Controlling Expansion and Cardiomyogenic Differentiation of Human Pluripotent Stem Cells in Scalable Suspension Culture

1Cardiovascular disorders can induce severe, progressive loss of contractile heart muscle tissue, including billions of cardiomyocytes (CMs). Because of the low regenerative capacity of the heart, this can ultimately lead to heart failure with very limited treatment options available at present. Given their proliferation and differentiation potential, human pluripotent stem cells (hPSCs), including embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs), are an attractive cell source for the mass generation of lineage- and (potentially) patient-specific progenies, including bona fide CMs. This possibility opens new avenues for the development of regenerative cell therapies and more specific drug discovery assays. Therapeutic and industrial applications of hPSCs, however, will require large cell quantities to be generated under highly robust, well-defined, and economically viable conditions.

2It was recently shown that hPSCs can be expanded as cell-only aggregates in serum-free suspension culture irrespective of matrix supplementation, which is mandatory for conventional surface-attached propagation of hPSCs. In contrast to matrix-attached 2D conditions, suspension culture (3D) provides a straightforward strategy for process upscaling, including cell cultivation in stirred tank bioreactors. Stirred tank reactors represent a universal, well-established vessel type for the production of recombinant proteins in industrial biotechnology and allow for cost-effective, multiparametric monitoring and optimization of mammalian cell culture processes. Once established, relative linear process upscaling is feasible since reactors from 0.1 to >1.000 l culture scale are available. However, the application of stirred bioreactors to hPSC expansion and their differentiation is still in its infancy.

3Single cell-based inoculation of suspension cultures establishes a well-controlled starting point at every passage. The inoculation density as well as the physical properties of the culture system (such as the reactor design and the stirring speed) can then be used to control formation of PSC aggregates and their subsequent growth. Importantly, when utilizing appropriate media such as mTeSR, hPSCs remain pluripotent over multiple passages in aggregate culture, thus providing the attractive option of directly switching from hPSC expansion to lineage-specific differentiation in a continuous suspension process.

4Recent work has demonstrated that Wnt pathway modulation by small molecules is an efficient strategy for hPSC cardiomyogenic induction, resulting in ∼60%–80% CMs content in defined media. A common feature of these protocols is the activation of the Wnt pathway at early stages of differentiation by the GSK3 inhibitor CHIR99021 (CHIR) aiming at enhanced mesoderm induction. Following cues from developmental biology, Wnt pathway activity is then inhibited using inhibitors such as IWP (inhibitor of Wnt production) or IWR (inhibitor of Wnt response). This later step aims at specifying cardiac differentiation of the mesoderm-directed cells. However, these protocols rely on confluent monolayer cultures limiting straightforward industrial scale production.

5In this study, we aimed at directly combining hPSC expansion with cardiomyogenic differentiation in suspension culture. Taking advantage of a NKX2.5-GFP reporter line, a multiwell screening assay was established to develop Wnt modulator-based CMs differentiation of hPSC aggregates in static suspension culture. By scaling up to rotated Erlenmeyer flasks and ultimately to fully equipped stirred tank bioreactors, we show the robustness of the method, as well as its applicability to dynamic suspension culture. The work provides insights on critical cellular and molecular process parameters and a straightforward strategy for the scalable mass production of CMs at up to 85% purity, which predominantly displayed ventricular-like action potentials (APs). Moreover, bioreactor-derived embryoid bodies (EBs) were directly applicable for the generation of human bioartificial cardiac tissue (BCT), a promising strategy for heart repair and novel in vitro drug discovery and drug safety assays.

6To establish combined hPSC expansion and cardiac differentiation, single cell-dissociated HES3 NKX2-5eGFP/w were seeded into 12-well dishes in mTeSR1 (Figure 1A). After 4 days, a reproducible pattern of floating aggregates (214 ± 65 μm average diameter; Figures S1A and S1B available online) formed comprising ∼1 million cells per well (data not shown); ∼98% TRA-1-60 positivity (Figure S1C) suggested maintenance of pluripotency in this culture format, in line with previous findings.

