pluripotency marker expression was impaired by soluble serum FN. EBs also showed a spatiotemporal correlation between expression of FN and GATA4, a marker of definitive endoderm (DE), and an inverse correlation between FN and Nanog, a pluripotency marker. Maintenance of mouse ESC pluripotency prevented fibrillar matrix production, but induction medium created lineage-specific ECM containing varying amounts of FN and laminin. Mouse ESC-derived matrix was unlike conventional fibroblast-derived matrix, which did not contain laminin. Naïve mouse ESCs plated onto ESC- and fibroblast-derived matrix exhibited composition-specific differentiation. With exogenously added laminin, fibroblast-derived matrix is more similar in composition to mouse ESC-derived matrix and lacks residual growth factors that mouse ESC matrix may contain. Naïve mouse ESCs in DE induction medium exhibited dose-dependent DE differentiation as a function of the amount of exogenous laminin in the matrix in an α3 integrin-dependent mechanism. These data imply that fibrillar FN is necessary for loss of pluripotency and that laminin within a FN matrix improves DE differentiation. Stem Cells2013;31:2084–2094
Embryonic stem cells (ESCs) use a wide variety of external cues to regulate their balance between self-renewal and differentiation. During development, these cues are expressed with precise temporal and spatial control to result in appropriate cellular arrangements . Complex sets of soluble growth factors, intended to recapitulate the early developmental regulation of a specific lineage, have been used in vitro to differentiate ESCs ; for example, addition of Activin A, a TGF-β family protein, and Wnt3a induces initial endoderm marker expression [3, 4]. However, differentiation protocols can also inadvertently include or stimulate the production of ECM proteins. For example, induction of mouse ESCs to definitive endoderm (DE) by Activin A involved changing the substrate on which the cells were plated from a collagen- to a FN-based substrate . Many growth factors have also been linked to extracellular matrix (ECM) upregulation, for example, TGF-β stimulates fibronectin (FN) production , and can sequester and regulate the presentation of soluble cues . Though growth factors are clearly an important differentiation regulator, these observations motivate the examination of whether combinations of ECM proteins, for example, FN and laminin, and the integrin-mediated signaling which they induce play a role in directing ESC fate in general .
FN is expressed and localized to mouse endoderm marker-expressing cells in vivo  as well as many endoderm-derived tissues [10-12]. Laminin is also upregulated during endoderm specification in vivo, where it forms a basement membrane between the primitive ectoderm and endoderm . In fact, β1-integrin, which binds FN, laminin, and other matrix proteins, has been shown to be required for endoderm differentiation in mouse embryoid bodies (EBs) . When secreted, FN and laminin bind to integrin receptors and are assembled into a three-dimensional (3D) fibrillar scaffold alongside other matrix proteins to support cells . Much like the signaling cascades initiated by growth factors, integrin activation via ECM binding initiates intracellular signaling pathways in mouse ESCs (e.g., MAPK/ERK and Rho-ROCK ) that mirror key aspects of embryonic development , including the regulation of pluripotency [17, 18], proliferation , and differentiation [20, 21]. Yet, these studies typically utilize a defined ECM without necessarily defining the endogenous ECM that is produced by differentiating ESCs. This ESC matrix is likely to be as complex as that found in mature tissues  and while many reductionist studies have examined the effects of matrix components on properties of mouse ESCs , examining how a combination of matrix-based cues or ECM proteins influence stem cells has been limited to high throughput two-dimensional assays . Therefore, we used multicellular EBs, ESC-derived matrix, and fibrillar matrix designed with specific protein composition to investigate how the presence of FN and laminin in a 3D matrix regulates mouse ESC differentiation, especially into DE. Together, these data implicate endogenous ECM as important coregulators of mouse ESC fate along with the growth factors that induce their production.
