Control of Human Embryonic Stem Cell Colony and Aggregate Size Heterogeneity Influences Differentiation Trajectories


  • Céline Liu Bauwens,

    1. Department of Chemical Engineering and Applied Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    2. Institute of Biomaterials and Biomedical Engineering, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Raheem Peerani,

    1. Department of Chemical Engineering and Applied Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    2. Institute of Biomaterials and Biomedical Engineering, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Sylvia Niebruegge,

    1. Institute of Biomaterials and Biomedical Engineering, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Kimberly A. Woodhouse,

    1. Department of Chemical Engineering and Applied Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Eugenia Kumacheva,

    1. Department of Chemical Engineering and Applied Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    2. Institute of Biomaterials and Biomedical Engineering, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    3. Department of Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Mansoor Husain,

    1. Department of Medicine, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    2. McEwen Centre for Regenerative Medicine, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Peter W. Zandstra Ph.D.

    Corresponding author
    1. Department of Chemical Engineering and Applied Chemistry, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    2. Institute of Biomaterials and Biomedical Engineering, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    3. McEwen Centre for Regenerative Medicine, University Health Network and Heart & Stroke Richard Lewar Centre of Excellence, University of Toronto, Toronto, Ontario, Canada
    • Terence Donnelly CCBR, 160 College Street, Office 1116, Toronto, Ontario, Canada M5S 3G9. Telephone: 416-978-8888; Fax: 416-978-2666
    Search for more papers by this author


To better understand endogenous parameters that influence pluripotent cell differentiation we used human embryonic stem cells (hESCs) as a model system. We demonstrate that differentiation trajectories in aggregate (embryoid body [EB])-induced differentiation, a common approach to mimic some of the spatial and temporal aspects of in vivo development, are affected by three factors: input hESC composition, input hESC colony size, and EB size. Using a microcontact printing approach, size-specified hESC colonies were formed by plating single-cell suspensions onto micropatterned (MP) extracellular matrix islands. Subsequently, size-controlled EBs were formed by transferring entire colonies into suspension culture enabling the independent investigation of colony and aggregate size effects on differentiation induction. Gene and protein expression analysis of MP-hESC populations revealed that the ratio of Gata6 (endoderm-associated marker) to Pax6 (neural-associated marker) expression increased with decreasing colony size. Moreover, upon forming EBs from these MP-hESCs, we observed that differentiation trajectories were affected by both colony and EB size-influenced parameters. In MP-EBs generated from endoderm-biased (high Gata6/Pax6) input hESCs, higher mesoderm and cardiac induction was observed at larger EB sizes. Conversely, neural-biased (low Gata6/Pax6) input hESCs generated MP-EBs that exhibited higher cardiac induction in smaller EBs. Our analysis demonstrates that heterogeneity in hESC colony and aggregate size, typical in most differentiation strategies, produces subsets of appropriate conditions for differentiation into specific cell types. Moreover, our findings suggest that the local microenvironment modulates endogenous parameters that can be used to influence pluripotent cell differentiation trajectories.

Disclosure of potential conflicts of interest is found at the end of this article.


Author contributions: C.L.B.: conception and design, collection and or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; R.P.: conception and design, data analysis and interpretation, final approval of manuscript; S.N.: collection and/or assembly of data, final approval of manuscript; K.A.W.: project conception, financial support, final approval of manuscript; E.K.: conception and design, financial support, final approval of manuscript; M.H.: manuscript writing, final approval of manuscript; P.W.Z.: conception and design, data interpretation, financial support, manuscript writing, final approval of manuscript.

The promise of human pluripotent cells as a renewable source of specialized cells has been limited by progress in the development of robust differentiation protocols. Since the first reports of successful human embryonic stem cell (hESC) derivation and maintenance [1, 2], significant efforts have been made to develop methods to control their differentiation into functional cells and tissues. The generation of induced pluripotent stem cells from human cells [3, [4]–5] further motivates our need to design controlled and reproducible mature cell production strategies. Initiating hESC differentiation either in attached colonies or by forming embryoid bodies (EBs) has demonstrated in vitro lineage potential in each of the three germ layers (endoderm, ectoderm, and mesoderm), including pancreatic β cells, neural cells, cardiomyocytes, and blood cells [1, 6, [7]–8]. Differentiation in EBs recapitulates many aspects of embryonic development [9]. However, although it has been recognized that the developmentally relevant emergence of specialized tissues and their subsequent differentiation to mature functional cell types can be influenced by local inductive cues, few studies have used microenvironmental control to prospectively regulate endogenous parameters that influence pluripotent cell differentiation trajectories.

In the embryo it is recognized that cardiogenesis is directed via the coordination of inductive cues from the anterior primitive endoderm [10, [11], [12], [13]–14] and inhibitory cues originating from neurogenic tissue [10, 15, 16]. Accordingly, methods to drive mesoderm and cardiac induction during hESC differentiation have included coculturing hESCs with the visceral endoderm-like cell line END-2 [17] or the addition of transforming growth factor beta (TGF-β) family proteins (most commonly activin A and bone morphogenetic protein [BMP]-2 or BMP-4) that are known to be secreted by primitive endoderm [18, 19]. These studies aim to exploit observations made in the embryo to direct cardiogenesis through exogenous factor-mediated control. We hypothesize that, in addition to these exogenous factors, endogenous parameters such as local cellularity and organization can impact in vitro hESC differentiation. These endogenous parameters may play a role in tissue development during embryogenesis. In this report we explore some of these parameters and document their effect on hESC differentiation trajectories.