7mTeSR1 was replaced with differentiation medium (3 ml/well; ∼0.33 million cells/ml) supplemented with 0–15 μM CHIR for 24 hr, followed by IWP2 supplementation on day 3 for 48 hr (Figure 1A). On day 10, fluorescence microscopy of floating EBs and flow cytometry revealed the highest content of 52% ± 7.3% NKX2.5-GFP-positive cells (GFP+; indicative of CMs) at 7.5 μM CHIR (Figures 1B and 1C). The highest total cell count was also observed under these conditions (Figure 1D). Somewhat lower cell yields and GFP+ cells were obtained at 5 μM CHIR, whereas significantly lower CMs induction was observed at 10 μM CHIR, which further declined in response to higher drug concentrations (Figures 1B–1D). In DMSO controls (0 μM CHIR), EBs disintegrated over time, suggesting loss of vital cells (Figure 1B).

8Replacing IWP2 by the alternative Wnt pathway inhibitor IWR1 resulted in a shift toward reduced GFP+ cells at 5 μM but elevated GFP+ at 10 μM CHIR (Figure S1M). However, cardiomyogenic induction remained highest at 7.5 μM CHIR, resulting in 58.7% ± 8.8% GFP+ irrespective of IWR1 or IWP2. Thus, IWP2 was maintained throughout further experiments.

9The established protocol was robust in experimental repeats and confirmed by two human iPSC lines differentiated at 7.5 μM CHIR. Immunofluorescence and flow cytometry specific to cardiac troponin T (cTNT), sarcomeric α-actinin, myosin heavy chain (MHC), and NKX2.5, respectively, revealed the induction of up to ∼60% CMs in these hiPSC lines (Figures 1E and 1F).

10Dynamic conditions can have a profound impact on stem cell culture and differentiation. We thus tested whether the static 12-well format can be directly transferred to rotated Erlenmeyer flasks (Figure 2A).

11In single cell-inoculated flasks, a relatively fast increase in aggregate size during the initial 48 hr was observed (data not shown) and aggregates with an average diameter of 389.4 ± 13.8 μm formed in 4 days, with ∼98% of cells remaining TRA-1-60 positive (Figures S1D and S1F). Differentiation by 7.5 μM CHIR resulted in robust GFP expression and formation of contractile EBs (Figure 2B). CM induction was confirmed by immunohistochemistry specific to sarcomeric markers on EB-derived sections and dissociated cells (Figure 2C). On average, 55% ± 7.4% GFP+ cells were obtained on days 7–10 (Figure 2D) with >60% positivity to cTNT and signal-regulatory protein alpha (SIRPα), respectively (Figure 2E).

12To increase the production of CMs, we moved our combined expansion/differentiation protocol to 100 ml stirred bioreactors and examined two alternative expansion strategies. First, we attempted “batch-feeding” (batch), in which cells were allowed to form stirring-controlled aggregates for 48 hr before 100 ml mTeSR1 was exchanged daily. The strategy resulted in a relative linear increase of “batch-aggregates” (BAs) in size over time (Figures 3A and 3B), in line with published observations.

13The second strategy termed “cyclic perfusion feeding” (C-perfusion) aimed at more homogeneous culture conditions by continuous, automated medium exchange: at 24 hr after inoculation, stirring was paused for 10 min every 2 hr to let cell aggregates settle by gravity. About 7 ml of medium was automatically replaced at every cycle, resulting in a turnover of ∼90 ml/day (approximately equivalent to the 100 ml daily medium throughput by batch). Online monitoring of metabolic activity by means of dissolved oxygen (DO) and pH revealed that both parameters remained more stable in C-perfusion compared with batch culture (Figure S2). The pattern of C-perfusion aggregate (CPA) formation was also different, as relatively large spheres readily formed at an early stage (Figures 3A and 3B), reminiscent of the Erlenmeyer flask approach. Such process-dependent aggregate patterns were reproducible in individual runs (Figure 3B), resulting in 283 ± 9.6 μm average diameter for BAs compared with 468 ± 14.4 μm for CPAs on day 0 (compare Figures S1G and S1H with Figures S1J and S1K). Notably, the pluripotency markers TRA-1-60 (Figures 1I and 1L), NANOG, and OCT4 (quantitative RT-PCR [qRT-PCR]; Figure 3F) had similar expression profiles in both culture conditions.

