Bioinspired Precision Engineering of Three‐Dimensional Epithelial Stem Cell Microniches

Maintenance of the epithelium relies on stem cells residing within specialized microenvironments, known as epithelial crypts. Two‐photon polymerization (2PP) is a valuable tool for fabricating 3D micro/nanostructures for stem cell niche engineering applications. Herein, biomimetic gelatin methacrylate‐based constructs, replicating the precise geometry of the limbal epithelial crypt structures (limbal stem cell “microniches”) as an exemplar epithelial niche, are fabricated using 2PP. Human limbal epithelial stem cells (hLESCs) are seeded within the microniches in xeno‐free conditions to investigate their ability to repopulate the crypts and the expression of various differentiation markers. Cell proliferation and a zonation in cell phenotype along the z‐axis are observed without the use of exogenous signaling molecules. Significant differences in cell phenotype between cells located at the base of the microniche and those situated towards the rim are observed, demonstrating that stem cell fate is strongly influenced by its location within a niche and the geometrical details of where it resides. This study provides insight into the influence of the niche’s spatial geometry on hLESCs and demonstrates a flexible approach for the fabrication of biomimetic crypt‐like structures in epithelial tissues. This has significant implications for regenerative medicine applications and can ultimately lead to implantable synthetic “niche‐based” treatments.

we hypothesized that spatial geometry plays a crucial role in determining stem cell fate. Our aim was to recreate the precise geometry of an epithelial stem cell niche using two-photon polymerization (2PP) and to determine the role of spatial geometry on stem cell phenotype in a 3D microniche setting.
Geometrical cues have been reported to influence gene expression on 2D micropatterned substrates, [11][12][13][14] but little is known about how geometrical cues influence cell behavior in 3D "niche-like" settings. In this study, the influence of spatial geometry on the differentiation of stem cells was investigated, independent of imposed chemical and environmental conditions. As an exemplar, we focused on the limbal stem cell niche. The limbus is the niche responsible for long-term renewal of the cornea. [15] It is a 1 mm-wide circular region separating the sclera from the cornea. It contains the palisades of Vogt, [16,17] some of which are associated with limbal epithelial crypts (LECs). Three types of stem cell niches have been identified in the limbal region: LECs, limbal crypts, and focal stromal projections. [18] The presence, location, and shape of LECs were described for the first time by Dua et al. [19,20] During healing and homeostasis, human limbal epithelial stem cells (hLESCs), also known as corneal epithelial stem cells, migrate towards the center of the cornea and differentiate [21] to constantly replace corneal epithelial cells. [22] Such niche structures are common in all epithelial tissues and share certain characteristics, including geometry, a stromal compartment, and zonation in epithelial cell phenotype. This zonation is such that there is a stem cell zone, a progenitor (transiently amplifying) cell zone, and a region where cells are terminally differentiated. [23] We used 2PP to create gelatin-based 3D crypts that mimic the precise geometry of the limbal stem cell niche. 2PP is an additive manufacturing technique capable of fabricating 3D micro/nanostructures without a mask and with a possible feature size of <100 nm. This is achieved through the absorption of multiple photons simultaneously, usually in the infrared range, and the requirements for such an event result in small region of illumination with a sufficient intensity of photons to initiate polymerization. [24] Using histological sections described by Dua et al., [19] crypt dimensions were identified and a 3D tapered geometry was designed to replicate the niche using computer-aided design (CAD) software (Figure 1). These structures were then fabricated by 2PP from gelatin methacrylate (GelMA) as concave scaffolds, with a 200 µm opening tapering to a 20 µm point at the base. GelMA is being increasingly used in tissue engineering due to its biocompatibility and tunable physicochemical properties, enabling control over the mechanical properties of the gel. [25] GelMA presented a degree of methacrylation of 80.2% ± 6.1%, confirmed by the trinitrobenzenesulfonic acid (TNBS) assay ( Figure S1, Supporting Information). The microniche constructs were successfully printed and they closely replicated the CAD model ( Figure 1).
Following the creation of the niche-mimicking geometries, hLESCs (commercial hLESCs and biopsy-derived td_hLESCs) were seeded inside the niches in xeno-free conditions to Adv. Biosys. 2020, 4,2000016  investigate their ability to repopulate the crypts and to study their differentiation. hLESCs (CELLnTEC; Switzerland) presented a homogenous monolayer of single cells ( Figure S2, Supporting Information). Td_hLESCs were seeded on an irradiated 3T3-feeder layer, and these feeder cells were later removed by changing the media upon keratinocyte confluency ( Figure  S2, Supporting Information), leaving the cells of interest on the scaffold. After seeding, cells proliferated and were distributed along the length of the concave crypt, from the base to the rim (Figure 2A,B). Cell differentiation was studied along the z-axis, using cytokeratin 14 (CK14) and cytokeratin 3/12 (CK3/12) as markers for limbal progenitor cells and differentiated corneal epithelial cells, respectively. [26,27] Zonation in marker expression was observed along the z-axis, presenting differentiated cells on the base (CK3/12; green) and limbal progenitor cells (CK14; red) at the rim ( Figure 2C; Figure S3, Supporting Information). As CK14 is a progenitor cell marker (rather than a stem cell marker, such as p63), we conclude that cells expressing this marker represent a cell population at an early stage of the differentiation pathway.