Materials and Methods
All cells were cultured at 37°C in a humidified incubator containing 5% carbon dioxide. Mouse fibroblast (NIH-3T3) cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% bovine calf serum (Thermo Scientific, Rockford, IL, www.thermoscientific.com), 4 mM l-glutamine, 1 mM sodium pyruvate, 100 U/mL penicillin, 100 µg/mL streptomycin, and 0.25 µg/mL amphotericin B for 4 days before passaging. Murine ESCs (cell line CCE, Stem Cell Technologies, Vancouver, Canada, www.stemcell.com, ; cell line R1, American Type Culture Collection, Manassas, VA, www.atcc.org, ) were maintained in DMEM containing 2 mM l-glutamine, 1 mM sodium pyruvate, 50 U/mL penicillin, 50 µg/mL streptomycin, 1 mM non-essential amino acids, 15% fetal bovine serum screened for mouse ESCs (Thermo Scientific), 100 µM 1-thioglycerol, and 103 U/mL leukemia inhibitory factor. ESCs were grown on 0.1% gelatin-coated plates and passaged 1:10 as a single-cell suspension once the plate reached 80% confluence, about every 2 days, to maintain their undifferentiated state. CCE cells that stably express green fluorescent protein (GFP) under the control of the Nanog promoter were kindly provided by Dr. Ihor Lemischka (Mt. Sinai Medical School) and are termed Nanog-GFP ESCs .
To differentiate ESCs as a monolayer, cells were seeded as a single cell suspension on the indicated substrate at a density of 104 cells/cm2 in a differentiation medium composed of 1:1 Iscove's modified DMEM and Ham's F-12 nutrient mixture, 2 mM l-glutamine, 50 U/mL penicillin, 50 µg/mL streptomycin, 0.1% bovine serum albumin, 450 µM 1-thioglycerol, and the indicated concentration of fetal clone II serum (Thermo Scientific). Differentiation medium was supplemented with 10% serum , 10% serum and 20 ng/mL BMP-4 (R&D Systems, Minneapolis, MN, www.rndsystems.com) , or 1% serum, 100 ng/mL Activin A, and 10 ng/mL Wnt3a  to induce neural progenitor (NP), mesoderm, and DE lineage specification, respectively (supporting information Table 1). A final concentration of 10 µg/mL of GoH3 α6-integrin function blocking antibody (Beckman Coulter, Indianapolis, IN, www.beckmancoulter.com), 5 µg/mL of Ralph 3.1 α3-integrin function blocking antibody (Developmental Hybridoma Bank, Iowa City, IA, http://dshb.biology.uiowa.edu), or 50 µg/mL of rabbit IgG control antibody were selectively added to differentiation cultures to disrupt integrin binding. In addition to including blocking antibodies in differentiation medium during culture (which was replenished every 2 days), ESCs in suspension were also incubated in differentiation medium containing the antibody at 37°C for 1 hour before seeding.
To differentiate ESCs in EBs, the multicell aggregates were formed by culturing 105 ESCs/mL in differentiation medium containing 15% serum in non-adherent 6 cm Petri dishes for the indicated time. Soluble heparin at a final concentration of 100 µg/mL or BIIG2 α5-integrin function blocking antibody at a final concentration of 6 µg/mL (Developmental Hybridoma Bank) was selectively added to EB cultures in the same manner as described above to competitively bind FN dimerization sites  or block α5-integrin function, respectively . As noted above, medium during culture was replenished every 2 days to continuously block the interaction. Unless otherwise noted, cell culture products purchased were from Invitrogen (Carlsbad, CA, www.invitrogen.com) and other reagents purchased from Sigma-Aldrich (St. Louis, MO, www.sigma-aldrich.com).
Preparation of Extracellular Matrix and Gelatin Culture Substrates
To prepare gelatin-coated substrates, tissue culture plates were incubated with 0.1% gelatin for 30 minutes. Fibrillar extracellular matrix was prepared from the indicated cell type using an established protocol . For fibroblasts, mouse laminin-111 (Southern Biotech, Birmingham, AL, www.southernbiotech.com) was added exogenously to the culture medium when indicated. Briefly, 104 cells/cm2 were grown in growth medium for fibroblasts or induction medium for mouse ESCs for 7 days. Cells were then washed with phosphate-buffered saline (PBS), wash buffer I (2 mM magnesium chloride, 2 mM EGTA, 100 mM sodium phosphate, pH 9.6), incubated at 37°C for 15 minutes in lysis buffer (8 mM sodium phosphate, 1% NP-40, pH 9.6), and replaced with fresh lysis buffer for an additional 60 minutes before final washes with wash buffer II (10 mM sodium phosphate, 300 mM potassium chloride, pH 7.5), PBS, and sterile water. Matrices were sterilized under ultraviolet radiation for 15 minutes before seeding.