14Global gene expression analysis comparing the undifferentiated aggregates derived from batch and C-perfusion on day 0 revealed a highly equivalent expression signature of established pluripotency, as well as cell proliferation genes and primitive-streak markers (Figure S3A). To identify potential differences, array data were processed in a group comparison by the programs Qlucore Omics and RCUTAS, filtering for genes that were statistically significant and >2-fold regulated, respectively (see respective gene lists in Figures S3B–S3D). A set of 21 genes was found to be upregulated in BAs by both programs (Figure 3C). Notably, these included six members of the metallothionein family (MT1M, MT2A, MT1H, MT1L, MT1E, MT1B), known to bind metal ions (gene ontology categories: cellular response to zinc ion, metal ion, inorganic substances, negative regulation of growth). In CPAs, only three genes were found to be upregulated under the same comparative criteria: the TGF-beta superfamily member BMP2, the caudal type homeobox transcription factor (CDX4), and the melanin-concentrating hormone receptor 1 (MCHR1; Figure 3C). Exemplary verification of the array data confirmed upregulation of BMP2 in CPAs and MT1M in Bas, as depicted in Figure S3E.

15Before inducing aggregate differentiation, cell numbers were determined. On average, batch expansion yielded ∼79 million HES3-NKX2-5eGFP/w cells per bioreactor, whereas ∼60 million cells were recovered by C-perfusion (n = 2 runs each; data not shown). Subsequently, aggregates were diluted in differentiation medium (7.5 μM CHIR) to establish ∼33 million cells per 100 ml (∼0.33 million cells/ml equivalent to 12-well cultures) per bioreactor. After this “equalization step,” the same differentiation scheme (Figure 3D) was applied in stirred reactors irrespective of the preceding expansion strategy. Surprisingly, we found that differentiation runs (n = 3) initiated with BAs repeatedly failed to form contracting EBs and essentially no GFP+ cells were detected (Figure 3E). In contrast, contracting EBs were observed upon CPAs differentiation, which was accompanied by GFP expression in all experimental repeats (n = 3; Figure 3E).

16NANOG and OCT4 expression was equivalent in both expansion approaches before differentiation (day 0; Figure 3F). However, after 24 hr of CHIR treatment (day 1), NANOG dropped substantially in BAs but remained almost unchanged in CPAs before downregulation occurred between days 1 and 3. A decline of OCT4 expression was also delayed in CPAs (Figure 3F).

17Regarding the metabolic activity, induction of differentiation of CPAs resulted in a steep DO drop from 100% to ∼65% in 24 hr, whereas a moderate reduction to ∼85% occurred in BA differentiation (exemplary depicted in Figure S2). Accordingly, the pH dropped from 7.5 to 6.9 in CPA but only to 7.2 in BA differentiation. Thus, although differentiation cultures were initiated at equivalent cell densities and performed under the same conditions, the overall metabolic activity was substantially higher in CPAs in the presence of CHIR and remained higher compared with differentiation processes of BAs during IWP2 treatment and thereafter (Figure S2).

18Robust upregulation of primitive streak markers Brachyury T (T) and MIXL1 was observed after CHIR treatment (day 1), irrespective of aggregate origin, suggesting equivalent progression of early differentiation in both conditions (Figure 3F). A differential pattern was observed for MESP1, a key regulator of mesoderm and cardiovascular cell-fate determination, and TBX3, promoting early lineage, mesendoderm, and CM specification. Upregulation of both genes was more rapid in differentiating BAs (which ultimately failed to form CMs) preceding expression in CPAs by 24–48 hr (Figure 3F). Subsequently, upregulation of the early cardiac progenitor markers ISL1 and NKX2.5 was almost exclusively observed in CPA differentiations, consistent with the GFP expression. On the other hand, GATA4 expression, a gene involved in cardiac mesoderm, as well as endodermal lineage specification, was highly equivalent in cells from both expansion strategies (Figure 3F).