Many parameters can influence hESC differentiation outcome, including the handling and status of the input hESC population, media composition, and the method of inducing differentiation [20]. Human ESC propagation typically requires paracrine and autocrine signals as well as physical cell-cell contact [21, 22]. As a result, single-cell dissociation is usually avoided during hESC passaging [1, 23] and human EB formation [24, 25]. Partial dissociation of hESC colonies results in a wide range of colony and aggregate (EB) sizes. This yields a potential source of variability in subsequent differentiation experiments [26]. Recently, control of human EB size has been reported by a number of groups using either forced aggregation of defined cell numbers [18, 27, 28] or microwells to form three-dimensional hESC aggregates of specified dimensions that can then be transferred to suspension to form monodisperse EBs [29, 30]. EB size may be an important parameter, not only to improve reproducibility of hESC differentiation experiments, but also to regulate endogenously influenced cell type-specific differentiation, as has been recently reported [18]. However, the interaction between controlling EB size and endogenously driven cell type-specific differentiation bias has not been explored.

Here we detail the development of a multistage EB-based differentiation method to control colony and EB size that takes advantage of the microcontact printing technique we have previously described [26]. Using this system we demonstrate that the ratio of endoderm- to neural-associated gene and protein expression can be manipulated as a function of colony size, and that this ratio varies between passages in conventionally propagated hESC cultures. Furthermore, by generating homogeneously sized colonies and EBs we revealed that optimizing this ratio is important for maximizing endogenous mesoderm and cardiac induction during hESC differentiation. These data suggest that improved reproducibility and efficiency of cell type-specific differentiation will require not only controlling colony and EB size, but also new technologies to control and characterize the status of the hESC population prior to initiating differentiation.

Materials and Methods

hESC Maintenance

The H9 and H2B cell lines used in these experiments were obtained from the Israel Institute of Technology; the use of the cells in this project has been approved by the Canadian Stem Cell Oversight Committee. Human ESC (passages 40–50) colonies were maintained on irradiated mouse embryonic fibroblasts (MEFs) in knockout Dulbecco's modified Eagle's medium (ko-DMEM; Invitrogen, Carlsbad, CA, supplemented with 20% knockout serum replacement (Invitrogen) and 4 ng/mL fibroblast growth factor (FGF)-2 (PeproTech, Rocky Hill, NJ, Passaging was performed every 4 days by dissociating the cells into small clumps using Collagenase type IV (2 mg/mL; Invitrogen) and replating on a MEF feeder layer at a subculture ratio of 1:6.

Matrigel Patterning

Microcontact printing was used to pattern Matrigel (BD Biosciences, San Diego, onto tissue culture surfaces. The protocol used (Fig. 1A) to print the patterns has been adapted from a published technique [31]. Briefly, a variety of stamps, made out of poly(dimethylsiloxane) (PDMS) using soft lithography [32], were used for printing different specified pattern geometries. The PDMS stamps were sterilized in 70% ethanol overnight, inked with an aqueous solution of pH 5 1:30 growth factor-reduced Matrigel (GFR-MG) for 1 hour, rinsed with sterile ddH2O, and dried with sterile N2. After rinsing and drying, a monolayer of protein remains on the surface. This layer is transferred to the substrate by placing the stamp in conformal contact with the substrate for more than 10 seconds. The substrates used were tissue culture-treated 60-mm dishes (Falcon; BD Biosciences). This substrate is optimal for both protein adsorption and passivation by 5% wt/vol Pluronic F-127 (mol. wt. = 12,600 Da) (Sigma-Aldrich, St. Louis,, which is used to prevent protein adsorption and cell attachment to unpatterned regions of the substrate [31]. Using this protocol, cells can be seeded on the substrate with little nonspecific binding to the passivated regions and no migration of cells between features.

Seeding hESCs on a GFR-MG-Patterned Surface

GFR-MG was micropatterned at three feature diameters: 200, 400, and 800 μm (referred to as 200MP, 400MP, and 800MP, respectively). Human ESC colonies were dissociated to single cells using TrypLE (Invitrogen). TrypLE is a dissociation enzyme reagent that effectively dissociates hESC colonies/aggregates to single cells with high viability (supplemental online Fig. 1). Single-cell suspensions of hESCs were resuspended in completely defined X-VIVO10 medium (Cambrex, Walkersville, MD, (XV medium) with 2 mM l-glutamine (Invitrogen), 1 × nonessential amino acids (NEAAs) (Invitrogen), and 0.1 M β-mercaptoethanol (Sigma-Aldrich), and supplemented with 40 ng/ml FGF-2 and 0.1 ng/ml TGFβ-1 (R&D Systems), and subsequently plated on the patterns at a concentration of 1.5 × 106 cells/mL. Twelve hours after seeding, the patterned cells were washed three times with XV medium to prevent cell attachment on unpatterned areas (Fig. 1A).

