Organ-specific ECM arrays for investigating cell-ECM interactions during stem cell differentiation

We used a perfusion decellularization protocol reported previously [29], to obtain organ-specific decellularized pancreas, liver and heart scaffolds explanted from female ICR mice (Taconic, ages of 6–12 weeks) after systemic heparinization. The decellularized organs were first lyophilized in Virtis Benchtop K freeze dryer, operating at 60 mTorr for 3 d. After lyophilization step, 30–50 mg dry weights of the lyophilized tissues from each organ were pulverized into ECM powder (figure 1(A(ii))). This was followed by treatment with 50 µg ml⁻¹ DNase (Roche) for 10 min and washed three times with PBS. For ECM extraction, each decellularized tissue were digested in high molar of chaotropic agents (4 M guanidine HCl, 8 M Urea in PBS) overnight with gentle shaking at 4 °C. Samples were be centrifuged at 13 000 g for 10 min at 4 °C to remove insoluble material and dialyzed on floating V-series membranes (Millipore) against PBS for one hour [45]. The final ECM solution was filter-sterilized through PVDF syringe filter units (0.22 mm, Millipore), and then stored at 4 °C until further use.

Zoom In Zoom Out Reset image size

Patterning substrates were prepared by first coating glass slides or tissue culture plates with nitrocellulose [46]. Static water contact angle measurements were performed to assess the hydrophobicity of the coated surface. Matrigel was used as a control ECM to pattern the surface and test protein immobilization efficiency. Matrigel was applied to nitrocellulose-coated surface in droplets of 1 µl–3 µl volume containing 100 µg ml⁻¹ concentration. After about 1 min, droplets were removed by aspiration and the substrate surface was blocked by washing twice with 1% bovine serum albumin (BSA)/PBS (Sigma Aldrich). To verify the efficiency of ECM protein patterning on nitrocellulose-coated surface, Lectin peanut agglutinin (PNA)—Alexa 647 (Invitrogen, Molecular Probes) was used to detect the immobilized ECM protein. PNA lectin stain was used at a concentration of 200 µg ml⁻¹ in PBS and followed by three PBS washes. Microspots' diameter were measured using slide scanning images recorded with Metamorph 7.5.6.0 (Molecular Device) software on an Olympus IX81 inverted microscope. The fluorescence intensities (RFU) of the PNA lectin stain was quantified with LI-COR odyssey scanner using the 700 nm wavelength channel.

After ECM extraction, individual organ's ECM was assessed for protein and ECM content as follows. For protein quantification, bicinchoninic acid (BCA) assay (Thermofisher) was done (n = 3 for each organ) according to manufacturer's instruction. For ECM content, organ derived ECM extracts (n = 3 for each organ) were analyzed with the Sircol Collagen and Blyscan GAG assay kits (Biocolor Life Sciences Assays, Carrickfergus, UK) according to the manufacturer's instructions in order to quantify the amount of each component present. The final content was normalized to the starting dry weight of each lyophilized decellularized organ.

For SDS-PAGE analysis, ECM samples were loaded at 4 µg per well. 20 µl of sample was mixed with 20ul of 2X NuPAGE LDS sample buffer +5% β-mercaptoethanol and heated at 98 °C for 5 min. This resulting mixture was loaded onto a NuPAGE 3%–8% Tris-Acetate Gel in 1X Tris-Acetate Running Buffer for 5 h at 60 V constant. Gel was stained using Bio-Rad Bio-Safe Coomasie stain for 4 h, and destained overnight in distilled water. The resulting gel was imaged using The Bio-Rad ChemiDoc XRS + system.

Each organ-derived ECM extracts were diluted into 100 µg ml⁻¹ and 1 µl of ECM solution per droplet were spotted with a repeater pipette (Eppendorf) onto the nitrocellulose-coated glass slides or tissue culture plates. After about 1 min, droplets of ECM solutions were aspirated and blocked by washing twice with 1% BSA/PBS. Five microspots were generated in square arrays for a well in a 6 well plate, and six by four droplets were generated in rectangular arrays for a glass slide. After fabrication steps, the arrays can be stored in blocking buffer at 4 °C until further use. All arrays used in this study were used within 3 months after fabrication.

