Stem Cell Colony Interspacing Effect on Differentiation to Neural Cells

Abstract

Efforts to enhance the efficiency of neural differentiation of stem cells are primarily focused on exogenous modulation of physical niche parameters such as surface topography and extracellular matrix proteins, or addition of certain growth factors or small molecules to culture media. We report a novel neurogenic niche to enhance the neural differentiation of embryonic stem cells (ESCs) without any external intervention by micropatterning ESCs into spatially organized colonies of controlled size and interspacing. Using an aqueous two-phase system cell microprinting technology, we generated pairs of uniformly-sized isolated ESC colonies at defined interspacing distances over of a layer of differentiation-inducing stromal cells. Our comprehensive analysis of temporal expression of neural genes and proteins of cells in colony pairs showed that interspacing two colonies at ~0.66 times colony diameter (0.66D) significantly enhanced neural differentiation of ESCs. Cells in these colonies displayed higher expression of neural genes and proteins, and formed thick neurite bundles between the two colonies. A computational model of spatial distribution of soluble factors of cells in interspaced colony pairs showed that the enhanced neural differentiation is due to the presence of stable concentration gradients of soluble signaling factors between the two colonies. Our results indicate that culturing ESCs in colony pairs with defined interspacing is a promising approach to efficiently derive neural cells. Additionally, this approach provides a platform for quantitative studies of molecular mechanisms that regulate neurogenesis of stem cells.

INTRODUCTION

Embryonic neurogenesis is a highly regulated process that involves a series of events including stem cell proliferation, spatiotemporally orchestrated differentiation, and highly regulated cell migration and patterning (Martynoga, Drechsel, & Guillemot, 2012). Various cues such as time-varying concentration gradients of endogenous soluble morphogens, cell-cell signaling, cell-matrix interactions, and biomechanical forces synergistically regulate neurogenesis of stem cells (Germain, Banda, & Grabel, 2010; Solozobova, Wyvekens, & Pruszak, 2012), and positioning and morphology of specific neural cells and tissues in the central nervous system (Hayashi, Kubo, Kitazawa, & Nakajima, 2015; Martynoga et al., 2012; Paridaen & Huttner, 2014). Availability of different cellular and animal models for embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) research has significantly progressed the understanding of embryonic neurogenesis. From a practical standpoint, exploiting this knowledge is critical to efficiently generate specific neural cells in vitro for cell replacement therapies of neurodegenerative diseases, and to offer neurological disease models that recreate patients’ disease pathogenesis in laboratory settings (López-Bendito & Arlotta, 2012; Prajumwongs, Weeranantanapan, Jaroonwitchawan, & Noisa, 2016). A major step toward implementing these strategies is to closely recapitulate the native cellular microenvironment (Bratt-Leal, Carpenedo, & McDevitt, 2009; Gattazzo, Urciuolo, & Bonaldo, 2014; Yan et al., 2018).

Cell and protein micropatterning technologies, micro-well arrays, and microfluidic systems enable stem cell cultures that mimic spatiotemporal organization and heterocellular interactions of cells and biochemical signaling associated with differentiation to specific cells (Béduer et al., 2012; Parekkadan et al., 2008; Peerani et al., 2009). Various research groups have used these approaches to confine stem cells into defined geometric patterns and study stem cell pluripotency and differentiation (Dennis E. Discher, David J. Mooney, & Peter W. Zandstra, 2009; Kolind, Leong, Besenbacher, & Foss, 2012). For example, culture of human ESCs (hESCs) in certain geometrical patterns resulted in defined colony sizes that determined self-renewal or differentiation of hESCs. Cells in colonies of 400 μm diameter better maintained a pluripotent state than those in smaller colonies of 200 μm diameter (Peerani et al., 2007). When hESCs were patterned on circular cell adhesive islands, colonies of 1200 μm diameter primarily showed mesodermal differentiation to hematopoietic progenitor cells, whereas smaller colonies of 200 μm diameter favored endodermal differentiation to primitive gut cells (Lee et al., 2009). Patterned mouse ESCs (mESCs) in bow-tie micro-wells showed neuroectoderm differentiation, potentially through a mechanism involving connexin-43 (Parekkadan et al., 2008). Microenvironmental parameters such as colony size and separation and degree of clustering modulate paracrine signaling through Jak-Stat pathway to determine stem cells fate (Peerani et al., 2009). Although various microtechnologies have enabled differentiation of stem cells to several lineages, generation of neural cells from stem cells in micropatterned environments is still underexplored.

To derive neural cells from ESCs, the stem cells may be grown as three-dimensional (3D) aggregates in suspension cultures and differentiated using retinoic acid; grown as a monolayer and differentiated using defined media; or grown on a stromal feeder layer to allow intercellular signaling direct neural differentiation (Kawasaki et al., 2000; Kim et al., 2009; Ying, Stavridis, Griffiths, Li, & Smith, 2003). Each of these approaches has some limitations. For example, retinoic acid can hinder the natural neural patterning and maturation of neural cells (Papalopulu et al., 1991). Neural precursors in the monolayer culture method need dissociation and selective re-plating to differentiate to neural cells (Bouhon, Joannides, Kato, Chandran, & Allen, 2006; Ying et al., 2003). The co-culture method uses stromal cells of non-human origin to induce neural differentiation of stem cells. Although several studies established that secreted soluble growth and neurotrophic factors are potent drivers of ESCs neural differentiation (Kawasaki et al., 2000; S. Park et al., 2004; Schwartz et al., 2012; Tavana, Mosadegh, & Takayama, 2010; Vazin et al., 2009; Vazin, Chen, Lee, Amable, & Freed, 2008), the mechanism underlying the neural differentiation is still not fully known Nevertheless, the co-culture approach mimics ESCs niche in vivo in terms of direct intercellular contacts and signaling, and avoids using differentiation-inducing chemicals. Therefore, it is more suitable to exclusively study effects of niche configurations on neural differentiation. Using the co-culture technique, we previously modulated neural differentiation of mESCs by varying the stem cell colony size (Joshi, Thakuri, Buchanan, Li, & Tavana, 2017). We used a cell printing technology based on a polymeric aqueous two-phase system (ATPS) to generate size-controlled individual mESC colonies over live stromal cells. Comprehensive protein and gene expression studies showed that colonies that reached 2.0 mm in diameter after one week of culture gave significantly and size-disproportionately greater neural cells than smaller colonies of 1.4 mm and 1.0 mm diameters.