19Focusing on CM formation, CPA-based differentiation processes were analyzed in more detail. Upregulation of cTNT (TNNT2) and α-MHC (MYH6) was observed by qRT-PCR from day 5 onward (Figure 4A) in line with contracting EBs from days 6 and 7 onward (data not shown). Interestingly, α-MHC expression dropped when comparing day 7 versus day 10 samples, whereas β-MHC (MYH7), a marker of CMs maturation, showed robust upregulation (Figure 4A).

20At endpoint analysis (day 10), the vast majority of EBs were bright GFP+ (Figure 4B) and contracting (Movie S1). On average, 62.9% ± 7.3% GFP+ and 68.6% ± 8.7% cTNT+ cells were observed in three individual bioreactor runs (53.9%–84.1%; Figure 4C and Table 1). When multiplied with total cell yields, on average, ∼40 million CMs were observed per process (Figure 4C), whereby more than ∼50 million CMs were generated in individual runs. Flow cytometry for sarcomeric markers α-actinin, MHC, and the cell surface marker SIRPα showed CM content of up 85% CMs at process endpoint (Figure 4E). This observation was further confirmed by immunofluorescence staining on EB sections or EB-derived, seeded cells displaying typical cross-striations (Figure 4D).

21To substantiate the robustness of the C-perfusion-based differentiation, additional experiments were performed using hHSC_F1285T_iPS2 cells. The CPA formation pattern of the hiPSC line (Figure S4A and S4B) was highly equivalent to the HES3 NKX2-5eGFP/w pattern (Figures 3A and 3B), resulting in an average aggregate size of 531.6 ± 35.3 nm before differentiation (day 0; Figure S4B). The subsequent cardiac differentiation led to homogeneously contracting EBs (Movie S2). On average, three independent runs revealed 53.2% ± 16.4% positivity for MHC, 58% ± 30.3% cTNT, and 51.2% ± 14.6% α-actinin (Figure S4C; range, 27.2%–83.5%; MHC, 27.7%–88.3% cTNT; Table 1).

22Reactor-derived CMs were seeded onto MEAs to measure field potentials in the presence of chronotropic and arrhythmogenic drugs. Consistent with published data, we have detected a positive chronotropic response by the beta-adrenergic agonist isoproterenol (Figure 5A). The class I antiarrhythmic compound quinidine, known to have complex interactions with ion channels, induced reversible spike amplitude reduction and prolonged the field potential duration (FPD) at 10 μM (Figure 5B), in line with recent findings. The class IV antiarrhythmic drug verapamil exerts its action by blocking calcium channels but is also a potent hERG blocker. The compound induced a dose-responsive negative chronotropic effect at 100 nM. This effect was more prominent at 300 nM and accompanied by a shortening of the FPD (Figure 5C), which most likely resulted from multiple channel blocks, thereby compensating the known hERG blocking effect, in agreement with clinical data.

23Electrophysiological properties were further analyzed by whole-cell patch clamp recordings in the current clamp mode. Cells were classified according to the shape of their APs (Figure 5D). Thirty five of 41 cells (comprising 20 and 21 cells, respectively, derived from two independent differentiations) displayed ventricular-like APs. Six were classified as atrial-like (Figure 5E), whereas no AP patterns indicating nodal-like CMs were identified. The data were corroborated by comparing spontaneous versus evoked APs applying short depolarizing current steps (Figure S5).

24Taken together, these pharmacological and electrophysiological results underscore that our method led to the formation of bona fide human CMs, displaying typical characteristics of early hPSC-derived cells.

25Bioreactor-derived EBs were also used for the generation of BCTs. After matrix solidification, EBs formed synchronously contracting syncytia during the initial cultivation phase (days 0–4; data not shown). The addition of fibroblasts to the EB/collagen-I matrix was not essential to achieve tissue consolidation in contrast to a prior approach using genetically enriched EBs, which consisted of >99% CMs. However, qualitative microscopic comparison of tissue formation on day 21 suggested slightly brighter NKX2.5-GFP intensity of constructs that were prepared with the addition of human foreskin fibroblasts (with HFF; n = 5) compared with the EB-only group (without HFF, n = 4; exemplary depicted in Figure 5F), suggesting that HFF might support CMs persistence in respective constructs. Furthermore, BCTs generated with HFFs showed significantly higher contraction forces compared with the control group (1.59 ± 0.2 versus 0.4 ± 0.23 mN) and a more physiological Frank Starling mechanism, whereas no significant differences in passive forces were observed between these experimental groups (Figure 5G).