For surface roughness measurement, AFM imaging of the ECM protein immobilized surface (MFP3D, Asylum Research) was performed in PBS at room temperature at a scan speed of 1 Hz with a resolution of 256 × 256 pixels. The scan size was 10 micron × 10 micron regions. Surfaces were scanned with a silicon nitride conical tip of k = 0.8 N m⁻¹( Veeco, Ltd), and root mean square (RMS) surface roughness analysis was performed on four regions per sample [47].

For stiffness measurement, 3D ECM array containing organ-derived alginate microspots were prepared and placed in room temperature PBS for AFM measurements. All AFM force-stiffness measurements were taken using silicon nitride cantilevers (k = 0.2 N m⁻¹) attached to silica microspheres (r = 3400 nm). Force measurements were taken at 16 locations over a 4 × 4 grid and two microspot samples were measured per condition. The resulting force curves were analyzed within the AFM-MFP3D software using Hertzian-fit for indentation depth of <10 microns [48].

The H1 hESC lines used in these experiments were obtained from the University of Pittsburgh Stem Cell Core, and are part of the NIH hESC registry eligible for NIH funding. Undifferentiated H1 hESCs were maintained in feeder-free culture in mTeSR medium (StemCell Technologies) on hESC-qualified Matrigel (BD Biosciences)–coated tissue culture plates. Cultures were fed every day with mTeSR medium and passaged with Accutase (StemCell Technologies) at 70% confluency. We have previously reported differentiation of hESC to pancreatic progenitor cells [49–51]. Briefly, first step of pancreatic differentiation is definitive endoderm induction by using 100 ng ml⁻¹ ActivinA (R&D Systems) with 25 ng ml⁻¹ Wnt3A (R&D Systems) for 4 d. This is followed by pancreatic progenitor induction with 0.2 μM KAAD-cyclopamine for 2 d (R&D Systems) and 0.2 μM KAAD-cyclopamine with 2 μM retinoic acid (Sigma-Aldrich) for 2 d. For generation of PP aggregates, an alginate encapsulation differentiation platform was utilized as previously described [49]. Encapsulated cells were grown into small aggregates and treated with the same induction protocol to differentiate into pancreatic progenitor stage. A basal media of DMEM/F12 plus 0.2% BSA and B27 supplement was used for the differentiation. Cultures were performed with p55–p70 hESCs in 37 °C incubator, 5% CO2, and 100% humidity.

Similar to regular immunofluorescence assay, attached 2D cells or encapsulated cells on 3D ECM arrays were first fixed for 15 min with 4% formaldehyde (Thermo Scientific). For 3D ECM array staining, we used dH₂O supplemented with 10 mM BaCl₂ to replace PBS for all the washing steps because cation in alginate could be dissolved by displacement of other stronger ions in PBS. For 2D ECM array staining with ECM proteins, ImmEdge Hydrophobic Barrier Pen (Vector Laboratories) was used to confine the antibody solutions to staining specific microspot of interest. The blocking buffer was also supplemented with 10 mM BaCl₂ when staining on 3D ECM arrays. Following fixing, samples were permeabilized with 0.1% TritonX100 for 15 min and blocked with Odyssey Blocking buffer (LI-COR Biosciences) for 1 h. For primary antibodies, we used the following antibodies: goat anti-PDX1 (1:400, R&D Systems), goat anti-NKX6.1, rabbit anti-Cpeptide (1:200, Cell signaling). Antibodies were diluted to half of the suggested immunofluorescence concentration as the combination of robust Near-IR dyes and sensitive LICOR scanner can provide sufficient to reliably detect signals. Incubation time for primary antibodies was overnight in 4 °C. Slides were washed three times (5–10 min) with PBS before secondary antibody staining. Infrared anti-goat and anti-rabbit IRDye800CW secondary antibody (1: 800, LI-COR Biosciences) and DRAQ5 (1: 10 000, Invitrogen) in blocking were then added to the arrays and incubated for 1 h at room temperature. This was followed by PBS washing for three times and incubated in PBS. The plates were imaged on an Odyssey infrared scanner using slide or plate settings with sensitivity of 5 in both the 700 and 800 nm wavelength channels. Data were acquired by using Odyssey image studio software, exported and analyzed in Excel (Microsoft). Primary antibody stained values were background subtracted from microspots treated only with secondary antibody, and then normalized to cell numbers by dividing by the total DNA (DraQ5) fluorescence signal.