Here, we use the ATPS cell micropatterning approach to control the interspacing between pairs of mESC colonies on stromal cells and study colony interspacing effect on neural differentiation. Using comparative neural gene and protein expression studies, we found that colony interspacing has a significant role on the yield of neural cells and identified potential molecular drivers. Our stem cell niche micropatterning approach may be used to modulate microenvironmental signals and spatial gradients of endogenous secretome to generate neural cells with improved efficiency.

MATERIALS AND METHODS

Maintenance of Cells and Preparation of Stromal Cells

The mESC line EB5 (Riken) was maintained on 0.1% gelatin-coated dishes in a maintenance medium comprising GMEM, 1% fetal bovine serum (FBS), 10% knockout serum replacement (KSR), 2 mM glutaMAX, 0.1 mM non-essential amino acids (NEAA), 1 mM sodium pyruvate, 0.1 mM 2-mercaptoethanol, and 2000 U/ml leukemia inhibitory factor (LIF). The stromal PA6 cell line (Riken) was also maintained on gelatin coated dishes in αMEM supplemented with 10% FBS and 1% antibiotic. Except for FBS and LIF that were purchased from Sigma and Millipore, respectively, the remaining reagents were obtained from Thermo Fisher Scientific. To prepare feeder layer for mESCs differentiation, 4×10⁴ PA6 cells were seeded on gelatin-coated 35 mm Petri dishes and grown for 2 days to form a confluent monolayer. The PA6 cells were then mitotically inactivated with 10 μg/ml mitomycin-c (Sigma) for 2 hours, washed three times with PBS, and incubated overnight at 37°C and 5% CO₂ in a “differentiation” medium comprising GMEM, 10% KSR, 2 mM glutaMAX, 0.1 mM NEAA, 1 mM sodium pyruvate, and 0.1 mM 2-mercaptoethanol before co-culturing with mESCs.

Microprinting of mESCs onto Stromal Cells

BioUltra polyethylene glycol (PEG), Mw: 35kDa, (Sigma) and dextran (DEX), Mw: 500kDa, (Pharmacosmos) were dissolved in the differentiation medium at concentrations of 5.0% (w/v) and 12.8% (w/v), respectively.(Atefi, Joshi, Mann, & Tavana, 2015) mESCs were suspended in an equal volume of the medium and the DEX phase solution to yield a cell density of 100 cells per 30 nl, and reduce DEX concentration to 6.4% (w/v). The cell suspension was loaded into a small number of wells from a 384-well plate (source plate). A robotic liquid handler (SRT Bravo, Agilent) was used to load 30 nl of the suspension into slot hydrophobic pins (VP Scientific). The pins were then slowly lowered into a 35 mm Petri dish containing a monolayer of PA6 cells immersed in 1 ml of the PEG phase solution (Fig 1a). The DEX phase containing mESCs autonomously dispensed from each pin and formed a drop on the PA6 cells layer (Fig 1b). The pins were then retracted from the dish and cleaned by dipping consecutively into a cleaning solution (VP Scientific), two water baths, and an isopropanol bath. The clean pins were reloaded with the mESC suspension in the aqueous DEX phase. The tip magazine of the liquid handler was offset with a defined distance and the printing process was repeated to generate a set of second DEX phase drops neighboring the first set (Fig 1c). The interspacing between the two drops was determined by the offset distance of the tip magazine. Polymeric solutions were replaced with the fresh differentiation medium after 3 hours of incubation that allowed the mESCs adhere to the PA6 cells layer (Fig 1d). Cultures were maintained in the differentiation medium for 8 days and with supplemented 1X N2 (Life Technologies) for an additional 6 days. Nine pairs of mESC colonies of similar size were generated on a PA6 layer in each 35 mm Petri dish and maintained for 14 days. This approach was used to generate colony pairs of three different interspacing distances to study the effect on neural differentiation of mESCs.

Figure 1.

Aqueous two-phase system microprinting of mESCs on PA6 stromal cells to generate a pair of mESC colonies with a defined interspacing. (a) Dispensing pins are loaded with mESCs suspended in the aqueous DEX phase. (b) Pins are lowered into the culture plate containing PA6 cells monolayer immersed in the aqueous PEG phase and the pins content autonomously dispenses. (c) The process is repeated once more to form two isolated drops confining mESCs and with a defined interspacing. (d) Microprinted mESCs from both drops attach to the stromal layer and proliferate to form two colonies. (e) A pair of distinct DEX phase drops confining mESCs. (f,g) Colony pairs at two different interspacing distances on day 8 of culture. (h) Measured edge-to-edge distance between the two colonies formed varies linearly with center-to-center distance between the corresponding DEX phase drops. (i, j) Neurite processes extending out between the two colonies in panels b and c, respectively. * p < 0.01. n=18. Error bars represent mean ± S.E.M.