Metabolic FN Labeling
ESCs and EBs were cultured for the indicated periods of time in the absence or presence of serum FN, and subsequently labeled with 100 µCi/mL [35S] methionine (MP Biomedicals; Solon, OH, www.mpbio.com) for 24 hours before collection of the medium and lysis of cells in modified radioimmunoprecipitation (mRIPA) buffer. FN was isolated from culture medium and lysates using gelatin–sepharose binding. Isolated FN was then reduced by addition of 0.1 M dithiothreitol and separated by electrophoresis. Gels were dried, placed on a phosphor storage screen, and bands detected and quantified using a Storm 860 system (GE Healthcare Life Sciences, Pittsburgh, PA, www.gelifesciences.com). Data plotted represent total FN produced in both lysates and media. Though cells were plated at similar densities, to normalize gel loading to account for any proliferation differences, protein concentration was determined from a β-N-acetylglucosaminidase activity assay performed on cell lysates.
A deoxycholate (DOC)-solubility assay was used to separate cellular components from DOC-insoluble cell-derived extracellular matrix for western blot analysis, as described previously . Briefly, cultured cells were washed with PBS and lysed with DOC lysis buffer (2% sodium deoxycholate, 20 mM Tris·Cl, 2 mM EDTA, 2 mM phenylmethanesulfonyl fluoride (PMSF), pH 8.8). The cell lysate was collected and passed through a 27-G needle five times to shear DNA and reduce viscosity. The lysates were micro-centrifuged at 18,400 g for 20 minutes to pellet the DOC-insoluble fraction (DOC-insoluble ECM). The supernatant (cell lysis component) was removed and the DOC-insoluble fraction was washed once with fresh DOC lysis buffer and then resuspended in SDS-solubilization buffer (4% SDS, 20 mM Tris·Cl, 2 mM EDTA, 2 mM PMSF, pH 8.8).
Cells were lysed using mRIPA buffer  or DOC lysis buffer  as indicated. If a DOC-lysis buffer was used, the DOC-insoluble fraction was solubilized in a SDS-solubilization buffer . Samples were separated by electrophoresis under reducing and denaturing conditions, transferred to a nitrocellulose membrane and immunoblotted using ab8245 mouse anti-GAPDH monoclonal antibody (1:104; Abcam, Cambridge, MA, www.abcam.com), ab18976 rabbit anti-Oct4 polyclonal antibody (1:500; Abcam), PAX6 mouse-anti Pax6 monoclonal antibody (1:200; Developmental Studies Hybridoma Bank), ab20680 goat anti-Brachyury polyclonal antibody (1:200; Abcam), sc-9053 rabbit anti-GATA4 polyclonal antibody (1:200; Santa Cruz Biotechnology, Inc., Dallas, TX, www.scbt.com), AF1924 goat anti-SOX17 polyclonal antibody (1:200; R&D Systems), R457 rabbit anti-fibronectin polyclonal antiserum (1:2000; ), ab11575 rabbit anti-laminin-111 polyclonal antibody (1:1000; raised against laminin from EHS basement membrane; Abcam) or ab34710 rabbit anti-collagen type I polyclonal antibody (1:2000; Abcam), and the appropriate horseradish peroxidase-conjugated goat or donkey IgG (1:104). Western blots were developed with ECL substrate (Pierce Biotechnology, Rockford, IL, www.piercenet.com) and the integrated densities of bands within the linear range of the film were analyzed using Image J. GAPDH was used to normalize all data, which was plotted as a fold change from EBs grown with complete serum (Fig. 1) or from ESCs grown in maintenance medium (Fig. 4).