Cell adhesion assays were carried out on 2D ECM array on glass slide or tissue culture plastic. If glass slides were used, we used chamber glass slides (Nunc™ Lab-Tek™) for the purpose of cultivation here. hESC-derived PP cells were treated with 10 μM of the rock inhibitor (RI) Y-27632 for 4 h before dissociation into single cells using accutase. Single-cell suspension of hESC derived-PP cells in serum-free media (DMEM/F12 with B27) were plated into the patterned ECM microspots. For short term attachment, cell suspension was incubated for 2 h in serum-free medium to allow for cell attachment (shaking the plates every 15 min to redistribute the cells). The arrays were then gently aspirated to remove unattached cells and fixed for cell staining and counting. For long term attachment, after first 2 h of cell attachment and removal of unattached cells, fresh culture medium with serum was added and the cells were allowed to attach overnight.

Similar to the protocol described by Fernandes et al[ 17], poly-L-lysine (PLL) (0.01% w/v) (Sigma Aldrich) and BaCl₂( 0.1 M) were mixed in 1:2 volume ratio and spotted on the nitrocellulose coated surface using repeater pipette (Eppendorf). A mixture of 1.1% (w/v) low-viscosity alginate (Sigma-Aldrich), ECM extracts and hESC-PP cell suspension (2.5 × 10⁶ cells ml⁻¹) in DMEM/F12 media was prepared and micro-spotted on top of the BaCl₂/PLL bottom layer using the repeater pipette.

For hESC-derived PP aggregates differentiation in 3D ECM array, the PP aggregates were first generated using the alginate differentiation platform described above. Alginate differentiated PP aggregates were treated with 10 μm of the rock inhibitor (RI) Y-27632 for 4 h before decapsulation with 100 mM EDTA. This was followed by passing the PP aggregates through 100 µm mesh-sized cell strainer to avoid larger aggregates being included in the microspots. Approximately 8000–12 000 PP aggregates were mixed in 1 ml of alginate and ECM extract solutions to micro-spot on top of the BaCl₂/PLL bottom layer using the repeater pipette. Subsequently, the cell-laden 3D ECM array was cultured in base maturation medium consists of DMEM/F12 plus 0.2% BSA and B27 supplement with nicotinamide (10 μM, Sigma Aldrich) and 10 μm of the rock inhibitor (RI) Y-27632 for 2 d. After that, fresh maturation media with notch inhibition (30 μM DAPT, Santa Cruz) was added and cultured for another 7 d. The medium was changed daily for a total for 9 d culture period.

Cell viability on the 2D and 3D ECM array was assessed via a Live/Dead assay (Invitrogen). Briefly, calcein AM (1 µM) and EthD-1 (2 µM) were added together with culture medium to the cell seeded arrays and incubated for 15 min under light protection in room temperature. Samples were washed twice with PBS and continued with epifluorescence microscopy to distinguish between live cells (green) and dead cells (red). Quantification of live/dead cells was accomplished by conversing all individual (red and green) channel to grayscale and followed by analysis with Metamorph Image software.

The immunofluorescence staining protocol is identical to the on-chip immunofluorescence protocol described above except for the following. For blocking buffer, 10% donkey serum (Jackson Immuno) was used. Anti-rabbit 488 and 555 Alexafluor secondary antibodies (1:500, Invitrogen) were used. The only primary antibody used here is Cpeptide (1:100, Cell signaling). Z–stacks images were acquired using a Nikon A1 laser confocal system (5 µm steps).

Microscope images of each array were acquired at 4x using an Olympus IX81 microscope. Images were recorded using Metamorph 7.5.6.0 (Molecular Device) software. The arrays were imaged using phase-contrast, DAPI, GFP, Cy3 and Cy5 optics. Scan slide module from Metamorph software were utilized and array images were montaged. For quantification purposes except for live/dead staining, the slides were acquired using Odyssey image studio software, exported and analyzed in Excel (Microsoft). Each array (on glass or tissue culture wells) were imaged using a focus height that gave the maximum signal for each wavelength channel (700 nm and 800 nm) at the center of the array.