Immunofluorescence, Imaging, and Image Analysis

The differentiated mESC colonies were fixed in 3.7% formaldehyde on days 8 and 14, and incubated for 1 hour at room temperature in a blocking solution containing 5% donkey serum. Both primary antibody and secondary antibody solutions were prepared in PBS containing 0.05% Tween 20 (Sigma). The following primary antibodies were used at the indicated dilutions: affinity purified chicken Nestin (1:100, Neuromics), rabbit monoclonal class III β-tubulin (1:400, Cell Signaling), and affinity purified chicken GFAP (1:2000, Neuromics). Secondary antibodies (Jackson Immunoresearch) raised in donkey, conjugated with aminomethylcoumarin (AMCA), Rhodamine red, or Alexa Fluor 488 were used for detection. Samples were double stained for multiple proteins. Each colony pair was imaged in sections at a 10× magnification using an inverted fluorescent microscope (Axio Observer, Zeiss) equipped with a high-resolution camera (AxioCam MRm, Zeiss). Sections were merged using Photoshop (Adobe) to generate a panorama view of each colony pair. The resulting images were used to quantify the expression of the proteins.

The exposure time during imaging was kept constant for each secondary antibody. To quantify the expression of a neural marker, a selection covering the entire differentiated area of the immunostained colony pair was made and total fluorescent intensity of the area was measured in ImageJ. Five random locations outside the region containing the main signal were selected to measure the background fluorescent. Net fluorescent intensity was then calculated for each colony after subtracting the background using the following relation:

$$\text{Net}\text{fluorescent}\text{intensity}=0.5\times [\text{integrated}\text{density}\text{of}\text{the}\text{selected}\text{area}-(\text{mean}\text{background}\text{fluorescence}\times \text{area}\text{selected})]$$

The net fluorescent intensity of each marker protein from each colony was normalized against the circumference of the respective colony for unbiased comparison (Joshi, Thakuri, et al., 2017).

Additionally, differentiation of mESC colonies to neural stem cells, neuronal progenitor cells, and glial cells was quantified using adaptive thresholding in ImageJ. Each image containing a colony pair was converted to a binary image. In TuJ-stained images, neurites density of the pair of colonies, defined as the total white pixels count that considered both length and thickness of neurites, was measured. The colony bodies were cropped out from each image after thresholding and following our established protocol (Joshi, Thakuri, et al., 2017). The total neurites density was measured from the portion of the image containing only neural processes, divided by two to account for the pair of colonies, and normalized by the respective perimeters of the pairs of colonies to give the neurites density. In Nestin-stained and GFAP-stained images, similar method of thresholding, colony cropping, and normalization was used to estimate positively stained neural stem cells and glial cells, respectively.

Finite Element Modeling

Diffusion of the endogenous soluble factors secreted by the cells in colony pairs was evaluated by finite element modeling using the following three-dimensional diffusion equation:

$$\frac{\partial c}{\partial t}=D{\mathrm{\Delta }}^{2}c$$

Here, c is the concentration of soluble factors and D ≈ 1.5×10⁻¹⁰ m²/s is the diffusion coefficient (Hui & Bhatia, 2007). From measurements with experimental images of mESC colonies, each colony was approximated as a circular region of 1 mm diameter. Colony pairs with 3 interspacing distances of 20 μm, 660 μm, and 1400 μm were modeled. The simulation domain comprised of a physics controlled extremely fine mesh in shape of a cylinder of 35mm diameter and 2mm height, and a no flux boundary condition to represent dimensions of a Petri dish. All simulations were run for 1 hour which was the minimum time duration between two consecutive media renewals during 2-week culture period. The concentration profile was modeled across the bottom surface where the colonies were located. The distribution of secreted factors from the colony pairs in multiple interspacing configurations was compared.

Gene Expression Analysis

Experimental samples were lysed using a TRK lysis buffer every other day for two weeks (Omega Biotek). The lysates were homogenized by passing through homogenizer mini columns (Omega Biotek). Total RNA was isolated from the samples using an RNA isolation kit (Omega Biotek). DNase was removed using RNase-free DNase kit (Omega Biotek). Purity and concentration of isolated RNA was determined using OD 260/280 spectrophotometry (Synergy H1M, Biotek instruments). cDNA was synthesized from 1 μg of total RNA using Transcriptor First Strand cDNA Synthesis kit (Roche) following the manufacturer’s instructions.

Real time q-PCR was performed with a LightCycler 480 II instrument using a SYBR Green Master Mix (Roche). Each PCR reaction comprised of 50 ng (i.e., 1 μl) of the synthesized cDNA, combined with 1 μl from each of forward and reverse primers and 12μl SYBR green Master Mix to yield a final reaction volume of 15 μl. The reactions were pre-incubated at 95°C for 5 min followed by 45 cycles of amplification, i.e., at 95°C for 10 sec, at 60°C for 10 sec, and at 72°C for 10 sec. For each sample, the expression levels of 28 genes that included reference genes, markers of pluripotency, neural progenitors, specific neuronal and glial cells, and mesodermal lineage cells were measured. Table S1 lists the genes and primers sequences. Expression levels of mRNA for different marker genes were calculated relative to reference genes, GAPDH and β-Actin, using the ΔΔC_t method. The fold change in mRNA expression was determined using the 2^−ΔΔCt method. The experiments were repeated twice. For each experiment, 8 samples were obtained every other day over the culture period. Therefore, for 3 colony interspacing experiments performed in duplicates, 2688 PCR reactions were performed (6 conditions × 8 days × 28 genes × 2 technical replicates).

Statistical Analysis

All experiments were performed in duplicates. For protein expression analysis using image processing, at least 18 pairs of colonies were imaged and analyzed for each colony interspacing. Statistical tests were performed using one-way ANOVA and Fisher’s post hoc test in Minitab 16 software. Statistical significance was defined at p<0.05.

For the statistical analysis of gene expression data, the fold change values of each gene obtained every other day were normalized to the highest fold change value of that gene among the samples from three different colony interspacing distances. The sole purpose of this approach was to scale the gene expression values to a fixed range of 0–1. Then, one-way ANOVA with Fisher’s post hoc test was performed on the normalized fold change values.

