Reproducible, ultra high-throughput formation of multicellular organization from single cell suspension-derived human embryonic stem cell aggregates

Abstract

Background: Human embryonic stem cells (hESC) should enable novel insights into early human development and provide a renewable source of cells for regenerative medicine. However, because the three-dimensional hESC aggregates [embryoid bodies (hEB)] typically employed to reveal hESC developmental potential are heterogeneous and exhibit disorganized differentiation, progress in hESC technology development has been hindered.

Methodology/principal findings: Using a centrifugal forced-aggregation strategy in combination with a novel centrifugal-extraction approach as a foundation, we demonstrated that hESC input composition and inductive environment could be manipulated to form large numbers of well-defined aggregates exhibiting multi-lineage differentiation and substantially improved self-organization from single-cell suspensions. These aggregates exhibited coordinated bi-domain structures including contiguous regions of extraembryonic endoderm- and epiblast-like tissue. A silicon wafer-based microfabrication technology was used to generate surfaces that permit the production of hundreds to thousands of hEB per cm(2).

Conclusions/significance: The mechanisms of early human embryogenesis are poorly understood. We report an ultra high throughput (UHTP) approach for generating spatially and temporally synchronised hEB. Aggregates generated in this manner exhibited aspects of peri-implantation tissue-level morphogenesis. These results should advance fundamental studies into early human developmental processes, enable high-throughput screening strategies to identify conditions that specify hESC-derived cells and tissues, and accelerate the pre-clinical evaluation of hESC-derived cells.

Figure 1. Conventional hESC differentiation protocols result in heterogeneous aggregates with inconsistent organization and structure: A.

Conventional differentiating hESC aggregates, formed by scraping colonies of hESC off the culture surface, are predominantly disordered. As hESC colonies differ widely in size and shape, this heterogeneity is passed on to the differentiating aggregate. Consequently the local microenvironment is neither consistent between nor within the aggregates. Scale bar represents 200 microns. Inset: Fusing aggregates exhibit expression of the transcription factor CDX2 (green, counterstain 7AAD, red) at points of contact, but not elsewhere, demonstrating the ability of microenvironmental cues to override macro-environmental conditions. Scale bar represents 200 microns. B–D. Rare hESC-derived aggregates exhibit self-organization. Within heterogeneous populations of scraped hESC-derived aggregates, a rare subpopulation of hEB can be observed. These hEB are characterized by the presence of two distinct domains, visible in phase contrast (B). An inner domain is positive for the pluripotency marker Oct4 (C/D – red), while the outer domain is positive for the endodermal marker FoxA2 (shown in green in panel C). A laminin-containing membrane at least partially defines the interface between these two domains (shown in green in panel D, counterstained with Hoechst in blue). Scale bars represent 100 microns.

Figure 2. Controlling aggregate formation and stability: A. Pre-differentiation improves aggregate formation and stability.

hESC cultured on mouse embryonic fibroblast (MEF) feeders were pre-differentiated with 20% serum for 72 hours prior to aggregate formation, resulting an overall reduction in the population level of Oct4 expression [left panel, red line: standard maintenance culture; blue line: pre-differentiated; black: control (unstained)]. Aggregates formed from 2,000 input cells were substantially larger with treatment (blue bar) than without (red bar). Y axis represents aggregate cross-sectional area in microns², error bars represent one standard deviation. B. The ROCK inhibitor Y-27632 promotes aggregate stability. hESC cells cultured on Matrigel in MEF-conditioned medium with and without pre-differentiation [left panel, red line: standard maintenance culture; blue line: pre-differentiated; black: control (unstained)] were used to form SISO-aggregates in the presence or absence of 10 µM Y-27632. Under these culture conditions, in the absence of both, no aggregates were formed (N.D. - size not determined). With 48 hours pre-differentiation in 20% serum, consistent aggregates were formed (first blue bar). When Y-27632 was added to the suspension of cells without (red bar) or with (second blue bar) pre-differentiation immediately prior to dispensing into the well plate, sizeable aggregates resulted.

Figure 3. SISO-aggregation allows for the generation of size-controlled aggregates.

(A) hEB were generated by scraping, and SISO-aggregates were generated from input populations of 400, 2,000 and 10,000 cells in 384-well plates, and recovered by centrifugation. After imaging in phase-contrast mode, images were thresholded and cross-sectional areas were calculated using the ImageJ software package. Values obtained were extremely consistent, with coefficients of variation of 0.09, 0.06 and 0.08 respectively, vs 0.72 for the scraped hEB. (B) The base-10 logarithm of cross sectional area is plotted on a histogram, demonstrating the clear separation between aggregate sizes and dramatic increase in size control over scraping techniques.