26Focusing on critical culture parameters, we have developed and upscaled the production of hPSC-derived CMs in suspension culture and show their utility for tissue engineering.

27Chemical Wnt pathway modulators were recently applied to direct cardiomyogenic differentiation of hPSCs seeded on matrigel or the synthetic matrix Synthemax. Disregarding upscaling limitations of monolayer cultures, respective matrices might be a prerequisite for these protocols to work. By successful transition to free floating aggregates, we demonstrate that matrices and direct cell-to-substrate contacts are dispensable, stripping away another (costly) layer toward fully defined conditions.

28In 2D, the generation of confluent monolayers ahead of differentiation was noted to be critical for efficient cardiomyogenesis. This suggests that the cell density, which impacts on cell-cell contacts and the concentration of paracrine factors, plays a role. thus provided cell numbers for monolayer inoculation, but the resulting cell densities at the induction of differentiation were not noted. In contrast to surface-attached cultures, cell density assessment and adjustment are straightforward in suspension. Particularly in stirred bioreactor scale, regular sampling allows close monitoring of cell counts and the assessment of other cell and aggregate properties, thereby ensuring good process definition and reproducibility. Here, we initiated differentiation when cultures reached a defined density of ∼0.33 × 106 cells/ml. Although further investigations on the effect of cell density are required, this ensures improved cross-platform and intercell line comparability and provides a valuable reference point.

29Keeping all other parameters constant, the concentration of the GSK3 inhibitor CHIR dictated the differentiation outcome in our screening platform; 7.5 μM CHIR was found to be optimal for cardiomyogenesis of hES3 NKX2.5-GFP cells in static and dynamic suspension culture and was also applicable to two independent hiPSC lines. In hESC monolayers, 10 μM CHIR was found to be optimal for inducing primitive streak markers, mesoderm, and ultimately cardiomyogenesis, whereas 12 μM worked best in a related setup. Here CHIR concentrations >7.5 μM significantly reduced the CM content and overall cell yields, suggesting differential activity of GSK3 in suspension versus monolayer cultures. Alternatively, or in addition, the activity of other canonical Wnt pathway components and/or other pathways effecting cardiomyogenesis might differ in 3D versus 2D.

30Numerous studies have suggested an impact of a defined sphere size on hPSC lineage differentiation. In microwell plates, for example, 1,000 hESCs per aggregate differentiated more efficiently into CMs compared with 100 or 4,000 hESCs per aggregate, respectively, whereby the size-dependent formation of definitive endoderm was found to trigger cardiac mesoderm induction by paracrine factors. In hydrogel microwells, cardiogenesis was enhanced in EBs of 450 μm diameter, whereas endothelial cell differentiation was increased at 150 μm, apparently due to EB size-dependent expression of noncanonical WNTs. In simple, agarose-based microwells, Dahlmann et al. generated defined sphere batches ranging from 666 to 2,666 hESCs per aggregate (representing 185 to 270 μm mean diameter; J. Dahlmann, personal communication), respectively, followed by aggregate harvest and differentiation in dynamic suspension. Although the initial aggregate size had an impact on hPSC growth kinetics and subsequently on overall cell and CM yields, the resulting CM content remained in a relative narrow range of ∼40%–50% at all aggregate sizes tested.

31Despite the excellent experimental utility of culture platforms generating uniform aggregates or colony patterns, respective studies often ignore culture aspects such as the overall cell density or lack informative controls of mixed sphere sizes. Comparing EB formation methods, such as random induction in suspension versus size-controlled spin EBs, noted that, despite a potential influence of the EB size, cell density-dependent, and (co) culture-dependent medium, conditioning dominantly defined hESC hematopoietic differentiation. Generating defined EBs by forced aggregation of mouse ESCs instructive clues on differentiation were rather induced by rotation speed-dependent hydrodynamic forces, whereas minor, if any, EB size-dependent effects were observed.