Quantification data were expressed as mean ± SD. Significant differences among groups were determined by two-tailed Student's t-test for two-group comparisons or ANOVA followed by post-hoc analysis for multiple group comparisons. Probability values at P < 0.05 (*) and P < 0.01 (**) indicated statistical significance.

Figure 1(A) illustrates the overall workflow used to create an ECM array using decellularized organ ECM extracts. Organ-specific ECM scaffolds were first derived by perfusion-decellularization of whole organs [29] using an in-house bioreactor, as previously described. Keeping the decellularization protocol consistent, we successfully decellularized whole organ pancreas, liver and heart. Figure 1(A(i)) depicts the translucent acellular organ scaffolds generated after the complete removal of the cellular materials from the native organs. This is followed by lyophilization and pulverization steps to obtain organ-specific ECM powder (figure 1(A(ii))). We then extracted the ECM using high molar of chaotropic agents (8 M Urea and GuHCl), a commonly used protocol for ECM preparation [52–54]. The resulting extracts were dialyzed against PBS to yield pancreas-, liver- and heart-organ-specific ECM extracts (denoted as P-ECM, L-ECM and H-ECM respectively). We first characterized the organ-derived ECM extracts through sodium dodecyl sulphate (SDS)–PAGE for the presence of ECM proteins or other peptides (figure 1(B)). We loaded the same concentration of organ-derived ECM extracts (4 µg well⁻¹) and controls (collagen I and matrigel) on each well of the gel. SDS page analysis revealed that all three organ-derived ECM extracts contained a number of high molecular weight ECM proteins (∼250 KD) and low molecular weight (<100 KD) protein components that were absent in commercial bovine collagen type I and matrigel (figure 1(B)). There were also considerable differences in terms of the bands detected among the three organ-derived ECM extracts. Next, the organ-specific ECM extracts were analyzed for protein (BCA assay) and for ECM content such as collagen and glycosaminoglycans (GAGs). BCA assay revealed highest protein content in H-ECM (8.07 ± 4.7 µg mg⁻¹ dry weight), followed by L-ECM (6.90 ± 3.1 µg mg⁻¹ dry weight) and P-ECM (3.98 ± 1.5 µg mg⁻¹ dry weight) but without statistical significance among the three extracts (n = 3, P > 0.05; figure 1(C)). For the ECM content, sircol collagen assay demonstrated significant higher collagen content in H-ECM (0.78 ± 0.2 µg mg⁻¹ dry weight, n = 3) as compared to L-ECM (0.48 ± 0.05 µg mg⁻¹ dry weight, P < 0.05; figure 1(D)) but not significantly higher than P-ECM (0.53 ± 0.07 µg mg⁻¹ dry weight, P > 0.05). There was also no significant difference between L-ECM and P-ECM in terms of the collagen content (P > 0.05). While L-ECM was low on collagen content, it was high for sGAG content, as quantified by Blyscan assay to be 0.14 ± 0.02 µg mg⁻¹ dry weight, significantly higher than H-ECM and P-ECM (0.08 ± 0.01 and 0.03 ± 0.001 µg mg⁻¹ dry weight respectively, P < 0.01; figure 1(E)). This observation indicated that P-, L-, and H-ECM preserved complex multi-faceted ECM proteins, suggesting better mimicry of the native organs' intricate ECM composition. It also indicates quantitative differences in the retained ECM composition across different organs.