RESULTS AND DISCUSSION

Stem Cells Colony Formation and Differentiation

We loaded the DEX phase solution containing mESCs into 30 nl slot pins mounted on the pipetting head. To facilitate the microprinting of mESCs onto the stromal cells, we robotically lowered the pins into the PEG phase covering the PA6 cells monolayer. This resulted in the dispensing of the DEX phase solution and formation of drops on the PA6 cells layer. The ultralow interfacial tension of ~12 μJ/m² between aqueous PEG and DEX phases was insufficient to hold the DEX phase solution in the pins against the gravitational pull that dispensed the content of the pins (Atefi, Mann, & Tavana, 2014). The resulting DEX phase drops containing mESCs were 285 ± 54 μm in average diameter. mESCs remained partitioned to the DEX phase drop (Fig 1e) and adhered to the stromal cells layer within three hours of printing. The two-phase printing approach did not exert any thermal, mechanical, or chemical stresses on cells. We previously showed that microprinted mESCs remain pluripotent (Joshi, Thakuri, et al., 2017). mESCs rapidly proliferated to generate isolated colonies while the mitotically-arrested PA6 cells remained viable and intact as a monolayer (Fig 1f–g). These colonies contained multiple layers of cells. However, the colonies were not spherical like embryoid bodies (EBs) and contained areas of varied cell density as evident from differences in the contrast within colonies. Printing 100 mESCs in the DEX phase drops yielded individual mESC colonies with an average diameter of 1.00±0.05 mm by day 8.

To print two interspaced DEX phase drops, we used the displacement of the pin tool defined in terms of percentage of radius of a well of a standard 96-well plate. Varying the interspacing by offsetting of 40% to 80% of a well radius resulted in the drops with center-to-center distances of 1.270 mm to 2.540 mm. The edge-to-edge distance between the two neighboring colonies varied linearly with the center-to-center distance between two drops (R²=0.986) as shown in Fig 1h. To study the effect of colony interspacing on neural differentiation of mESCs, we used three offset percentages of 40%, 60%, and 80% that gave the edge-to-edge colony interspacing distances of 0–40 μm, 540–790 μm, and 1300–1500μm, respectively, on day 8 of culture. For the ease of discussion, we will use the average of each range and express it in terms of diameter of a single colony, i.e., 0.02D, 0.66D, and 1.40D, respectively

The size of the mESC colonies and thus, the interspacing between each two neighboring colonies, did not significantly change after the first week of culture. Proliferation of pluripotent mESCs and resulting multipotent neural stem and progenitor cells causes growth and formation of the colonies during the first week of culture. Cells subsequently acquire less proliferative neuronal or glial progenitor states, or terminally differentiate to neurons, astrocytes, or oligodendrocytes during the second week (Joshi, Buchanan, Paruchuri, Morris, & Tavana, 2016; Kawasaki et al., 2000). Our temporal gene expression study showed transition in the state of mESCs and supported this explanation (Fig S1–S3). Over the culture period, mESCs in the colonies differentiated to neural cells and extended out dense neurite processes from the colonies. Extensive neural processes formed between each two adjacent colonies (Fig 1i–j). It is important to note that mESCs-PA6 cells signaling induced neural cell differentiation without a need for additional chemicals.

Colony Interspacing Effect on the Expression of Neural Cell Proteins

Our previous studies showed that expression of specific markers of neural stem and progenitor cells, such as Nestin and TuJ, and neurite processes are detectable in mESCs colonies starting from the fourth day of co-culture with PA6 cells (Joshi et al., 2016). The expression of neural markers gradually increased until day 8 and plateaued thereafter. Therefore, we quantified expression of neural stem cell marker Nestin and post-mitotic neuronal marker TuJ using immunocytochemistry analysis of samples from day 8, and additionally using image processing techniques. Specific neuronal and glial cell markers, such as TH and GFAP, were detectable from day 8 of culture and showed increased expression until day 14. Therefore, we quantified expression of specific neuronal and glial cell markers from day 14 samples.

TuJ (Beta III Tubulin)

Fig 2a–c shows TuJ-stained colony pairs at three different interspaced distances on day 8. The colony pairs at 0.66D had considerably longer and denser neurite processes extending out from the colony periphery than those at 0.02D and 1.40D (Fig 2a–c). A magnified view of the selected area between the two colonies from Fig 2b shows the formation of neurite bundles between the colonies (Fig 2d). To objectively compare differentiation of cells in colony pairs at different interspacing, we subtracted the average background fluorescence from the total fluorescent intensity measured from each colony pair, and normalized the resulting net fluorescent intensity to the total perimeter of the pair of colonies. TuJ levels were significantly higher in the colonies interspaced at 0.66D than in the colonies interspaced at 0.02D and 1.40D (p < 0.01) (Fig 2e). Cells in the 0.02D configuration also showed a significantly greater TuJ expression than those in the 1.40D configuration (p < 0.01).

Figure 2.

TuJ expression of interspaced mESC colonies. (a–c) Immunostained images of TuJ-positive colony pairs at three different interspacing distances on day 8 of culture. (d) Magnified view of the boxed portion of panel (b) showing the neural processes. (e) TuJ fluorescent intensity normalized to the total perimeter of both colonies shows the highest intensity at the interspacing of 0.66D, followed by the 0.02D and then 1.40D spacing. (f) Total neurites density normalized to the total colony perimeter shows results similar to those from the fluorescent intensity measurements in panel (e). * p < 0.01, n=18. Error bars represent mean ± S.E.M.