32In our static 12 wells, the average aggregate diameter was ∼210 μm, representing a relatively large spread of ∼100–400 μm in each well, typical of random aggregation. However, at 7.5 μM CHIR, aggregates of all shapes and sizes differentiated into contracting, GFP+ EBs, whereas, at suboptimal CHIR concentrations, GFP expression was reduced or failed in respective wells, again without apparent aggregate/EBs size correlation. In stirred Erlenmeyer Flasks, the mean aggregate diameter was ∼400 μm (∼300–600 μm), and in bioreactors, at C-perfusion, it was ∼450 μm (∼350–600 μm), representing about double the mean diameter in 12 wells. These culture-dependent disparities were still compatible with efficient CM formation applying the same preoptimized differentiation protocol. In contrast, following batch expansion, cardiomyogenesis completely failed in repeated bioreactor runs, although the mean aggregate diameter of ∼300 μm (∼150–450 μm) was within the range of successful conditions.

33Taken together, our cross-platform findings suggest that the initial aggregate size before induction of differentiation per se is not the dominant factor determining cardiac differentiation outcomes in suspension culture. Within a given size range, variations can be tolerated.

34In the hPSC expansion phase, batch feeding was applied in static 12-well dishes and in stirred Erlenmeyer flasks, whereas in bioreactors, (cyclic) perfusion feeding led to successful cardiac differentiation. This demonstrates that different cell feeding protocols ahead of differentiation were compatible with subsequent cardiomyogenesis.

35In the bioreactor setup the spatiotemporal pattern of aggregate development was the major determinant of subsequent differentiation results. We and others have shown that the pattern of PSC aggregation and further sphere development can be controlled well in stirred bioreactors by means of the inoculation density, the impeller design, and the stirring speed, including the on/off patterns applied in this study. In addition to the feeding-dependent differences of the culture milieu, the striking distinction was the continuous growth of initially small aggregates at batch, whereas at C-perfusion relative large CPAs were formed at 24 hr with a minor increase in size thereafter.

36Although global profiling revealed rather uniform gene expression patterns typical of pluripotent cells under both expansion conditions, expression of the key morphogen BMP2 was highly upregulated in C-perfusion cultures. Modest BMP2 and BMP4 expression is suggested to act as an endogenous prodifferentiation signal in hESC, although these levels are insufficient to promote differentiation at pluripotent culture conditions. However, upon differentiation, the dominant role of BMP signaling on posterior primitive streak and subsequently on mesoderm formation in hPSCs was recently underscored. The upregulation of BMP2 as well as CDX4 (involved in anteroposterior axis specification regulated by Wnt activity; ) in CPAs might suggest the priming of these cells for mesoderm differentiation ahead of CHIR supplementation.

37further noted the necessity to neutralize endogenous BMP to eliminate mesoderm induction of hPSCs. The upregulation of BMPER (BMP endothelial cell precursor derived regulator), an established antagonist of BMP2, BMP4, and BMP6, in BAs (bottom panel in Figure 3C) might act as a neutralizing signal and, at least in part, explain the entire lack of cardiomyogenesis upon differentiation of these cells.

38The control of cardiac differentiation by BMP signaling is well established. However, the role of autocrine/paracrine hPSC-derived BMP agonists and antagonists in modulating WNT pathway controlled differentiation is not well studied. Although future investigations on this are necessary, our data indicate the importance of this interplay.

39However, primitive streak marker expression, including T and MIXL, was clearly present in BA-derived EBs after CHIR supplementation, suggesting that mesendodermal differentiation proceeds normally. Our data suggest that the temporal shift in the expression of key lineage transcription factors, such as MESP1 and TBX3, in CPA compared with BA cultures might underlie the divergence in differentiation outcomes.

40We further found that the expression of 6 of 13 tested subtypes of the major metallothionein isoforms 1 and 2 (MT1, MT2) was upregulated in BAs. Since MTs bind to metal ions, a role of these proteins in hematopoietic cell proliferation and differentiation was suggested, but regulation of MTs expression by numerous stimuli, including oxidative stress, has been described. In mouse ESCs, upregulation of MT1 was discovered after addition of the p38 mitogen-activated protein kinase inhibitor PD169316 at LIF starvation-induced stress conditions. Overexpression of MT1 was notably sufficient to protect mESC from differentiation-induced apoptosis. To date, little is known about the functional role of MT in human PSCs, but we hypothesize that the feeding-induced fluctuation of the culture environment at batch conditions induces MT expression, as compared with more homogeneous conditions at C-perfusion. Although further investigation is required, metallothionein might thus present an interesting marker of stress response in hPSC culture. In this context, it is noteworthy that the stress-related genes HSPA1A and HSPA1B (HSP70 protein) were both found to be ∼2-fold upregulated in BAs compared with CPAs (data not shown).