Having extracted the ECM from individual native organs, we next fabricated the patterning substrates to present the organ-derived ECM as individual microspots. We first coated glass slides with a thin nitrocellulose layer to immobilize the ECM proteins. Nitrocellulose coating rendered the surfaces more hydrophobic and resulted in water contact angles of 53.3º ± 0.2º, significantly higher than uncoated glass surface (36.9º ± 1.1º, n = 3, P < 0.01; figure 2(A)). This method allows adequate protein adsorption while providing sufficient optical clarity for microscopy and scanning applications [11]. To verify the efficiency of protein deposition, we used matrigel as our testing ECM solution and coated it on tissue culture plastic (TCP) pre-treated with nitrocellulose and compared with non-treated TCP for 5 min and 60 min. Immunostaining by Lectin peanut agglutinin (PNA) detected significantly higher intensity of relative fluorescence unit (RFU) on surfaces pre-coated with nitrocellulose (137 500 ± 2210 RFU versus 58 900 ± 3370 RFU for 5 min coating; 240 000 ± 20 405 RFU versus 110 050 ± 28 520 RFU for 60 min coating, P < 0.01, n = 3; figures 2(B) and (C)), indicating efficient deposition of ECM-associated carbohydrate structures on nitrocellulose surface. After confirming the efficient adsorption of matrigel, we then evaluated our organ-derived ECM by microspotting them onto nitrocellulose-coated glass slide. All organ-derived ECMs (P-, L- and H-ECM) and control ECM (matrigel) were spotted at the same concentration of 100 µg ml⁻¹. Atomic force microscopy (AFM) of microspotted organ-derived ECM extracts showed significant higher surface roughness—root mean square (RMS) than nitrocellulose only surface (P < 0.01, n = 5; figure 2(D), supplemental figure 1 (available online at https://stacks.iop.org/BF/13/015015/mmedia)), confirming their immobilization onto the nitrocellulose coated surface. Furthermore, the RMS value were about the same for all the organ-derived ECM extracts and matrigel control, indicating similar amount of proteins were immobilized [55]. In order to reproducibly pattern protein microspots of similar amount of protein, we used a repeater pipette that dispenses constant volumes of ECM protein solution. Matrigel measuring in 1–3 µl volume were spotted with a diameter ranging from 1500 μm–2300 μm to yield spatially separated ECM microspots (figure 2(E)). Each array contains up to 24 microspots for a glass slide (figure 3(A)) or 25 microspots for a well in 6 well plate (figure 6(A)). To render the glass surface in between the patterns resistant to unspecific protein and cell attachment, the remaining nitrocellulose layer were blocked with BSA. LI-COR slide scanning images of PNA lectin staining showed that each volume of microspot had similar diameter (<5% difference) and the diameter correlated linearly (R² = 0.95) with increasing volume (figure 2(F)). The same trend was observed with LI-COR quantification of near-infrared (680 nm) PNA stain—it correlated linearly with increasing volume and there was narrow variability of RFU for each volume (figure 2(G)), confirming the reproducible spotting quality that can be achieved by the repeater pipette dispersion. Taken together, these data suggest that this spotting technique is reproducible, and the platform can be efficiently patterned with organ-derived ECM proteins to generate an organ-specific ECM array.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Next, we characterized the individual matrix protein components of the organ-derived ECM spotted on the array by on-chip, immunofluorescence staining with IR-dye, and fluorescence quantification by LI-COR Odyssey imager (figure 3(A)). As a validation for this assay, we microspotted matrigel to study the assay sensitivity and range for the detection of the two most abundant ECM protein contained in Matrigel—Laminin and Collagen IV proteins. LI-COR image analysis detected increasing fluorescence intensity for both Laminin and Collagen IV staining with increasing concentration of spotted matrigel. The protein fluorescence intensity correlated linearly (R² > 0.97; supplemental figure 2) with matrigel concentration for both proteins over a range of dilution factors (1:4 to 1:16). Having validated our array platform, we next used it to characterize the ECM composition in P-, L-, and H-ECM, testing for a range of ECM proteins including Collagen I, Collagen IV, Fibronectin and Laminin. These ECM molecules were chosen because they are known to be critical in determining stem cell fate [56] and are the primary ECM components in various organs, including the pancreas [57, 58]. We quantitatively determined the levels of individual ECM proteins and compared to matrigel, rat tail collagen I (positive controls) and gelatin (negative control) (figures 3(B)–(E)). As expected, the on-chip immunofluorescence assay was able to detect the ECM protein profiles on both the positive ECM controls (high laminin and collagen IV for matrigel; high collagen I for rat tail collagen; figures 3(B) and (C)). Negative ECM control—gelatin is a denatured form of collagen and resulted in lower collagen I fluorescence intensity (figure 3(B)). Interestingly, despite the identical ECM preparation protocols and amount of protein being microspotted, the composition of the organ-derived ECM varied dramatically, illustrating the complex and different ECM compositional profiles among the three organs. There was no significant difference detected for Collagen I and Collagen IV ECM protein levels among the three organs (P > 0.05, n = 3; figures 3(B) and (C)), consistent with the previous assessment by Sircol collagen assay (figure 1(D)). However, laminin level was detected at much higher level in H-ECM (363.2 ± 15.8) than P- and L-ECM (84.9 ± 7.9; 153.5 ± 22.1 respectively, P < 0.05, n = 3; figure 3(D)). In addition, fibronectin level was also detected to be significantly higher in L-ECM (16.9 ± 3.4) than P- and H-ECM (2.9 ± 1.2, 4.2 ± 1.2 respectively; P < 0.01, n = 3; figure 3(E)). These data indicate that the ECM array combined with immunostaining and LI-COR analysis allows sensitive quantitative evaluation of different organ-derived ECM.