Figure 4. Micropatterned surfaces allow ultra-high-throughput (UHTP) production of size-specified aggregates.

Surfaces patterned with arrays of microwells were generated (A) in poly(dimethylsiloxane) via serial replica moulding from a pattern etched into a silicon wafer. Imaging of the silicon master, the PDMS negative cast, and the PDMS positive cast show conservation of form across these steps with wells 100, 200, 400 and 800 microns square (B). Note that the 800 micron wells were generated in the form of a truncated pyramid. Dashed yellow square represents one square millimeter. Sections of PDMS textured with 400 micron wells (C) were inserted into individual wells in a 24-well plate. A single-cell suspension of hESC grown on MEF, predifferentiated with serum for 48 hours was dispensed into the well such that each microwell was predicted to capture the desired number of cells. After 24 hours, the well contents were imaged, extracted, and re-imaged. Scale bar represents 400 microns. D. Centrifugation is not required in the presence of ROCK inhibition. Cells grown on Matrigel in defined conditions were dispensed over the microwells in the presence or absence of 10 µM of the ROCK inhibitor Y-27632. Aggregates formed only in the presence of 10 µM of the ROCK inhibitor Y-27632, in the presence or absence of centrifugation. Scale bar represents 400 microns.

Figure 5. SISO-EB are able to self-organize into highly ordered structures.

A. Aggregates formed from 2,000 hESC, imaged immediately after recovery (“day 0”)(Ai), or after 1, 2, 3 or 4 days respectively (Aii-Av). Note self organization into an ordered domain (white arrow) and a disordered domain (black arrow), and progressive encirclement of the ordered domain by the disordered domain over time (red arrows). Scale bar represents 500 microns. B. An aggregate fixed on day 3 of differentiation exhibits Oct4 positive nuclei (red) located basally within the cell over a laminin-containing basement membrane (green) in the ordered domain, with morphologically distinct Oct4 negative cells located outside the membrane in the disordered domain (counterstained with Hoechst, blue). Scale bar represents 50 microns. C. A day 5 aggregate exhibits staining for the endodermal marker GATA-6 (green) in the encircling disordered domain, and the pluripotency marker Oct4 (red) within the ordered domain. Note the tendancy of the Oct4 positive nuclei to align along the interface between the two tissue types, and a tendancy towards columnar morphology (white arrow), both characteristics of epiblast tissue, and the absence of mixing between the two cell types. The actin cytoskeleton, probed with phalloidin, is shown in blue. Scale bar represents 50 microns. D–F. Day 3 aggregates, showing staining for the endodermal markers GATA4, GATA6, AFP and FoxA2, the primitive-endoderm marker Sox7, and FGF5, a marker characteristic of epiblast and overlying VE. Scale bars represent 50 microns.

Figure 6. SISO-EB are able to give rise to cardiac, hematopoetic and neural cells.

Cardiac differentiation from hEB differentiated for 12 days in suspension culture in serum-containing medium. A) two frames from a video recording of a contractile aggregate, shown before (Ai) and during (Aii) a contraction. Panel Aiii is derived by subtracting panel Ai from panel Aii. The contraction trace (B) was generated by integrating the subtraction image derived from successive frames in the video over the area of contraction, and plotting a 5-point moving average. Neural rosettes (C) were observed after 11 days in culture (4 days in suspension followed by 7 days adherent), staining positive for Pax6 (shown in green) and Sox2 (shown in red). Hematopoetic differentiation was observed after 28 days using Cytospin (D), CFC (E) and flow cytometric (F) assays. G. Quantitative RT-PCR results from aggregates formed from 10,000 pre-differentiated cells in the absence of ROCK inhibition, differentiated for 4 days in suspension followed by an additional 3 days in adherent culture shows down-regulation of pluripotency genes, and up-regulation of markers for endodermal, ectodermal and mesodermal lineages. H. Quantitative RT-PCR results from aggregates formed from 4,000 untreated cells in the presence of ROCK inhibition, differentiated for 4 days in suspension, showed upregulation of mesodermal and endodermal markers, including markers for mesendodermal precursors, but only limited upregulation of the primitive endoderm marker Sox7.