41Following fundamental analysis of pharmacological and electrophysiological features of suspension-derived CMs, we have demonstrated direct applicability of bioreactor-produced EBs for tissue engineering. While BCT generation from undissociated, contracting spheres was recently shown, this required genetic enrichment of CMs before tissue formation. On the other hand, enrichment to ∼99% CM purity was incompatible with direct BCT generation, but required 15% fibroblast addition to remodel the initial collagen I matrix and hence support tissue formation. Here, we show that the ∼80% CM content resulting from our differentiation protocol was directly compatible with the production of functional tissues, suggesting that residual non-CMs within EBs provided structural support. A detailed assessment of the phenotype and features of respective non-CMs is currently in progress.

42Taken together, we show that bioreactor-controlled programs of hPSC culture can be used to direct the subsequent fate of hPSCs on differentiation. As recently noted, it is now necessary to focus on the improvement of mass suspension culture for hPSC production and differentiation. Our study provides a substantial step along this path.

43HES3 NKX2-5eGFP/w, hCBiPS2, and hHSC_F1285T_iPS2 were maintained at standard conditions on MEFs. Before transition to suspension culture, cells were accutase-treated (PAA Laboratories) and seeded at 5 × 104 cell/cm2 on Geltrex-coated flasks either in KnockOUT-SR medium (Life Technologies) conditioned with MEFs or in mTeSR1 (STEMCELL Technologies) supplemented with 10 μM Y-27632 for 24 hr. Medium was replaced daily, and cells were passaged twice per week.

44For the inoculation of suspension culture, dissociation into single cells was performed by accutase treatment for 5 min at 37°C followed by dilution to 3.3 × 105 cells/ml in mTeSR1 plus 10 μM Y-27632 and seeding into 12-well suspension plates (GreinerBioOne) or Erlenmeyer flasks (125 ml scale; VWR-International) in 1.5 ml/well or 20 ml/flask, respectively. Flasks were agitated at 75 rpm (orbital-shaker, Infors-HT).

45Stirred DASbox minibioreactors (DASGIP/Eppendorf) were inoculated as described previously (Supplemental Experimental Procedures). For subsequent expansion strategies, see the Results.

46Differentiation of resulting aggregates was induced on day 0 (4 days after single cell inoculation) using CHIR99021 (synthesized by the Institute of Organic Chemistry, Leibniz University Hannover or purchased from Millipore) at indicated concentrations for 24 hr. On day 3, IWP2 (Tocris) was added at 5 μM for 48 hr. During the differentiation, cells were kept in RPMI1640 (Life Technologies) supplemented with B27 (minus insulin). Medium was entirely replaced on days 0 (+CHIR), 1, 3 (+IWP2 or IWR1), and 5 applying a volume of 3, 20, or 100 ml depending on the respective culture format, i.e., 12-well dish, Erlenmeyer flask, or stirred bioreactor. Aggregates were cultured in RPMI1640 supplemented with B27 from day 7 onward.

47Aggregate or EB samples were monitored by light microscopy; images were captured (AxiovertA1; Zeiss) and processed in AxioVision (Zeiss) to define diameter and size distribution. Mean diameters represent the arithmetic average of 400–700 independent spheres. For cell seeding, samples were dissociated by 1 mg/ml collagenase B (Roche) for 15–60 min (progressing with differentiation) at 37°C.

48For standard methods and prepublished techniques, see the Supplemental Experimental Procedures.

49Data are presented as mean ± SEM. Unless otherwise noted, statistical significance was calculated using Student’s t tests. To take account of multiplicity in group comparisons, one- or two-way ANOVA followed by Bonferroni’s posttest was conducted. Statistical significance was assigned as ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.