In addition to the fluorescent intensity measurements, we also analyzed TuJ expression using an adaptive thresholding method that accounted for the length and thickness of the neurite processes. Unlike the fluorescent intensity measurements, the adaptive thresholding method corrected for fluorescent intensity variations within an image. Results showed that neurites density in the 0.66D configuration was 1.9-folds and 2.9-folds higher than that in the 0.02D and 1.40D interspacing distances, respectively, (p < 0.01) (Fig 2f). To make a direct comparison of the thickness of the neurite bundles formed between the adjacent colonies in each configuration, we measured widths of 5 random neurite processes from the TuJ-stained fluorescent images. Overall, colonies interspaced at 0.66D had 4 times thicker neurite bundles formed between the colony pairs than the colonies at 0.02D and 1.4D configurations (p < 0.01) (Fig S4a).

Nestin

Nestin-positive neural stem cells distributed at the periphery of the colony pairs in all three conditions (Fig 3a–c). Closer scrutiny showed numerous Nestin-positive neural stem cells migrating out of the colonies, particularly in the areas between the two interspaced colonies (Fig 3d). Therefore, we normalized the measured fluorescent intensities to the combined perimeter of both colonies to quantitatively compare Nestin expression among all three groups. Nestin-positive cells were significantly more abundant in the 0.66D colony interspacing than in the 0.02D interspacing by 1.3 folds and the 1.4D interspacing by 1.8 folds (Fig 3e). We also quantified the relative ratio of Nestin-positive neural stem cells to TuJ-positive neuronal cells in each interspacing configuration using adaptive thresholding to determine if any configuration favored differentiation of neural stem cells to neuronal cells. The results showed that the colony interspacing did not affect the ratio of neural stem cell to differentiated neuronal cells. (Fig S4b)

Figure 3.

Nestin expression of interspaced mESC colonies. (a–c) Immunostained images of Nestin-positive colony pairs at three different interspacing distances on day 8 of culture. (d) Magnified view of a portion of panel (b) showing Nestin stained neural processes. (e) Nestin fluorescent intensity normalized to the total perimeter of both colonies shows the highest intensity when the two colonies are interspaced at 0.66D, followed by at 0.02D and then at 1.40D. * p < 0.01, n=18. Error bars represent mean ± S.E.M.

GFAP (glial fibrillary acidic protein)

In addition to the neural stem and neuronal progenitor cell markers, we immunostained the colonies for astrocytes using GFAP on the final day of culture to allow the differentiating cells to acquire the specific neuronal or glial traits (Fig 4a–d). We quantified GFAP expression through fluorescent intensity measurements and normalized the results with respect to total perimeter of pairs of colonies. In contrast to expression of TuJ and Nestin, GFAP expression consistently increased with increasing the interspacing distance between colony pairs. Cells in the 1.40D colony interspacing showed significantly higher GFAP expression by 1.2 folds and 2.8 folds than cells in the 0.66D and 0.02D interspacing distances, respectively (Fig 4e). To estimate neuronal to glial differentiation ratio in each colony interspacing configuration, we calculated the ratio of TuJ-positive and GFAP-positive pixels count in each colony pair. The results showed that interspacing of 0.02D and 0.66D promoted greater neuronal cell differentiation than glial cell differentiation (i.e., ratio of TuJ to GFAP> 2), whereas the interspacing of 1.40D favored glial cell differentiation (i.e., ratio of TuJ to GFAP< 1). The ratio of neuronal to glial cells differentiation decreased significantly with increasing colony interspacing (p <0.01). (Fig S4c) We emphasize that this graph merely shows the ratio of neuronal to glial cells in each interspacing configuration and does not compare the magnitude of neuronal or glial differentiation among different interspacing configurations.

Figure 4.

GFAP and TH expression of interspaced mESC colonies. (a–c) Immunostained images of GFAP-positive colony pairs at three different interspacing distances on day 14 of culture. (d) Magnified view of a portion of panel (b) showing GFAP stained astroglial progenitors. (e) GFAP fluorescent intensity normalized to the total perimeter of both colonies increases with increasing the colony interspacing. (f) An image of TH-positive cells at the periphery of a colony on day 14. (g) TH-positive cell count per colony shows the highest number at 0.66D interspacing, followed by at 0.02D and then at 1.40D. * p < 0.01, n=18. Error bars represent mean ± S.E.M.

TH (Tyrosine hydroxylase)

We quantified dopaminergic neuronal cells in the colonies using expression of TH per colony through counting and averaging the number of positively stained cells around and within the pairs of colonies. The results showed that the 0.66D interspaced colonies contained the largest number of TH-positive cells per colony, with 2.2-folds and 1.9-folds greater cell count than the 0.02D and 1.40D interspaced colonies, respectively (Fig 4f–g).

The above results showed that among the three different interspacing distances, the 0.66D configuration resulted in the highest expression of neural cell proteins in the differentiating cells while the 1.4D configuration provided a more favorable niche for glial cell differentiation. This could be indicating the dominance of neural inducing factors such as neurogenins in 0.66D interspacing that suppress the glial differentiation, and dominances of glial differentiation promoting and neural differentiation inhibiting factors like BMPs in 1.4D interspacing (Bond, Bhalala, & Kellser, 2012; Sun et al., 2001). Because of a similar number of stromal cells in all three groups, the effect of paracrine signaling from PA6 stromal cells on inducing neural differentiation is expected to be similar in all three groups. We previously showed that the self-regulatory secreted factors secreted by differentiating mESCs also impact their neural differentiation (Joshi, Buchanan, & Tavana, 2017a). Higher expression of neural proteins in a specific colony pair is most likely due to differences in the concentration, gradients of concentration, and distribution of secreted endogenous factors among the pair of colonies. Our previous work and the result presented above showed that differentiating cells primarily exist in the peripheral regions of the colonies. We hypothesized that differentiating cells in a colony sense the adjacent colony through secreted factors and respond by migrating along the concentration gradient and extending neurite processes. We conducted a finite element modeling diffusion of the secreted endogenous factors in the colony pair configuration to test this hypothesis.