Fluorescence-Activated Cell Sorting Analysis
The indicated ESCs maintained in self-renewing conditions or grown as EBs in either complete fetal clone II serum (Thermo Scientific) or FN-depleted fetal clone II serum (FN removed by gelatin-Sepharose affinity chromatography ) were analyzed by flow cytometry for their expression of GFP under the control of a Nanog promoter. After washing in PBS, EBs were dissociated using a 5-minute treatment of 0.05% trypsin under light agitation. Cells were centrifuged before being resuspended for analysis in a FACScan cytometer (Becton Dickinson, Franklin Lakes, NJ, www.bd.com). Fluorescence was measured at 488 nm and data were gated to measure single cells.
Cell cultures to be immunofluorescently stained were fixed with 3.7% formaldehyde for 20 minutes at room temperature. When staining for intracellular proteins, samples were subsequently permeabilized with 0.5% Triton X-100 for 5 minutes at 37°C. Samples were blocked for 1 hour with 2% ovalbumin at 37°C and stained using R457 rabbit anti-fibronectin polyclonal antiserum (1:500; ), sc-9053 anti-GATA4 polyclonal antibody (1:200; Santa Cruz Biotechnology, Inc.), ab3280 mouse anti-actin polyclonal antibody (1:1000; Abcam), rhodamine-phalloidin (1:500; Invitrogen), ab11575 rabbit anti-laminin-111 polyclonal antibody (1:100; raised against laminin from EHS basement membrane; Abcam), or AF1924 goat anti-SOX17 polyclonal antibody (1:20; R&D Systems) and the appropriate Alexa dye- conjugated goat or donkey IgG (1:500). Cells were additionally labeled with Hoechst (1:2000) stain as indicated. Samples were examined by either a CARV or CARV II confocal microscope (Becton Dickinson) mounted on a Nikon Eclipse TE2000-U microscope with IP Lab software or Nikon Ti-S microscope with Metamorph 7.6 software. Image J software with a custom analysis macro was used to determine line profiles (Fig. 2). Briefly, a line was drawn from the center to edge of the EB at its widest confocal cross-section (Fig. 2A). This line was averaged over all 360° and then averaged across a population of EBs in the same culture condition, resulting in line plots from the center to the edge of EBs (Fig. 2B–2D; supporting information Fig. 1).
Quantitative Polymerase Chain Reaction
RNA was isolated from adherent ESCs or cultured EBs at the indicated time points using Trizol according to the manufacturer's instructions and cDNA was prepared from 2 µg of RNA as described elsewhere . Quantitative polymerase chain reaction was performed (40 cycles, 95°C for 15 seconds followed by 60°C for 1 minute) for differentiation markers using an ABI Prism 7900 HT, the primer sets indicated in supporting information Table 2 as designed via PrimerQuest software (Integrated DNA Technologies, San Diego, CA, www.idtdna.com), and the iQ SYBR Green Supermix. Data were analyzed using SDS 2.3 software (Applied Biosystems, Foster City, CA, www.appliedbiosystems.com)), which calculated expression based on a standard curve generated by a FN plasmid . GAPDH was used to normalize all data, which was plotted as a fold change from undifferentiated mouse ESC control samples. Note that the variance in threshold cycle for GAPDH across all samples from Figures 5 and 6 was <0.03 and 0.08 cycles, respectively.
All statistical analyses were performed using Prism 5 (GraphPad Software, Inc., La Jolla, CA, www.graphpad.com). Differences among three or more groups were assessed by analysis of variance with Tukey's post hoc analysis to identify statistical differences when the p-value is less than 0.05. Unpaired t-tests were used when comparing two groups as indicated. All data are presented as mean ± SE of the mean. Experimental data are shown for experiments performed in triplicate.