Formation and Differentiation of EB Aggregates from hESC Colonies

To form EBs, 2- or 3-day-old confluent micropatterned (MP)-hESC colonies were scraped off the dishes gently (to preserve intact colonies), and resuspended in hESC differentiation medium containing ko-DMEM, 20% fetal bovine serum (Gibco, Grand Island, NY,, 1 × NEAA, 2 mM L-glutamine, 0.1 M β-mercaptoethanol, and 50 U/mL penicillin/streptomycin (Gibco). Cell numbers in hESC colonies and suspension aggregates were measured by staining cells with a fluorescent DNA binding dye (Cyquant; Invitrogen) and measuring average fluorescence intensity with a fluorescence microplate reader (Spectramax Gemini; Molecular Devices Corp., Union City, CA, Aggregates were transferred to suspension on ultra-low attachment non-tissue culture-treated plates (Corning Inc., Corning, NY, for 4 days and then plated onto 0.5% gelatin-coated tissue culture-treated dishes to induce cardiac differentiation in EB outgrowths. Cultures were carried out for 16 days following EB formation. EBs formed using published EB formation protocols served as controls [24].

Quantitative Analysis of EB Size Distribution

The number of aggregates formed was determined by manual counting on a microscope 1 day after transferring colonies to suspension culture. To measure aggregate size distribution under the different aggregate formation conditions, image analysis was performed with Image-Pro1 Plus software (Media Cybernetics, Crofton, MD, The number of cells per EB was determined by dissociating a known number of EBs to single cells and counting cells.

Flow Cytometry

Pluripotency of the starting (input) hESC population was evaluated by flow cytometric analysis of Oct4 protein expression. Human ESC colonies were enzymatically dissociated to single cells and fixed and permeabilized with the IntraPrep fixation and permeabilization kit (Immunotech, Luminy, France, Fixed and permeabilized cells were incubated at a 1:100 dilution of Oct-4 antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA, for 20 minutes at room temperature, followed by an incubation with secondary fluorescein isothiocyanate-conjugated IgG antibody (Molecular Probes Inc., Eugene, OR, at a 1:100 dilution for 20 minutes at room temperature. Viability was assessed using 7-aminoactinomycin D (Invitrogen). Cells were analyzed with a flow cytometer (XL; Beckman Coulter, Fullerton, CA,, using EXpoADCXL4 software (Beckman Coulter). Positive staining was defined as the emission of a level of fluorescence that exceeded levels obtained by 99.5% of cells from the control population stained with only the secondary antibody.

Quantitative Reverse-Transcription–Polymerase Chain Reaction Analysis

Cells were digested in Trizol reagent (Invitrogen), followed by chloroform extraction and precipitation with isopropyl alcohol. The RNA was then purified using RNAEasy columns (Qiagen, Hilden, Germany, with an on-column DNaseI digestion step. cDNA was generated from purified RNA using Superscript-III reverse transcriptase (Invitrogen) as per the manufacturer's instructions. cDNA (10 ng) was used per polymerase chain reaction (PCR) reaction using iQ-SYBR-green master mix (Bio-Rad, Hercules, CA, in triplicate. Relative expression was determined by δ-delta Ct method with the expression of β-actin as internal housekeeping reference (expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a second internal housekeeping reference). Total RNA from human fetal heart, human fetal liver, and human fetal brain were used as positive controls (Stratagene, La Jolla, CA, Primer sequences, derived from the MGH PrimerBank [33], were as follows: Oct4, (5′ primer) CTTGAATCCCGAATGGAAAGGG and (3′ primer) CCTTCCCAAATAGAACCCCCA; Gata6, (5′ primer) AGGGCTCGGTGAGTCCAAT and (3′ primer) CGCTGCTGGTGAATAAAAAGGA; Pax6, (5′ primer) AAGAGCAACGTCACCAGTTTC and (3′ primer) GGAGCCCGGTTGATACCAG; Brachyury, (5′ primer) TGCTTCCCTGAGACCCAGTT and (3′ primer) GATCACTTCTTTCCTTTGCATCAAG; Mixl1, (5′ primer) CCGAGTCCAGGATCCAGGTA and (3′ primer) CTCTGACGCCGAGACTTGG; α-actinin, (5′ primer) GGGTCCGTTTGCCAGTCAG and (3′ primer) GGCTTTCCTTAGGTGGGAGTT; Nanog, (5′ primer) CAAAGGCAAACAACCCACTT and (3′ primer) TCTGCTGGAGGCTGAGGTAT; CXCR4, (5′ primer) CACCGCATCTGGAGAACCA and (3′ primer) GCCCATTTCCTCGGTGTAGTT; Sox3, (5′ primer) GCCGACTGGAAACTGCTGA and (3′ primer) CGTAGCGGTGCATCTGAGG; TroponinT, (5′ primer) GGACGAAGACGAGCAGGAG and (3′ primer) CTTCCGGTGGATGTCATCAAA; β-actin, (5′ primer) CATGTACGTTGCTATCCAGGC and (3′ primer) CTCCTTAATGTCACGCACGAT; and GAPDH, (5′ primer) CTCCACGACGTACTCAGCG and (3′ primer) TGTTGCCATCAATGACCCCTT. Specificity of amplification was assured with analysis of dissociation curves of all reactions. For all samples a control was performed without reverse transcriptase and no amplification was detected in these controls.