Cell adhesion to the ECM is essential for a coordinated morphogenesis and growth of functional tissue [59, 60]. In order to measure the adhesive properties of the organ derived ECM, we used pancreatic progenitor (PP) cells derived from hPSC as described in our previous published studies [50, 51], and seeded them onto the organ-derived ECM array. As depicted in step 1 of schematic in figure 4(A), we first generated definitive endoderm from human embryonic stem cells (hESCs) using activin A (100 ng ml⁻¹) and wnt3a (25 ng ml⁻¹) for 4 d (day 0–4). We then induced pancreatic progenitor cell fate by sonic hedgehog inhibition (KAAD Cyclopamine, 0.2 µM) for 4 d (day 4–8) and retinoic acid from day 6–8. With this reported protocol, we were able to efficiently generate up to 60% of PP cells expressing the transcription factor pancreatic and duodenal homeobox1 (Pdx1), which is the master regulator giving rise to all pancreatic lineage cells [61]. The pre-differentiated PP cells were then seeded onto the arrays in serum-free media and allowed to adhere for 2 h at 37 °C (step 2 of schematic figure 4(A)). To ensure uniform seeding, the slides were agitated every 15 min. We first tested this approach with matrigel spotted arrays. The PP cells adhered preferentially to matrigel spotted regions and did not attach to the nitrocellulose coated regions lacking ECM proteins. The cell patterning was robust over the array surface (3 × 3 ECM microspots), yielding near confluent cellular microspots as shown in figure 4(B). To quantify cells bound to each spot, cell nuclei were stained with DraQ5 DNA stain, a DNA-binding dye that fluoresces in the infrared spectrum and the slides were scanned and quantified with LI-COR to detect the fluorescent intensity of DraQ5. As shown in figure 4(C), the cell density correlated strongly with the cell seeding density of the hPSC-PP cells, which confirms both the quantification technique and compatibility of the array with hPSC derived cell types. Live/dead images of the adhered cells demonstrated good viability 24 h after cell seeding, indicating cytocompability of the ECM array (figure 4(D)).

Zoom In Zoom Out Reset image size

Using the same methodology, we next seeded the hPSC-PP cells on organ-derived ECM (P-, L- and H-ECM) spotted arrays and evaluated their adhesion (figure 4(E)). Live/dead images after 24 h showed most adhered cells were viable but there were considerable differences in cell adhesion between the three organ ECMs (figure 4(F)). This was further confirmed by quantification using LI-COR analysis of DRAQ5 DNA stain. Adhesion of the hPSC-PP cells to L-ECM was the highest among the three organ-derived ECMs, while H-ECM exhibited the lowest hPSC-PP cell adhesion (P < 0.01). Interestingly, in spite of having pancreatic tissue origin, P-ECM exhibited significantly lower adhesion of hPSC-PP cells as compared to L-ECM (P < 0.05; figure 4(G)). These results clearly suggest that adhesion of hPSC-PP cells was responsive to the composition differences between different organ-derived ECMs.