Fibronectin Is Necessary for Loss of Pluripotency and Correlates with Endoderm Differentiation in EBs
Embryos lacking FN have early developmental defects  and fail to complete gastrulation , suggesting that FN may be required for initial fate specification. To examine the influence that FN may have on differentiation, mouse CCE ESCs that express GFP under the control of a Nanog promoter  were cultured as multicellular EBs to induce their differentiation while monitoring GFP expression and endogenous FN production. Between days 4 and 8 in culture, a decrease in GFP expression occurred as assessed by flow cytometry (Fig. 1A; left), resulting in a threefold reduction of the GFPHIGH population (Fig. 1B), and thus indicating lower Nanog promoter activity. Metabolic labeling showed that FN production preceded the change in GFP expression, resulting in a threefold increase in FN production by day 8 when normalized to undifferentiated ESCs at day 0 (Fig. 1C). To determine whether changes in FN levels affected Nanog promoter activity, EBs were grown in FN-depleted serum with or without small interfering RNA (siRNA) targeting FN (see supporting information Table 3 for sequence). Unlike complete serum, a roughly equal number of cells remained GFPHIGH as lost fluorescence when cultured in FN-depleted serum (Fig. 1A; center and 1B), and the dramatic increase in FN production was no longer detected by metabolic labeling (Fig. 1C). EBs grown in FN-depleted serum with FN siRNA contained many GFPHIGH cells for the duration of the culture period (Fig. 1A; right and 1B) and exhibited decreased FN production versus untreated cells as a confirmation of the FN siRNA's effect (Fig. 1C).
Though these data indicate a sequential relationship between FN and the loss of self-renewal, they do not show whether the absence of FN would also delay the expression of differentiation markers. To address this, EBs from both CCE and R1 mouse ESCs cultured in complete serum and in FN-depleted serum for 6 days were analyzed by western blotting to assess expression of germ layer markers. When compared with complete serum, CCE and R1 EBs cultured in FN-depleted serum exhibited increased expression of the self-renewal marker Oct4 and reduced expression of the ectoderm marker Pax6, the endoderm markers SOX17 and GATA4, and the mesoderm marker Brachyury, although R1 cells saw a slight but not statistically significant increase in Brachyury expression and no change in Oct4 expression (Fig. 1D).
As FN is localized to endoderm cells in vivo  and is abundant in many endoderm-derived tissues [10-12] the spatial and temporal localization of endoderm-markers and fibronectin in EBs were compared. Nanog-GFP, GATA4, and FN images of the widest confocal cross-sections of CCE EBs (Fig. 2A) grown with medium containing complete serum were taken at days 2 and 6, and subsequent analyses produced line plots of the average staining intensity from EB center to edge. Both GATA4 (Fig. 2B) and FN (Fig. 2D) showed increased staining as a function of culture time. Levels of GATA4 and FN also increased with distance from the EB center at day 6 (arrows) but not at day 2. To quantitatively measure the correlation between protein expression and position for FN and GATA4, Pearson's Product-Moment Coefficient, a measure of positively (+1) or negatively (−1) correlated variables, was calculated. This coefficient increased from 0.11 on day 2 to 0.70 on day 6, showing that as EBs develop, FN and GATA4 exhibit similar distributions. In contrast, Nanog-controlled GFP expression (Fig. 2C) decreased with both time and position towards the EB edge, and thus its correlation coefficient with both GATA4 and FN was negative at day 6: −0.58 and −0.85, respectively. FN and GATA4 expression in EBs treated with heparin, which interferes with FN matrix assembly but not production , did not increase with position towards the EB edge but did increase over time (supporting information Fig. 1C–1E). Similarly, treatment with the antibody BIIG2, which blocks α5-integrin binding to FN , prevented the increase in FN and GATA4 at the EB periphery (supporting information Fig. 1F–1H). It is important to note that mouse ESCs do not express many other fibronectin-binding alpha integrins . It should also be noted that these confocal microscopy data are not likely affected by antibody diffusion as actin antibody and phalloidin staining produce line profiles with a correlation coefficient of 0.78 in EBs made from CCE ESCs (supporting information Fig. 1A, 1B). Together these data suggest that spatial and temporal changes in FN assembly may be coupled with mouse ESC fate.