Immunostaining Analysis

Fluorescent images of α-actinin (EA-53, 1:800; Sigma-Aldrich), connexin-43 (1:25; Molecular Probes), Gata6 (1:20; R&D Systems Inc., Minneapolis,, and Oct4 (1:200; Santa Cruz Biotechnology) protein expression were obtained and quantitatively analyzed using the Compartmental Analysis or Target Activation assay algorithms available with the Cellomics Arrayscan VTI platform (Thermo Fisher Scientific, Pittsburgh,


To evaluate differences in gene expression level between conditions, statistics were manually computed using the Wilcoxon signed rank test for a paired difference experiment at a significance level of p < .5. To evaluate differences in EB size between conditions, statistics were computed in Origin 7.5 (OriginLab, Northampton, MA, using the paired t test at a significance level of p < .5. Error bars on plots represent the standard deviation of the mean of three or more replicates (n > 3).


An Approach for Generating Size-Controlled hESC Colonies and EBs

Typically, EB-based hESC differentiation involves partial enzymatic dissociation of hESC colonies resulting in a suspension of cell clumps of various sizes (Fig. 1B). We have developed a process, illustrated in Figure 1A, for generating uniform EBs of controlled size based on micropatterning hESC colonies at defined diameters. Briefly, hESC colonies are dissociated to single cells and replated at high density onto micropatterned Matrigel islands. Upon confluence, MP-hESC colonies are transferred to suspension culture by detaching them intact with a cell scraper and resuspending the colonies in hESC differentiation medium. Figure 1C depicts the overall approach, outlining three aspects of this system that could affect endogenous control of differentiation trajectory: (a) status of the input hESC population; (b) MP-hESC colony size; and (c) MP-EB size.

Figure Figure 1..

Overview of conventional and micropatterned (MP) human embryonic stem cell (hESC) colony and embryoid body (EB) formation. (A): Outline of the process for micropatterning Matrigel (MG) on a cell culture substrate, and subsequently seeding cells onto MG-patterned substrate. (B): In conventional EB formation, hESC colonies (input hESC) are enzymatically dissociated to cell clumps and transferred to ultra-low attachment tissue culture plates in hESC differentiation medium. (C): To generate size-controlled colonies, hESCs are dissociated to single cells and plated at high density (1 × 106 cells/mL) onto patterned Matrigel islands and cultured to confluence (2–3 days) in serum-free maintenance medium. To form MP-EBs, intact colonies are detached using a cell scraper and transferred to suspension in differentiation medium. Scale bar = 250 μm. Abbreviation: PDMS, poly(dimethylsiloxane).

Even under maintenance conditions, hESC cultures do not consist of a homogeneous pluripotent population but contain hESC-derived differentiated cells [34, 35]. Therefore the composition of the input hESC population varies between passages and could impact cell fate trajectory during differentiation. Colony size is also an important parameter in determining self-renewal and differentiation fate, as hESC maintenance relies on paracrine signaling from neighboring undifferentiated cells to antagonize prodifferentiation signals [26]. Finally, EB size could affect differentiation as control of this parameter may modulate spatial signaling within the aggregate.

The Influence of Colony Size on Differentiation Trajectory in MP-hESC Colonies

To evaluate the effect of colony size on differentiation trajectory, hESC colonies were patterned at three diameters (Fig. 2A): 200, 400, and 800 μm (200MP, 400MP, and 800MP, respectively). In addition, nonpatterned (non-MP) conventional hESC colonies were cultured in parallel for comparison. A flow cytometry analysis of six separate input hESC populations demonstrates that in each run the majority of cells express the pluripotency protein Oct4 with expression frequencies ranging from 57% to 78.4% in six independent input populations (Fig. 2B). To determine the effect of colony size on differentiation trajectory, quantitative reverse transcription (qRT)PCR was performed to compare gene expression levels between input hESC populations and their corresponding day-2 or -3 colony size-controlled MP-hESC derivatives. Oct4, Gata6, and Pax6 gene expression levels were tested as a surrogate measure for undifferentiated cells, endodermal, and neural differentiation/commitment, respectively. Importantly, these trends were reproduced with a second gene set (Nanog for undifferentiated cells, CXCR4 for endoderm, and Sox3 for neural [supplemental online Fig. 2]). As demonstrated in the three representative runs shown in Figure 2C, variability in lineage-specific gene expression in the input-hESC population led to variability in gene expression in the MP-hESC populations. Despite variations between runs, similar trends were observed in each run in terms of the effect of colony size on gene expression level (Table 1 presents statistical validation of these trends). Non-MP hESCs expressed similar levels of expression for all three genes as the input hESC populations, thus demonstrating that the process of passaging onto Matrigel in serum-free medium itself does not skew the population. Non-MP and 200MP colonies maintained the Oct4 gene expression level of their input hESCs, whereas a significant increase in Oct4 gene expression levels was observed in the larger colonies (400MP and 800MP). It should be noted that our gene expression analysis detects the Oct4A and Oct4B isoforms. It has recently been reported that expression of the Oct4B isoform is not exclusive to undifferentiated embryonic stem cells and is predominantly expressed in the cytoplasm [36]. Using immunofluorescence, we have demonstrated that Oct4 protein expression in the input hESC population is localized in the nucleus (supplemental online Fig. 3), suggesting that we are measuring Oct4A isoform expression. Compared with input hESCs, the Gata6 gene expression level was maintained in non-MP hESCs. In MP-hESCs, expression of this endoderm marker rose significantly in small colonies (200MP), with gene expression level decreasing as colony size increased. Conversely, Pax6 gene expression levels suggest that markers of neural differentiation significantly increase in larger colonies (400MP and 800MP) and decrease in smaller colonies (200MP) (Table 1). We followed the gene expression analysis with an examination of protein expression of these markers in our MP-hESC colonies. The frequency of Oct4 protein-expressing cells was similar at each colony size (Fig. 2E), ranging between 80%–85%, despite increasing Oct4 gene expression with increasing colony size. Conversely, Gata6 protein expression frequency, consistent with the Gata6 gene expression data, was significantly higher in the small 200MP colony condition (55%–60%). Pax6 protein expression was not detectable at this stage of differentiation. Interestingly, whereas Gata6 and Oct4 coexpression was observed in the small 200MP colonies (Fig. 2Di), in the larger 400MP and 800MP hESC colonies coexpression was generally not observed (Fig. 2Dii, 2Diii). These data suggest that controlling colony size can impact changes in gene expression in cells that still express pluripotency markers to skew the probability of the cells taking one differentiation trajectory over another.