Finite Element Modeling (FEM) of Distribution of Secreted Factors by the Colony Pairs

We used FEM to determine the distribution of secreted factors within the area of the Petri dish and compare the results among colony pairs with different interspaced distances. With nine mESC colony pairs generated in each 35 mm Petri dish, the minimum distance between any two colony pairs was always greater than 5.5 mm. Our preliminary results showed that the concentration of cell secreted factors drops to only 10% of the maximum concentration within 5 mm from the edge of each colony. Therefore, each colony pair was considered isolated and independent from other colony pairs in the dish.

The modeling results of concentration of diffusive factors within 1×1 cm² area are shown in Fig 5a–c. In all three configurations, the concentration was highest over the colonies and slightly toward the inner edges that face each other. Moving away from the outer edges, the concentration dropped and quickly diminished. Overlaying the concentration profiles and TuJ stained images showed abundance of neuronal extensions in regions of the concentration profiles highlighted with orange-to-yellow colors (Fig 5d–f). TuJ1 expression levels were the highest in areas between and around pair of colonies where a concentration gradient was present, as represented by orange-to yellow-hues. TuJ1 expression significantly reduced by decrease in the concentration of soluble factors.

Figure 5.

Finite element modeling of spatial distribution of mESCs-secreted soluble factors. (a–c) Concentration profiles of diffusive signaling factors of mESC colonies at different interspacing distances. (d–f) Overlay of magnified concentration maps and TuJ expression of cells in the colony pairs. (g) Line scans of concentration profiles for colony pairs along the dashed lines in panels (d–f).

Analysis of the concentration profiles along the line scan in Fig 5d–f showed that the maximum concentration for the 0.02D configuration occurred where the inner edges of the two colonies almost overlapped (Fig 5g). The concentration showed a rapid decrease away from this region. It is plausible that very close proximity of the two colonies and lack of sufficient space between them hindered formation of significant concentration gradients that would support extension of neurite processes. As such, there was minimal staining for neural cell proteins in this region, considerably less than that in the outer edges of the colonies where stem cells were exposed to gradients of soluble factors (Fig 5d, 2a, 3a, and 4a). This observation is consistent with a tropic hypothesis that the concentration gradient, and the direction and relative steepness of the gradient of secreted molecules are more important factors that determine neurites growth and neural cell proliferation, than the absolute concentration of the molecules (Duncan Mortimer et al., 2010). On the other hand, colony pairs at the interspaced distances of 0.66D and 1.40D exhibited lower maximum concentrations than the 0.02D group, but they resulted in concentration profiles consisting of a gradient in the region between the two colonies (Fig 5g). The concentration profile showed a larger slope for the colonies interspaced at 1.40D (slope: 914) than the colonies interspaced at 0.66D (slope: 863). Closer scrutiny of Fig 5e–f showed that differentiated cells extended neurites more toward and around the inner edges where the two colonies face each other, and a local concentration gradient exists. Higher yield of neural cells in the 0.66D configuration suggests that a shallow concentration gradient with a larger minimum concentration in the space between the two colonies provides a more favorable condition for neural differentiation. Other studies also associated shallow concentration gradients with elevated growth rate of neural population and steep concentration gradients with turning of the axonal tip to determine its direction (D. Mortimer et al., 2009; Duncan Mortimer et al., 2010). This model therefore provides a qualitative explanation that short range concentration gradients of endogenous soluble factors are in part responsible for the differences in neural differentiation observed in colony pairs of various interspaced distances.

Several studies have identified specific factors such as sonic hedgehog (Shh), nerve growth factor (NGF), and neurotrophins that increase neuronal cell count and promote neurites outgrowth in a concentration gradient-dependent manner (Moore, Macsween, & Shoichet, 2006; Xu et al., 2013). The actual concentrations of the soluble factors and their mechanisms of action in these co-culture niches remain mostly unknown. We recently identified several self-regulatory mESCs-secreted soluble factors in our microengineered co-culture system that promote neural cell differentiation of mESCs (Joshi, Buchanan, & Tavana, 2017b). Similar to the neurogenic soluble factors reported in the literature, we identified several factors such as neurotrophins, VEGFs, Shh, Pax1, and Wnts that augmented neural differentiation of mESCs. Interestingly, we found that these soluble factors were secreted by differentiating stem cells and not by the co-cultured stromal cells. Therefore, we suggest that concentration gradients of these mESCs-secreted soluble factors in different interspacing configurations gave rise to the corresponding differences in gene and protein levels in this study.

Our results and explanation are supported by previous studies on effects of concentration gradients of soluble factors on stem cells response and differentiation to neural cells (Tavana, Mosadegh, Zamankhan, Grotberg, & Takayama, 2011). In a microfluidic bioreactor, mESCs subjected to gradient of nutrients differentiated to Nestin positive cells within three days, whereas lack of gradient prevented differentiation and cells remained pluripotent (Khoury et al., 2010). In another study, it was shown that a recombinant human IgM (rHIgM12) that was presented with a concentration gradient guided the neurites navigation of hippocampal neurons. Neurites of spinal neurons followed patterns of rHIgM12 printed on a substrate and elongated to form a physical network (Xu et al., 2013). Neuronal cell count and the length of neurite extensions were directly proportional to the concentration gradient of sonic hedgehog (Shh) (Xu et al., 2013). Concentration gradient of immobilized nerve growth factor (NGF) guided the neurites outgrowth. And a lower concentration gradient of immobilized NGF combined a concentration gradient of neurotrophin-3 (NT-3) acted synergistically to guide neurites growth due to co-localization of their respective receptors on the neurons (Moore et al., 2006). Similarly, synergism between factors Shh and Netrin-1 amplified the sensitivity to the guidance cues by eliciting a growth response from axons at shallow concentration gradients (Sloan, Qasaimeh, Juncker, Yam, & Charron, 2015). Steeper concentration gradients were otherwise required to elicit any axonal response when either molecules were presented individually.