Endoderm-Inducing Growth Factors Cause Assembly of a Fibrillar ECM with Composition that Is Distinct from Other Lineages
Unlike EBs, monolayer culture can be highly controlled to induce mouse ESC differentiation down a specific lineage. To further establish a link between extracellular matrix proteins that are expressed as a result of the growth factors present in ESC cultures and endoderm induction, mouse ESCs were plated onto gelatin-coated substrates and cultured for 7 days in medium that either maintained pluripotency (PP)  or induced differentiation to mesoderm (MD) , DE , or NPs  (supporting information Table 1). Fibrillar FN matrix was observed in the DE induction conditions, compared with pluripotent mouse ESCs and the other differentiation conditions of CCE cells where FN was mostly punctate (Fig. 3). Quantitative comparison of assembled matrix using a DOC-solubility assay  indicated differences in both composition and amount of DOC-insoluble matrix produced in each condition (Fig. 4A); for example, CCE and R1 ESCs in differentiating conditions produced two- to sixfold more FN matrix than when cultured in PP medium (Fig. 4B). These assembly differences are not likely due to different serum FN concentrations as PP medium has the highest serum concentration (supporting information Table 1). DE medium also induced 30–550% more laminin-111 expression than ESCs in NP or MD media (Fig. 4A, 4B). In contrast, no matrix from either ESC line produced substantial type I collagen (Fig. 4A). These cells, regardless of culture conditions, also produced matrix with different composition from the conventional matrices of 3T3 fibroblasts  which are enriched in fibronectin and also produce some type I collagen (Fig. 4A). These data show that more matrix is assembled in DE medium than in other differentiation media, which is supported by decellularization studies. After 7 days in culture, CCE ESC and 3T3 fibroblast cultures were decellularized, removing cellular but preserving matrix components . Only matrices derived from 3T3 fibroblasts and ESCs in DE medium remained on the substrates following decellularization (Fig. 4C).
Laminin-111 Incorporated into Fibroblast-Derived ECM Dose-Dependently Improves Endoderm Differentiation
To examine whether differentiation depends on or is affected by ECM components in the presence of induction medium, pluripotent ESCs were cultured on decellularized matrix derived from fibroblasts or ESCs induced to express DE markers and grown in either NP or DE inductive medium. Lineage specific transcription factor expression was then measured to assess the population's fate. In all conditions and for both CCE and R1 cells, Oct4 expression decreased (Fig. 5). NP medium was sufficient to induce up to a twofold increase in Pax6 mRNA expression and inhibit expression of mesoderm or endoderm lineage markers after 6 days of culture on ESC-derived matrix (black data, Fig. 5). On the other hand, DE medium inhibited Pax6 and induced mesoderm and endoderm marker expression; the mesoderm marker Brachyury peaked at day 6 coincident with or followed by endoderm markers Foxa2 and GATA4 (green data, Fig. 5). At least during a portion of the time course, Foxa2 and GATA4 were differentially expressed on ESC-derived matrix compared with on fibroblast-derived matrix, suggesting that ECM differences may enhance DE differentiation. No detectable levels of mRNA were measured from decellularized ECMs before reseeding (data not shown), indicating that these results were not affected by residual mRNA transcript.
To compare matrices that differ only in the presence or absence of a single ECM component, fibroblasts were grown in the presence of exogenous laminin-111, which was incorporated into the DOC-insoluble matrix in a dose-dependent manner (Fig. 6A). Naïve CCE and R1 ESCs cultured on decellularized, fibroblast-derived ECM with varying amounts of laminin again showed induction media-dependent expression of Pax6 or Brachyury, Foxa2, and GATA4 in NP or DE induction medium, respectively. More specifically, a laminin dose-dependent increase in Foxa2 and GATA4 was observed after 6 and 10 days for the R1 and CCE cell lines, respectively (Fig. 6B). R1 and CCE cells also showed a twofold laminin-dependent increase in the percentage of cell nuclei expressing the endoderm marker SOX17 after 6 or 10 days, respectively (Fig. 6C; supporting information Fig. 2). The number of population doublings was similar for all conditions (supporting information Fig. 3), indicating that proliferative differences could not account for laminin-111-dependent increase in DE marker expression.