While the ECM array is a useful tool to interrogate cell-ECM interaction under adherent culture, the increasing appreciation of the importance of three-dimensional (3D) spatial organization of cells has prompted the development of 3D cellular arrays [17, 18, 20, 21, 62] which better mimics the complexity of the native microenvironment. In addition, many ECM proteins do not have an adhesive role but are still critical for stem cell growth, survival, differentiation and morphogenesis [7, 56, 63]. Hence, it is important to develop a 3D system to globally present the ECM proteins while supporting 3D culture of hPSCs. Alginate is a suitable hydrogel matrix which adequately supports hPSC growth and differentiation while being inert to cell adhesion [49, 64]. Further, alginate hydrogels have been previously developed into a 3D cell on-a-chip array for mouse embryonic stem cell [17] and human neural stem cell [18] differentiation. Here, we modified this technique to incorporate our organ-derived ECM into alginate, while encapsulating pre-differentiated hPSC cells. Instead of using poly (styrene-co-maleic anhydride) (PS-MA)-treated surface, we used nitrocellulose coated surface, which has been shown to have a higher protein binding capacity than PSMA-treated surface (80–100 µg cm⁻²[ ] versus 68.2 µg cm⁻²[ 65, 66]). After pretreating the glass slides with nitrocellulose, mixture of poly-L-lysine (PLL) and BaCl₂ were microspotted onto predetermined positions using the same repeater pipette microspotting technique as described above. Alginate was pre-mixed with specific organ-derived ECM. hPSC-PP cells were harvested and suspended within the alginate-ECM matrix material before micro-spotting onto the predetermined positions on nitrocellulose-coated glass slides. 3D cell-laden gel were formed by the deposition of the pre-mixed alginate and ECM solution containing hPSC-PP cells on top of the PLL/BaCl₂ bottom layer as shown in schematic figure 5(A). Under these conditions, the positively charged PLL serves as a substrate to bind Ba²⁺ ions and assist the attachment of the negatively charged alginate. The divalent Ba²⁺ ions bind preferentially to the G-blocks in the alginate and cause instantaneous gelation of the alginate chains to give rise to 3D cell-laden microspots (1 µl) with a height and diameter of 250 µm and 1600 µm respectively (figure 6(B)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

With the platform now being 3D, we wanted to test the sensitivity of LI-COR imaging platform in detecting the protein expression throughout the thickness of the 3D gel microspots with the same on-chip, immunofluorescence staining method. Previous studies have shown that LI-COR was able to detect near infrared and infrared fluorescence signals of up to 2 mm thickness [67, 68], which is within the thickness range of our 3D gel microspots (250 µm). In addition, we wanted to ensure that the ECM proteins were stably incorporated into the 3D gel microspots throughout the culture period (up to 10 d). To validate this, we microspotted premixed matrigel and alginate (cell free) at various concentration (0 µg ml⁻¹–200 µg ml⁻¹) onto the PLL/BaCl₂ pre-spotted array and maintained it in maturation media for 10 d at 37 °C culture incubator. The range of matrigel concentration on the 3D gel array were stained with PNA lectin and quantified by LI-COR Odyssey imager at day 0 and day 10 (figure 5(B)). We observed good linearity among the range of matrigel evaluated (Day 0, R² = 0.94; Day 10, R² = 0.98; figure 5(C)), indicating sensitivity of the immunostaining and LI-COR imaging under 3D configuration. Furthermore, majority of the matrigel was retained within the alginate throughout the 10 d culture period with less than 20% loss of ECM protein at higher matrigel concentration (i.e. 100 µg ml⁻¹ and 200 µg ml⁻¹; figure 5(D)). Specific ECM antibody staining also detected similar protein levels (data not shown) compared to the quantitative evaluation by 2D ECM array as described above (figure 3), indicating retained ECM after inclusion into alginate gel array, as would be expected.