Figure Figure 2..

Analysis of size controlled micropatterned human embryonic stem cell (MP-hESC) colonies reveals the influence of colony size on differentiation trajectory. (A): Bright-field images, at high (top panels) and low (bottom panels) magnification, of nonpatterned (non-MP) hESC colonies and colonies patterned at 200-, 400-, and 800-μm diameters (200MP, 400MP, and 800MP, respectively). Scale bars = 250 μm. (B): Representative flow cytometry results of input hESC Oct4 protein expression from six independent runs. Blue: Oct4-stained population. Percentages indicate frequency of positive-staining cells compared with secondary antibody control (red). (C): Representative quantitative reverse transcription-polymerase chain reaction analysis for three of seven trials for expression of Oct4, Gata6, and Pax6 (collective results of all trials found in Table 1). (D): Immunofluorescence images displaying Oct4 (red) and Gata6 (green) protein expression in 200MP (i), 400MP (ii) and 800MP (iii) colonies. Hoechst was used as a nuclear stain. Scale bars = 175 μm. (E): Gata6 and Oct4 protein expression frequencies in 200MP, 400MP, and 800MP hESC colonies. Abbreviation: Exp., experiment.

Table Table 1.. Gene expression level ranges of prepatterned (Input human embryonic stem cells [hESCs]) and postpatterned (nonmicropatterned [Non-MP] and MP-hESC) hESC cultures (n = 7)
original image

Next, we used an analysis of the ratio of endoderm-to-neural gene expression levels (Gata6/Pax6) in the input hESC and output MP-hESC populations to investigate relationships between colony size and differentiation trajectory with respect to the input hESC composition (Fig. 3A). In runs initiated with intermediate input Gata6/Pax6 ratios between 0.35 and 0.55 (i.e., a “balanced” population), it was observed that the output MP-hESC ratio of endoderm (Gata6) to neural (Pax6) gene expression was always higher than the input hESC ratio, and that Gata6/Pax6 decreased with increasing colony size. Input hESC populations with Gata6/Pax6 ratios that were at high (>0.55) and low (<0.35) extremes resulted in almost no change in Gata6/Pax6 ratio in the non-MP, 400MP, and 800MP conditions, possibly indicating that these ratios represent cell populations that are not easily manipulated. The smallest colony condition (200MP) displayed an increase in the Gata6/Pax6 ratio after MP regardless of the input hESC ratio. The increase in endoderm gene expression after MP in the 200MP condition may be attributed to the rapid differentiation of hESCs occurring at low localized cell densities (localized cell density refers to the number of cells per unit radius; in this case the number of cells per 500-μm radius). Not surprisingly, non-MP colonies showed no changes in ratios after a conventional passage was performed in parallel with the patterned conditions. As illustrated in Figure 3B, these results reveal two important parameters that affect differentiation trajectory in this system, input hESC composition and MP-hESC colony size. Schematically, input hESCs typically display a range of Gata6 and Pax6 expressions levels (Fig. 3Bi), which subsequently shift upon micropatterning according to colony size. In these studies it was observed that larger colonies support enrichment of neural-fated cells, whereas smaller colonies support the enrichment of endoderm-fated cells (Fig. 3Bii).

Figure Figure 3..

Gata6 to Pax6 gene expression ratios in micropatterned human embryonic stem cell (MP-hESC) cultures are influenced by input hESC composition and colony size. (A): The effect of (output) MP-hESC colony size on the gene expression ratio Gata6/Pax6 with respect to the corresponding input hESC Gata6/Pax6. (B): Schematics illustrating (Bi) variations in distribution of Gata6/Pax6 gene expression ratios between each input hESC population (I), and (Bii) the relationship between the Gata6/Pax6 gene expression ratio of any given input hESC population (I, dotted line) and its corresponding MP-hESC Gata6/Pax6 ratio at each colony size (200-, 400-, and 800-μm diameters; referred to as 200MP, 400MP, and 800MP, respectively; solid lines). Abbreviations: d0, day 0; EB, embryoid body.