Laminin-111 Enhances Endoderm Differentiation Through α3-Integrin Mediated Signaling
To understand how ECM-incorporated laminin might enhance DE differentiation, we considered three mechanisms: laminin-mediated Activin A sequestration, Activin A-independent integrin signaling, and an Activin A-laminin feed-forward loop. First, decellularized fibroblast-derived ECM, grown with 0–50 µg/mL laminin-111, was exposed to Activin A-containing induction medium, either in presence or absence of CCE cells, to determine if the laminin-containing matrix was regulating the presentation of soluble cues as has been shown with FN . The DOC-insoluble matrix fractions contained Activin A after 1-day only when CCE cells were reseeded onto the decellularized matrices; however, Activin A sequestration in the matrix was laminin-independent (Fig. 7A), which suggests that laminin is not enhancing DE differentiation through ECM-sequestration of Activin A. To assess whether laminin-mediated integrin signaling induced DE differentiation independent of Activin A, R1 cells were cultured on fibroblast-derived ECM with 50 µg/mL laminin-111 but in DE medium that did or did not contain Activin A. ESCs in DE medium on laminin-containing matrix had an almost two orders of magnitude increase in Sox17 positive nuclei versus cells grown on the same substrate but in medium lacking Activin A (Fig. 7B). Cells exposed to Activin A but lacking laminin within their matrix showed approximately twofold less Sox 17 compared with cells plus Activin A and laminin, consistent with previous data (Fig. 6C) and indicating that laminin does not act independently of but rather in concert with Activin A. As Activin A is a TGF-β family protein and stimulates FN production [6, 38], it is possible that Activin A signaling could be sufficient to increase endogenous laminin expression and drive DE marker expression. The matrix of R1 cells grown in the presence of exogenous Activin A contained more laminin-111 and similar amounts of fibronectin as matrix from cells grown without Activin A (Fig. 7C).
While Activin A enhanced matrix incorporation of endogenous laminin-111, its role in inducing DE marker expression could act via non-matrix mediated pathways. To examine whether the effects of Activin A on laminin-111 matrix induce DE marker expression as a result of integrin signaling, R1 cells were cultured in DE induction medium on fibroblast-derived ECM, with or without 50 µg/mL laminin-111, and also with function-blocking antibodies to disrupt common laminin receptors α6-integrin (GoH3 antibody ) or α3-integrin (Ralph 3.1 antibody [40, 41]) . The laminin-111 associated increase in Sox17 positive nuclei was not observed when the α3-integrin blocking antibody was included, either alone or in conjunction with an α6-integrin blocking antibody, despite the presence of Activin A (Fig. 7D). Together these data indicate that laminin-111 enhances DE differentiation through a feed-forward mechanism in which Activin A enhances endogenous laminin matrix production and laminin-mediated α3-integrin signaling enhances the DE induction efficiency of Activin A (Fig. 7E).
This study sought to determine the effects of stem cell derived matrix on mouse ESC differentiation. Fibronectin was found to inhibit pluripotency and correlate with DE differentiation in EBs. In monolayer culture, mouse ESCs exposed to DE-inducing growth factors, that is, Activin A and Wnt3a, assembled a DOC-insoluble matrix that contained more laminin-111 than the matrix derived from other ESC lineages. Laminin-111 was found to enhance Activin A-mediated DE differentiation and appeared to depend on α3β1-integrin signaling.
Our results suggest that conventional growth factor cocktails used to induce NP , mesoderm (MD) , and DE lineages  each may be effective in part due to production of and signaling from the ECM; indeed we observed that each cocktail produced a distinct matrix. We also found that specific components, for example, FN, are required for EBs to lose pluripotency marker expression and acquire lineage specific differentiation markers; in fact, there is a direct temporal correlation between the onset of FN production and the loss of Nanog as well as a spatial correlation between FN and the endoderm marker GATA4. Human ESCs differentiated on a FN-coated substrate have shown improved DE differentiation in comparison with collagen and laminin substrates , which together suggests that the ECM may be an early regulator of cell fate by causing ESCs to exit pluripotency.