Gene Level Regulation of Neural Differentiation of mESCs by Colony Interspacing

To understand molecular level effects of interspacing mESC colonies on neural differentiation, we complemented the above protein expression and FEM studies with a comprehensive temporal expression analysis of 26 prominent genes implicated in differentiation of ESCs (Table S1). This selection was based on a comprehensive literature survey and included genes representing pluripotent stem cells, neural stem/progenitor cells, specific neuronal and glial cells, and several mesodermal markers.

Temporal gene expression profiles of cells in different colony interspacing groups

The temporal trajectories of expression of all 26 genes were consistent with our previous studies of mESCs – PA6 cells co-culture (Joshi et al., 2016), and similar in all three colony interspacing conditions (Fig S1–S3, S5). Expression of Oct4, Nanog, Notch 1, Notch 2, Notch 3, and Wnt8a that mark the undifferentiated mESCs steadily decreased over time (Fig S1). Oct4, Nanog, and Wnt8a promote self-renewal and proliferation of pluripotent stem cells and rapidly downregulate at the onset of differentiation (Jaenisch & Young, 2008; Lindsley, Gill, Kyba, Murphy, & Murphy, 2006). The three Notch genes inhibit differentiation of stem cells until correct cues become available (Gaiano & Fishell, 2002; Woo et al., 2009). Moreover, Notch proteins restrain the expression of neural genes and promote differentiation to glial subtypes.³⁹ Low levels of the Notch genes over time suggest that this co-culture system favors neurogenesis over gliogenesis. We note that all six genes quickly downregulated with minimal differences among the three colony interspacing conditions. Therefore, we did not consider these genes for further statistical analysis. Concurrent with the decreased expression of the pluripotency genes, early stage neuroectodermal and neural stem cell genes Wnt1, Sox1, CDH2, Nestin, and Pax6 showed increasing expression. These genes showed tens to hundreds of folds change around days 6–10 of culture (Fig S2a–e). Subsequent decline in the expression of neural stem cell markers coincided with a steep rise in the expression of neuronal progenitor cell markers NCAM and TuJ (Fig S2 f–g), as well as all nine specific markers of neuronal and glial cells including GAP43, TH, GFAP, SYNAPTOPHYSIN, NEUN, MAP2, CHAT, OLIG1, and GAD1 (Fig S3).

Among the three colony interspacing groups, the neural genes had the highest mRNA fold change over time in the 0.66D interspacing, followed by in the 0.02D and then 1.40D groups (Fig S2–S3). Major conclusions of the gene expression results are as follows: (i) Temporal dynamics of neural differentiation of mESCs was independent of interspacing and the expression levels of gene markers of neural stem cells and specific neuronal and glial cells reached a maximum around the same time in all three colony interspacing groups. (ii) The difference in the expression levels of neural stem cell genes among the different configurations varied significantly over time, but the maximum difference always occurred when the expression of the genes reached a maximum. (iii) The expression of neural genes was the highest when the colony pairs were interspaced at 0.66D.

To determine the specificity of neural differentiation of mESCs and potential differences among the different colony interspacing groups, we performed q-PCR analysis for several markers of mesodermal cells (Fig S5). There was a negligible fold change of less than one in the expression of mesodermal genes NKX 2.5 and PECAM. The expression of GATA4 and FLK1 was slightly but significantly elevated during the culture. The 0.66D configuration showed the lowest expression of the mesodermal markers, indicating that it was the most permissive condition for neural differentiation of mESCs in the microprinted colony pairs.

Statistical analysis of differences in temporal gene expression profiles

Next, we performed an intermediate step of data normalization to enable statistical analysis of the gene expression data using conventional statistical tools. We normalized the fold change of each gene at each time point from all three interspacing groups to the highest fold change of that gene, scaling the expression levels between 0–1. Then, we subjected the normalized fold change data for each gene to one-way ANOVA followed by Fisher’s post hoc test to determine whether the gene was differentially expressed among the three colony interspacing groups. Fig 6 represents the average normalized fold change values of the neural genes for all three colony pairs, and the results of the statistical analysis.

Figure 6.

Average normalized mRNA fold change of neural genes from the two-week culture in three different colony interspacing distances. * p < 0.05. Error bars represent mean ± S.E.M.

The gene expression differences among the three colony interspacing groups showed a close correlation with the corresponding protein expression differences results. Neural stem and progenitor cell markers such as Nestin and TuJ had the highest expression in the 0.66D configuration, both at gene and protein levels. A correlation analysis between gene and protein expression results among different colony interspacing on day 8 resulted in a Pearson correlation of 0.996 (p-value: 0.05) for Nestin, and Pearson correlation of 0.998 (p-value: 0.04) for TuJ. Nestin is a neural stem cell marker that regulates self-renewal and axonal growth of the cells (D. Park et al., 2010). The highest expression of Nestin gene in the 0.66D configuration resulted in phenotypic changes such as longer and denser radial extension of neurites from differentiating cells in this colony pair group (Fig 6a, 2f, 3e). The greater NCAM expression in the 0.66D group than the other two groups (Fig 6b) suggests that this colony interspacing best promotes neural differentiation and maturation. The 0.66D group also showed the highest expression levels for TuJ, Sox1, CDH2, Pax6, and Wnt1 genes, which promote proliferation of neural progenitors and maturation of neuronal circuits by contributing to axonal extension (Osumi, Shinohara, Numayama-Tsuruta, & Maekawa, 2008; Solozobova et al., 2012; Venere et al., 2012). These genes also regulate neurites outgrowth and help induce sensory and midbrain dopaminergic neurons (Chen et al., 2013; Kasai, Satoh, & Akiyama, 2005). Higher fold change values of these genes in the 0.66D group was consistent with the higher TuJ protein expression, longer neurites length, and higher number of TH positive cells in this group.