The Influence of Colony Size on Differentiation Trajectories in hESC-Derived EBs

To evaluate our ability to control EB size using our patterning method, image analysis was performed on light microscopy images of day-1 EBs (Fig. 4A) to calculate size distributions based on EB diameter. Narrower size distributions were observed in MP-EB cultures compared with non-MP EB cultures. Furthermore, significantly different EB size distributions were observed between EBs cultures generated from each hESC colony size examined (200MP, 400MP, and 800MP) (Fig. 4Bi). In addition to EB diameter, colony and aggregate size control was also evaluated using a fluorescent DNA method (Cyquant; Invitrogen) to measure cell number (Fig. 4Bii). This analysis demonstrated that the number of cells in each colony is comparable to the number of cells in each aggregate upon EB formation. Furthermore, based on cell number, significant differences in EB size were observed between 200MP and 800MP, and 400MP and 800MP (Fig. 4Bii).

Figure Figure 4..

Formation of size-controlled micropatterned embryoid bodies (MP-EBs). (A): Bright-field images of EBs generated under the four conditions tested: nonpatterned colonies (non-MP) and colonies patterned at 200-, 400-, and 800-μm diameters (200MP, 400MP, and 800MP, respectively). Scale bars = 250 μm. (B): Quantitative demonstration of EB size control in EBs generated from size-controlled human embryonic stem cell colonies (Bi) by EB diameter and (Bii) by comparing cell number in colonies to cell number in subsequently generated aggregates using fluorescent DNA staining.

Within the EB, differentiation is guided by spatial cues and interactions between various cell types. Accordingly, it is probable that the specific composition (or gene expression status) of the hESC population used to initiate EBs is an influential factor in determining cell fate during EB development. Given that mesoderm commitment and cardiogenesis in the embryo are thought to be regulated through coordinated inductive and inhibitory signals from neighboring endoderm [12, 14, 37] and neural tissue, respectively [38, 39], and that the ratio of neural- and endoderm-associated gene expression in hESC cultures can be manipulated by varying colony size, we next examined whether this ratio influenced mesoderm and cardiac induction in size-controlled EBs. Using qRTPCR, mesoderm (Brachyury [Bry] and Mixl1)- and cardiac (α-Actinin)-associated gene expression levels were measured in EB outgrowths 8 days (day 8) after EB formation. Brachyury is transiently expressed in early mesoderm [40], and has been shown to be expressed at elevated levels between day 4 and day 10 of EB differentiation [41, 42]. Mixl1 is expressed in the primitive streak [43, 44], and in hESC-derived EBs from day 3 to day 10 [45]. α-Actinin is a cardiac contractile protein gene that has also been detected in differentiating hESC-derived EBs by day 12 [46, 47]. Plotting gene expression levels of day-8 EB outgrowths with respect to the neural and endoderm composition of their starting MP-hESC populations (Fig. 5A) revealed similar expression profiles for both mesoderm and cardiac markers. It was observed that gene expression levels of all mesoderm and cardiac markers examined were maximized in 400MP and 800MP EBs generated from MP-hESC populations with Gata6/Pax6 expression level ratios above 0.1 (Fig. 5A), whereas gene expression was virtually undetectable in EBs generated from neural-enriched MP-hESCs (Gata6/Pax6, <0.01; mostly 800MP) and endoderm-enriched MP-hESCs (Gata6/Pax6, >1; 200MP colonies). Non-MP cultures gave rise to starting population compositions across the range observed for cardiac gene induction in MP-hESC cultures (0.1–1); however, this resulted in notably lower gene expression levels for all mesoderm and cardiac differentiation markers, likely due to heterogeneity in responsiveness. Given that non-MP EBs were generated from non-size-controlled hESC colonies (Fig. 2A) the measured Gata6 and Pax6 gene expression levels for non-MP hESCs represent an average of different colony (Fig. 4B) sizes with various ratios of endoderm-fated and neural-fated cells. Although the average Gata6/Pax6 ratio for the non-MP condition falls within the optimum range for cardiac induction, it includes high ratios for small Gata6-expressing colonies and low ratios for large Pax6-expressing colonies, which were observed to result in poor cardiac induction (Fig. 5B).

Figure Figure 5..

The interactive effect of micropatterned human embryonic stem cell (MP-hESC) colony size and cell composition on mesoderm and cardiac induction in MP-embryoid bodies (EBs). (A): Gene expression levels measured for mesoderm markers Brachyury (Bry) and Mixl1, and cardiac marker α-Actinin in day-8 (d8) EBs plotted with respect to the Gata6/Pax6 gene expression ratio in the corresponding MP-hESC starting population, and as a function of MP-hESC colony size. Neural-enriched and endoderm-enriched starting MP-hESCs generate MP-EBs that display poor mesoderm and cardiac induction, as indicated by the marked regions on the plots. (B): A schematic summarizing the observed effect of controlling colony size on MP-hESC composition, and the consequence of MP-hESC composition on cardiac induction. Abbreviations: 200MP, 200-μm diameter; 400MP, 400-μm diameter; 800MP, 800-μm diameter.

