Interrogating cellular fate decisions with high-throughput arrays of multiplexed cellular communities

Abstract

Recreating heterotypic cell-cell interactions in vitro is key to dissecting the role of cellular communication during a variety of biological processes. This is especially relevant for stem cell niches, where neighbouring cells provide instructive inputs that govern cell fate decisions. To investigate the logic and dynamics of cell-cell signalling networks, we prepared heterotypic cell-cell interaction arrays using DNA-programmed adhesion. Our platform specifies the number and initial position of up to four distinct cell types within each array and offers tunable control over cell-contact time during long-term culture. Here, we use the platform to study the dynamics of single adult neural stem cell fate decisions in response to competing juxtacrine signals. Our results suggest a potential signalling hierarchy between Delta-like 1 and ephrin-B2 ligands, as neural stem cells adopt the Delta-like 1 phenotype of stem cell maintenance on simultaneous presentation of both signals.

Figure 1. Two-step patterning process and single-cell-tethering workflow.

(a) Microisland patterns were produced by UVO (185 nm) patterning into thin polyHEMA coatings (<0.5 μm). An aldehyde-functionalized organic silane was then vapour deposited to prepare for DNA printing. (b) Profilometry measurements show representative microisland features of 200 nm. (c) Spot arraying of NH₂-terminated oligonucleotides within each microisland was performed using the Nano eNabler system. After arraying of single-cell-sized spots, the entire slide underwent reductive amination using NaBH₄. (d) Representative image of four-component printed DNA patterns (scale bar, 100 μm). (e) Multiple cell populations are labelled with distinct DNA molecules presenting sequences complementary to the microisland DNA strands, washed and passed through a PDMS flow cell affixed to the patterned slides either sequentially at a density of ∼800,000 cells per cm² or in mixed solutions at a density of ∼400,000 cells per cm². Untethered cells are washed away, and the process is repeated for each cell type.

Figure 2. Customizable capabilities of two-step surface-patterning platform for modulating cellular interactions.

(a) Both cell number and identity can be precisely controlled. (b) As an example of the latter, four MCF10A cell populations, each coloured with a different dye, were labelled with distinct DNA strands and arrayed onto microislands printed with four of the complementary DNA oligonucleotides. Seven out of the nine displayed microislands possessed the correct cellular community, with yellow arrows indicating microislands containing incorrect cellular components. (c) Using this DNA-based cell tethering, the efficiency of exact MCF10A cell patterning (red circles) was considerably higher than the same four cell populations plated at a low cell/surface ratio for random Poisson seeding (green triangles) of single-, double-, triple- and quadruple-cell communities. (d) Efficiency, or fold improvement, of our DNA-patterned compared with Poisson-loaded arrays. (e) Variations to the microisland features for further modulation of cell–cell communication can be achieved by changing the size and shape of the photomask used during ultraviolet (UV) etching. (f) DNA printing enables precise control over the initial cell positions of NSC–astrocyte (bottom cell–top cell) pairs. All error bars are s.e.m. and n=4. All scale bars, 100 μm.

Figure 3. Arrays of cellular communities yield insights into cell dynamics and NSC differentiation, proliferation and signal arbitration of opposing juxtacrine signals at the single-cell level.

(a) Migration and cell–cell contact for each microisland can be tracked with time-lapse microscopy. Representative 48-h time-lapse images illustrating the dynamics of two NSC–astrocyte pairs initially patterned at different separations. NSC highlighted in red, and astrocyte highlighted in blue. (b) Percent of cellular communities that showed contact increased as the initial distance separating NSC and astrocyte decreased. (c) Total contact times also increased as initial cell–cell distance decreased. (d) Cell communities could be repeatedly imaged over long timescales with subsequent visualization of differentiation markers. Representative, stitched montages of NSCs (upper) and cortical astrocytes (lower, green) immediately after patterning (left), then after immunostaining after 6 days for the neuronal marker Tuj1 and astrocyte marker GFAP (right). Higher magnification of a representative adhesive microisland shows that all progeny of this particular single NSC founder differentiated into Tuj1⁺ neurons. (e) NSC differentiation can be tracked for each community. When patterned with single naive astrocytes, NSCs exhibited enhanced Tuj1 differentiation and similar GFAP differentiation when compared with low-density and high-density bulk co-cultures. (f) Microisland confinement enabled analysis of proliferation rates. Proliferation rates (r) for Tuj1-biased lineages (lineages in which no GFAP cells were present) were higher than proliferation rates for GFAP-biased lineages (P=8e⁻⁴). (g) NSCs patterned with a single hEfnB2-overexpressing astrocyte exhibited enhanced Tuj1⁺ differentiation. (h) NSCs patterned with a single hDll1-overexpressing astrocyte displayed low Tuj1 expression. (i) When a single NSC was in the presence of both a Dll1 astrocyte and an EfnB2 astrocyte, the Dll1 phenotype (that is, reduced Tuj1) dominated. The left graph represents immunostained proportions of NSCs in each condition, and the right graph depicts immunostaining changes compared with NSCs patterned 1:1 with a naive cortical astrocyte. All error bars are 95% confidence intervals; all P values obtained from t-test. ***P<0.001, **P<0.01, *P<0.05. All scale bars, 100 μm.