Next, we used this miniaturized 3D alginate-ECM array to investigate the differentiation of encapsulated hPSC-PP cells. The pre-differentiated hPSC-PP were harvested with enzymatic digestion to yield single cell suspension (cell density of 2.5 × 10⁶ cells ml⁻¹) and mixed with specific organ ECM-alginate solutions. Arrays of 25 alginate microspots (figure 6(A)) (n = 5 for each ECM condition) were prepared and cultured in maturation media with notch inhibition, a known inducer of beta cell phenotype [49, 51, 69] for 9 d. During this period, the cells remained confined within the microspots, with no spot breakage (figure 6(C)) and no visible gel detachment from the slide, indicating the microspots were structurally stable. Confocal microscopy images also indicated that the encapsulated cells were evenly distributed inside the alginate-ECM microspots (figure 6(B)). However, live/dead staining showed significant cell death (up to 80%) in the 3D cell-laden alginate-ECM microspots (figures 6(D) and (E)). This is most likely due to cell-cell contact inhibition while harvesting the hPSCs, which is known to activate the apoptotic pathway in hPSCs [70]. While we added Rho-associated protein kinase (ROCK) inhibitor (i.e. Y-27632), which has been illustrated to boost cell survival in this process [70, 71], we still observed significant cell death.

In order to retain high viability of hPSC-PP cells in our 3D ECM array platform, we chose to encapsulate cell clusters instead of single cell hPSC-PPs. We hypothesize that the preserved cell-cell contact in the clusters will not activate apoptotic pathway and hence will retain high viability. In our previous study we have reported the excellent viability and pancreatic differentiation of hPSC-PP aggregates in alginate capsules [49]. hPSCs were encapsulated as single cells within the alginate capsules, where they formed small aggregates upon propagation and subsequently differentiated into pancreatic progenitor cells upon induction. These hPSC-PP aggregates can be easily harvested by mild treatment of EDTA to dissolve away the alginate capsules. Schematic in figure 6(F) illustrates the differentiation protocol to generate 3D hPSC-PP aggregates and encapsulate into the 3D ECM gel array (figure 6(G)). To ensure relatively homogenous-sized aggregates were encapsulated, we passed the hPSC-PP aggregates through a cell strainer (100 µm mesh size) to avoid inclusion of cell clusters larger than the microspot gel size. This will avoid cellular overgrowth and protrusion out of the alginate microspots as shown in our supplemental figure 3. After culturing for 9 d in different organ ECM gel array, live/dead staining of encapsulated hPSC-PP aggregates showed excellent viability (>90%, figures 6(H) and (I)) much higher than what was experienced with the encapsulated single cell hPSC-PPs. In comparing across different organ-derived ECM, no significant difference was observed in the viability of PP aggregates (P > 0.05, figure 6(I)).

Having achieved high viability, we next investigated the differentiation outcome of hPSC-PP aggregates exposed to different organ-derived ECM in the 3D alginate microspots. We cultured the encapsulated hPSC-PP aggregates in different organ-derived ECM alginate microspots (n = 5 for each organ) for 9 d in the presence of notch inhibition (DAPT). Parallel cultures on alginate with and without matrigel were used as controls. Immunostaining and LI-COR imaging was performed to assess pancreatic maturation using pancreatic and islet-specific transcription factors such as Pdx1, Nkx6.1, and insulin protein marker, C-peptide. The result showed that higher PDX1 expression was detected in all the alginate-ECM microspots compared to alginate only controls (figures 6(J) and (K)). NKX6.1 did not show any differences between P-ECM, H-ECM, matrigel and the alginate only control. However, L-ECM elicited a substantial increase of all PP markers, including insulin (C-peptide) protein expression (figure 6(K)). Interestingly, P-ECM only showed a modest increase in C-peptide level but did not show significant increase of PP markers or C-peptide when compared to H-ECM or controls (P > 0.05, figure 6(K)). Confocal sectioning also corroborated this finding and detected higher expression of C-peptide staining in L-ECM alginate microspots (figure 6(l)). To ensure that this observation was attributed to the ECM-specific differences and not any physical parameters of the gel, like stiffness, we performed AFM stiffness measurement of the alginate microspots after incorporation of ECM proteins. Our result revealed no stiffness differences between the alginate microspots alone and alginate microspots incorporated with ECM proteins (supplemental figure 4). Collectively, these data suggest that the hPSC-PP aggregates were responsive to organ-derived ECM differences and further matured into insulin-expressing cells, with highest differentiation phenotype observed in L-ECM matrix. Furthermore, such stem cell differentiation phenomena can be studied in a miniaturized 3D alginate-ECM array.