Beyond pluripotency, ECM composition has also been suggested as a lineage-specific regulator of differentiation by high throughput screening and tissue-specific matrix peptide coatings on planar substrates [21, 43, 44]. However using these defined ECM compositions is complicated by the matrix produced by ESCs themselves, which depends on the exogenous cues presented, that is, serum proteins and growth factors. We found that maintaining pluripotency inhibited ECM assembly, but as most stem cell methods focus on differentiation, ESC-produced matrix is likely to complicate which differentiation pathways protocols use, for example, matrix- versus growth factor-mediated. That said, ESC-produced matrix could be advantageous as it defines specific components likely to drive lineage-specific differentiation; indeed, we observed that ESCs exposed to Activin A and Wnt3a  produced an ECM that was distinct from other ESC-derived or fibroblast-derived ECM, which contained substantially less or completely lacked laminin, respectively. Moreover, in matrix that originally lacked laminin, for example, fibroblast-derived ECM, ESCs still expressed Foxa2 and GATA4, likely because they produced the laminin that was not already present. On the other hand, substrates where laminin-111 was titrated into the matrix exhibited a dose-dependent expression of these markers. These data suggest that laminin is not only supportive of DE expression in the presence of Activin A and Wnt3a, but it can also augment growth factor signaling. Composition specific matrix has been implicated not just in differentiation but also in patterning; laminin-111 expression within EBs establishes a polarized basement membrane . Progenitor cells also have been found to increase ECM production in response to TGF-β family proteins, including Activin A [6, 38, 46, 47], indicating that growth factor-induced matrix expression is likely conserved, though the cell response is context specific.
Matrix production caused by induction medium can initiate integrin-mediated signaling, which may enhance cell fate changes brought on by growth factors. For example, we found that laminin incorporated into a fibroblast-derived ECM enhanced DE differentiation in the presence of DE induction medium, but that chronic treatment with an α3 integrin function blocking antibody prevented this response. Other laminin receptors, for example, α6 integrin, did not exhibit the same blocking capabilities, though in addition to laminin-111 binding, both are receptors for other ECM proteins including fibronectin, collagen I, and thrombospondin [40, 42]. Integrin specific signaling also appears to extend to fate choices between visceral and DE , suggesting that adhesion complex composition may play a larger role in matrix-mediated signaling than previously thought. We also noted that differentiation was not caused by matrix sequestration of growth factors, and laminin-containing FN matrix in the absence of exogenous Activin A was insufficient to induce DE marker expression, and thus neither Activin A binding nor integrin-signaling alone were responsible for the improved differentiation observed on laminin-containing ECM. These results suggest that Activin A directs DE differentiation and laminin production and that laminin-mediated α3β1 integrin signaling improves DE differentiation efficiency.
These findings that ECM composition augments growth factor signaling fit into a larger body of work showing that stem cell differentiation is affected by ECM properties, including stiffness [48, 49], topography [50, 51] and ligand composition [21, 44]. These findings also demonstrate that defined ECM composition, regardless of whether it was initially presented to ESCs or is the result of subsequent matrix production, should be considered when evaluating differentiation protocols.
This study examined the role of extracellular matrix proteins in the early lineage commitment of mouse embryonic stem cells. Fibronectin was found to be required for loss of pluripotency marker expression and was correlated with endoderm marker expression in embryoid bodies. In monolayer culture, definitive endoderm induction medium caused mouse embryonic stem cells to assemble an extracellular matrix that contained more laminin-111 than other lineages. Laminin-111 was found to dose-dependently enhance Activin A-mediated definitive endoderm differentiation, in part through α3β1-integrin signaling. This demonstrates a role for specific extracellular matrix proteins in mediating signals that guide early embryonic stem cell fate.
The authors thank Dr. Ihor Lemischka for kindly providing the Nanog-GFP ESCs, Dr. Melissa Micou for assistance with statistical analyses, and Dr. Maike Sanders for fruitful discussions. This work was supported by grants from the National Institutes of Health (NIH R01-CA044627 to J.E.S. and NIH R21-EB011727 to A.J.E.). Predoctoral fellowship support was also provided by a National Science Foundation Graduate Research Fellowship (to H.T.-W.).
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.