Mature Cell Differentiation in MP-EBs and the Effect of Input hESC Composition and Colony/EB Size on Cardiac Induction

Starting at day 8 of EB culture, contracting areas appeared in MP-EB outgrowths, suggesting that these cells exhibit spontaneous electrical activity. Optical mapping with di-4-ANEPPS demonstrated that changes in membrane potential corresponded visually with contractions (data not shown). Immunostaining revealed positive expression of contractile protein markers α-sarcomeric actin (α-SR-1) and α-actinin (EA-53), as well as the gap junction protein connexin-43 (Cx-43) (Fig. 6A). Staining day-20 EB outgrowths with α-SR-1 demonstrated a cluster of cardiac cells, whereas EA-53 expression in day-20 replated single cells exhibited the clear cardiac morphology of striated, highly organized fibers. Immunofluorescence analysis of Cx-43 expression on day 22 from dissociated MP-EBs that had been replated at high density displayed a web-like pattern of staining highly concentrated between cell membranes, indicating functional cell coupling.

Figure Figure 6..

Micropatterned embryoid bodies (MP-EBs) can be used to form mature cell types and the output is dependent on colony and EB size. (A): Immunostaining on a replated monolayer demonstrated expression of (Ai) alpha sarcomeric actin in day-21 (d21) EB outgrowths, (Aii) α-actinin at day 20, and (Aiii) gap junction protein connexin-43 at day 22. (B): Representative α-actinin immunofluorescence images from Cellomics analysis demonstrating (Bi) the algorithm used to detect positive α-actinin and (Bii) the corresponding overlay image. (C): A comparison of α-actinin expression frequency at two EB sizes (400-μm diameter [400MP] and 800-μm diameter [800MP]) from independent runs initiated with either a neural-enriched (Gata6 low/Pax6 high) or an endoderm-enriched (Gata6 high/Pax6 low) MP-human embryonic stem cell (hESC) population. (D): Tracking neural rosettes and spontaneously contracting areas in day-8 EBs (three representative trials). (E): The observed effect of differences in endoderm (E) and neural cell (N) composition in the input hESC (ES) population that are further modulated in MP-hESCs by controlling colony size, and ultimately impact MP-EB-based cardiac induction. Abbreviation: 200MP, 200-μm diameter.

To quantitatively evaluate mature differentiation from MP-EB-induced differentiation cultures, day-16 MP-EB outgrowths were dissociated to single cells and cultured for 2 days as a monolayer on fibronectin-gelatin-coated plates and then immunostained with EA-53 to detect cardiac differentiation. Quantitative immunofluorescence was used to image the cells and detect positive protein expression, revealing the presence of EA-53+ cardiomyocytes (Fig. 6B). The effect of MP-EB size and MP-hESC composition (level of Gata6 and Pax6 gene expression) on cardiomyocyte frequency was analyzed. In runs initiated from neural-enriched MP-hESC populations (high Pax6/low Gata6 gene expression), higher frequencies of EA-53 expression were observed in smaller MP-EBs (Fig. 6Ci). Conversely, in runs initiated from endoderm-enriched MP-hESC populations (low Pax6/high Gata6 gene expression), the frequency of EA-53 expression was higher in larger MP-EBs (Fig. 6Cii). In addition, tracking the appearance of beating areas and neural rosettes in the MP-EB cultures (Fig. 6D), it was observed that within individual runs, neural rosettes were always present in larger EB sizes than those in which beating areas were detected. But between runs, the specific MP-EB size that gave rise to the appearance of neural rosettes or beating areas shifted. A proposed mechanism for our observations is summarized in Figure 6E, wherein, mirroring what happens upon manipulating hESC colony size, the endoderm composition is higher in smaller MP-EBs and is lower in larger MP-EBs, however the specific MP-EB size that optimizes cardiac induction shifts based on the cell composition of the input hESC population.


EB-based differentiation is a powerful system because it can mimic in vivo embryonic development and provides a platform to study the effects of cell-cell interactions and spatial organization on cell type-specific commitment. During embryogenesis, factors secreted by anterior lateral endoderm are thought to direct cardiac differentiation of neighboring mesoderm, as well as to promote cell survival and proliferation [12, 14, 37, 48], whereas parietal endoderm [49] and visceral endoderm [17, 50, 51] have been shown to enhance embryonic stem cell differentiation to cardiomyocytes. Given that one of the first events in EB development is the formation of an endoderm layer surrounding a pluripotent ectoderm core [9, 52], we hypothesized that controlling EB size would modulate the ratio of cardiac-inducing visceral endoderm cells to pluripotent hESCs within the EB proportionally to the surface area (endoderm) to volume (hESC) ratio of the EB. In other words, we expected that the endoderm/hESC ratio would decrease with increasing EB size. Accordingly, there would be an ideal EB size corresponding to the ideal ratio of endoderm cells to hESCs to optimize cardiac induction.