In addition to higher expression levels of the genes representing neural stem and progenitor cells in the 0.66D group, cells in this configuration showed greater expression of genes implicated in neuronal growth and maturation such as GAP43, NeuN, and MAP2, which help extend and stabilize neurite processes in post-mitotic neurons. Increased expression of Synaptophysin implies that the neuronal cells from the 0.66D group are more developed, functionally active, capable of generating synapses, and potentially transmitting signals (Ding et al., 2002). Higher expression of specific markers of functional or terminally differentiated neural cells such as dopaminergic neuronal marker, TH, GABAergic/glutamatergic neuronal marker, GAD1, cholinergic neuronal marker, ChAT, astrocytic marker, GFAP, and neuronal markers NeuN and GAP43 implies a greater number of functional neurons generated in the colonies of the 0.66D group. We note that while our results of protein expression for TuJ, Nestin, and TH were consistent with the corresponding gene expression results, the overall GFAP protein expression was higher in the 1.40D group than the other two configurations. Several factors such as mRNA degradation, translation time, translational efficiency, and protein stability can potentially limit the correlation between gene and protein expressions to as low as 40% (Edfors et al., 2016; Schwanhäusser et al., 2011). Close scrutiny and analysis of the result in Fig S3b shows that the fold change values of the GFAP gene were similar in the 0.66D (557 folds) and 1.40D (534 folds) configurations up to and including day 12. However, GFAP gene expression escalated to several thousands fold change on day 14 with a significant difference between the 0.66D (2513 folds) and 1.40D (1757 folds) configurations. Because we quantified the protein expression on day 14, it is highly likely that the large difference in the gene expression did not manifest at the protein level on the day of analysis. Longer-term experiments for GFAP protein expression would be required to validate this point.

Significant impact of spatial organization of stem cell colonies on neural differentiation

In a previous study, we investigated the effect of single colony size on neural differentiation of mESCs microprinted on PA6 stromal cells (Joshi, Thakuri, et al., 2017). Our results showed that increasing the size of mESCs colony disproportionately enhanced differentiation to neural stem/progenitor cells and neuronal and glial cells. The smallest colony generated from 100 mESCs had the lowest differentiation level confirmed through gene and protein expression studies. In our current study, we still generated the mESC colonies from 100 cells to determine if interspacing two colonies can enhance the differentiation efficiency due to inter-colony signaling. We compared TuJ and Nestin protein expression normalized per colony between an isolated colony from the previous work and interspaced colony pairs from this study. Results showed that the interspaced colonies show significantly higher protein expression regardless of the interspacing distance. The expression levels of TuJ and Nestin were, respectively, up to ten folds and four folds higher in the interspaced groups (Fig S6a–d, p< 0.01). Neurites density per colony was five times higher in the 0.66D colony pair than in the single colony of the same size (Fig S6e–f, p< 0.01). The number of TH-positive cells counted per colony was more than two times higher in colony pairs than that counted in the single colony (Fig S6g–h, p< 0.01). This is consistent with the computational modeling result that gradients of soluble factors have a major effect on enhancing neural differentiation of mESCs.

Conclusions

We used an aqueous two-phase system cell microprinting technology to generate pairs of isolated mESC colonies with defined interspacing distances on a layer of stromal cells. Using comprehensive gene and protein expression studies, we established that the spatial organization of stem cell colonies has a major effect neural differentiation of mESCs. Our molecular studies showed that the colonies at the 0.66D interspacing had the highest differentiation of mESCs to neural cells. Colony pairs with a smaller of larger interspacing than 0.66D showed reduced neural differentiation. Our computational modeling revealed that modulating the interspacing of the colonies changes the concentration profile of soluble signaling molecules between the colonies. The results suggested that presence of stable concentration gradients of the soluble factors has a significant impact on enhancing the efficiency of neural differentiation of stem cells in interspaced colonies. Overall, our study offers a mechanistic understanding of the role of a biophysical factor of stem cell microenvironment, i.e., colony interspacing, on generation of neural cells from mESCs. This study provides an approach to engineer an efficient stem cell microenvironment to obtain differentiated neural cells that may be used in translational applications for regenerative therapies of neurodegeneration, or to generate biomimetic neurodevelopmental models for in vitro disease modeling or drug screening.

Supplementary Material

Supp info

Supplementary Figure S1. mRNA fold change values of pluripotency marker genes and regulatory factors during the two-week culture. n=2. Error bars represent mean ± S.E.M.

Supplementary Figure S2. mRNA fold change values of neural stem and progenitor cell genes during the two-week culture. n=2. Error bars represent mean ± S.E.M.

Supplementary Figure S3. mRNA fold change values of specific neuronal and glial cell genes during the two-week culture period. n=2. Error bars represent mean ± S.E.M.

Supplementary Figure S4. (a) Comparison of thickness of neurite bundles formed between the colony pairs at different interspacing distances. * p <0.01, n=18. (b) Ratio of neural stem cells to neuronal cells within each colony pair at different interspacing distances. (c) Ratio of neuronal cells to glial cells within each colony pair at different interspacing distances. Error bars represent mean ± S.E.M.

Supplementary Figure S5. mRNA fold change values of mesodermal markers during the two-week culture. n= 2. Error bars represent mean ± S.E.M.

Supplementary Figure S6. (a–b) Comparison of normalized TuJ fluorescent intensity in colony pairs at different interspacing distances (blue) with that of single colony of the same size (green). (c–d) Comparison of normalized Nestin fluorescent intensity in colony pairs at different interspacing distances (blue) with that of single colony of the same size (green). (e–f) Comparison of normalized neurite density in colony pairs at different interspacing distances (blue) with that of single colony of the same size (green). (g–h) Comparison of TH⁺ cell count in colony pairs at different interspacing distances (blue) with the cell count in single colony of the same size (green). * p <0.01. n=18. Error bars represent mean ± S.E.M.

Supplementary Table 1. List and sequence of primers for the genes analyzed