A number of groups are targeting EB size control to improve hESC differentiation [18, 27, 29, 30]. One limitation of our system is that EBs are formed from 2-dimensional (2D) colonies prior to transferring them to suspension. Ultimately, it would be advantageous if the cell trajectory control we have in our 2D system could be translated to direct methods for 3-dimensional cell patterning [29, 30]. We observed that controlling EB size [53] alone was insufficient to achieve reproducible cardiac induction. In fact, comparing cardiac induction in separate EB size-controlled runs with respect to EB size alone yielded more variability than in parallel non-size-controlled runs, with no clear pattern between EB size and cardiac induction efficiency. It became apparent that EB size-controlled cultures were more sensitive to differences in the starting population, which is typically composed of cells expressing pluripotent hESC markers and markers of early hESC derivatives. We thus concluded that analyzing the expression of pluripotency markers (Oct4, Nanog) was inadequate for assessing the input population, and that the composition of cell types expressing markers associated with lineage commitment is a useful indicator for differentiation trajectory biases. Interestingly, using quantitative image analysis, we have observed both the coexpression and unique expression of pluripotency markers and differentiation-associated markers at the protein level. These data suggest that both cell-cell interactions and endogenous parameters influence cell differentiation trajectories. In our system the effect of EB size on mesoderm/cardiac induction was dependent on the ratio of endoderm gene-expressing to neural gene-expressing cells. Importantly, this ratio could exist at a wide range of values within an input hESC population expressing robust levels of typical pluripotency-associated markers. Although signals from neighboring endoderm promote cardiac induction, we did not observe an increase in cardiac induction in endoderm marker expression-enriched starting populations, which demonstrates the necessity to balance the ratio of undifferentiated cells and endoderm/endoderm precursor cells in differentiating EBs. Input populations exhibiting high endoderm-associated marker expression produced differentiation cultures with increased mesoderm/cardiac induction in larger EBs, whereas neural marker-enriched input populations resulted in increased mesoderm/cardiac induction in smaller EBs. This observation corresponds with our previously stated hypothesis that you can modulate the ratio of outer endoderm cells and inner hESCs by controlling EB size; however, this ratio is also a function of the initial ratio of endoderm-associated and neural-associated cells in the hESC population prior to initiating micropatterned colonies.

Controlling the EB size revealed that hESC differentiation studies are hampered by variations in the gene expression profiles of hESC input populations. Our data suggest that colony size can be manipulated to generate the appropriate input cell composition for increased mesoderm/cardiac induction efficiency during EB-mediated hESC differentiation. The ratio of endoderm to neural precursors may be an indicator of an input population's mesoderm/cardiac induction efficiency because, as previously discussed, endoderm is believed to secrete procardiogenic factors, whereas signals (such as Wnts) from the neural tube have been shown to block heart formation in the embryo [38, 39]. By micropatterning hESC colonies of defined diameters, we demonstrated that the level of pluripotency gene expression as well as the ratio of endoderm-associated to neural-associated cells is a function of colony size (Fig. 3). Larger colonies exhibit higher levels of expression of pluripotency genes (Oct4 and Nanog) and lower ratios of endoderm-associated to neural-associated cells, whereas smaller colonies exhibit lower expression levels of pluripotency-associated genes and higher ratios of endoderm-associated to neural-associated cells. These observations are consistent with the literature, including our previous report that enhanced maintenance of the hESC phenotype in larger colonies was attributable to an antagonistic interaction between secreted factors from pluripotent and extraembryonic endoderm cells that regulates small mothers against decapentaplegic (Smad)1 activation [26]. In a larger colony, the higher local cell density (cell number per unit radius) leads to higher levels of BMP antagonists (such as GDF3), which are secreted by hESCs, thereby supporting self-renewal by suppressing Smad1 activation. Meanwhile in smaller colonies, with lower local hESC densities, the secretion of BMP-2 by extraembryonic endoderm activates Smad1 thus suppressing self-renewal. Further, it has been demonstrated that strong interference of endogenous BMP-2 signaling in hESC cultures leads to the induction of cells expressing the characteristic neuroectoderm markers Pax-6 and nestin [54]. Therefore, in smaller colonies (low local cell densities) lower levels of BMP-2 antagonist factor promote endoderm differentiation (higher Gata6/Pax6), whereas in larger colonies (higher local cell densities) higher levels of BMP antagonist suppress endoderm differentiation leading to higher levels of Oct4 and Pax6 expression.


Our findings here have revealed three critical parameters currently neglected in most hESC differentiation experiments: (a) the status of the input hESC composition, (b) hESC colony size, and (c) EB size. Specifically, reproducible, efficient (endogenous) mesoderm/cardiac induction is dependent on (a) the ratio of endoderm to neural precursors in the input hESC population, which can be modulated by controlling hESC colony size, as well as on (b) EB size possibly through a mechanism in which size modulates the ratio of endoderm cells to hESCs during EB development prior to the mesoderm/cardiac induction phase. In combination with future approaches aimed at selecting input hESC populations with more permissive gene expression profiles for specific differentiation trajectories, we posit that micropatterning will enable more rigorous optimization of other tissue environment factors that depend on colony and EB size.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.


The authors thank the following people for their assistance in completing the experiments presented in this report: Manuel Alvarez and Ting Yin for maintaining and expanding the input hESC, and Andra Nica and Chirag Joshi for support with the qRTPCR analyses. This work was supported by grants from the Heart and Stroke Foundation of Ontario (HFSO) and the Canadian Institutes of Health Research (CIHR) (K.A.W. and P.W.Z.).