Atomic force microscopy (AFM) was conducted to identify the stiffness of the synthetic microniche, as it allows elastic modulus measurements to be performed at the same length scales at which cells interact with material. [28] Due to the length of the AFM tip, the base of the scaffold was not accessible; hence, its properties could not be measured. Therefore, the base was measured using a flat structure fabricated with the same printing parameters used for making the base of the concave microniche ( Figure 3A). Results indicated that there was a notable difference between the mean values of elastic moduli in the z-direction with the top always stiffer than the base (75.60 vs 25.87, 92.39 vs 16.05, and 25.11 vs 5.46 kPa). However, the inherent variability associated with AFM measurements resulted in statistical nonsignificance (p = 0.09, n = 3). Elastic moduli were in the same order of magnitude of the cornea, which was also assessed by AFM (7.5-110 kPa [29] ). This range of values describes the local stiffness variations occurring in the transition from the limbal to the surface of the cornea. [30,31] To our knowledge, there are no studies so far that have shown correlation between stiffness and z-position of similar constructs. Future work will investigate this aspect in 2PP additive manufacturing, allowing the production of artificial microniches and tissue-engineered constructs with stiffness gradients.
To explore whether the spatial z-position influences differentiation, hLESCs were seeded on two flat scaffolds with different heights, 20 µm (4.9 ± 2.5 kPa) and 200 µm (28.7 ± 7.2 kPa),  Commercially available a-c) hLESCs or d-f) a feeder layer of irradiated 3T3-J2 cells+td_hLESCs were seeded, and distribution along the microniche was monitored using Hoechst 33258. C) Three z-positions representing a,d,g) the base, b,e,h) the middle, and c,f,i) rim of the microniche. For the base, the area surrounding the center has been darkened to highlight the structure's base. hLESCs were stained with CK3/12 (green; differentiated cells), CK14 (red; limbal progenitor cells), and Hoechst (blue; nuclei). a-c) Green, d-f) red, g-i) blue, and j-l) merged channels are presented. The majority of the cells on the base expressed the differentiation marker CK 3/12 (green), while the majority at the rim expressed the progenitor cell marker CK14 (red). Images are representative of three independent experiments. which were produced using the same printing parameters used for the base of the crypt structure. Since 2PP is a high-resolution printing method, whereby excitation is restricted to a tiny focal volume, [32] solidification of the photopolymer is strongly confined inside the focal volume throughout the multilayer processing procedure, and therefore the degree of polymerization is uniform throughout the printed microstructure (unpublished data). Results demonstrated that cells expressed the limbal progenitor cell marker (CK14) predominantly on both materials, independent of the differences in height and stiffness ( Figure 3B).
There is some evidence that stiffness in the limbus may impact cell response. [31,33] However, Moers et al. have reported that stiffness impacted differentiation of corneal epithelial cells only on unphysiologically stiff substrates. [34] After 3 days of culture, corneal epithelial cells grown on physiologically relevant substrates (40 and 80 kPa) both maintained an early differentiation state, represented by the expression of cytokeratin 19. However, growth on unphysiologically stiff substrates (1.5 MPa) induced cell differentiation, denoted by the expression of the late differentiation marker cytokeratin 12. [34] To delineate the effects of stiffness versus spatial geometry of the fabricated microniches, hLESCs were seeded on flat GelMA substrates of varying stiffness. The average elastic moduli (±SD) for 15%, 25%, and 40% w/v GelMA substrates, obtained by compression testing, were 173.5 ± 78.2, 509.7 ± 60.4, and 1014.4 ± 75.5 kPa, respectively. Varying stiffness did not have a pronounced effect on stem cell differentiation. Numbers of stem cells and differentiated cells ( Figure 3C,D) did not vary significantly (analysis of variance (ANOVA), p > 0.05, post hoc with Bonferroni correction) by increasing stiffness of flat substrates over the timeframe of the study in agreement with previous studies. [34] These results point to elimination of stiffness, within this range, as a contributor to cell fate and confirm that geometry is a major decisive factor for control of cell behavior.
To confirm the zonation in cell phenotype observed with commercial human primary stem cells (hLESCs), the same protocol was performed using freshly isolated cells from biopsy samples (td_hLESCs), supported by an underlying feeder layer of irradiated 3T3-J2 fibroblasts. Limbal epithelial cells were extracted from human cadaveric limbal biopsies and cultured using serum-supplemented media. CK3/12 was used as a differentiation marker and p63 as a stem cell marker ( Figure S4  www.adv-biosys.com www.advancedsciencenews.com expression as was observed with commercially available cells, with cells at the base of the microniche expressing CK3/12 and cells at the rim expressing the stem cell marker p63. We postulate that spatial geometry created a milieu of cellular signals inside the niche that enhanced cell differentiation at the base and not at the rim of the microniche construct, resulting in the observed cell differentiation gradient. A previous study reported that adipose-derived stem cell differentiation was localized within regions containing small, elongated cells (e.g., inner edge of a ring). [11] Therefore, the absence of complementary cues within the microniche in this study will have had an impact on the observed geometrically specific differentiation pattern. Maintenance of tissue homeostasis requires spatiotemporal cell organization to ensure tissue function. [35] While replicating the geometry and dimensions of the stem cell niche has a major influence over cell phenotype, as we report herein, cellular organization and phenotype within the niche is also influenced by other factors. Another potential parameter is the topography at the base. Previous studies have suggested that specific configurations, such as pillars (micropatterns of 11 vs 5 µm), can enhance the expression of late corneal keratinocyte markers. [34,36] Additionally, restricting the adhesive area of human epidermal stem cells triggered the initiation of terminal differentiation. [37] Further studies to understand the interplay between geometry and cell-cell signaling are needed to fully understand this phenomenon.
Mimicking intestinal crypt architecture using micromolded scaffolds has been previously reported, with a focus on recreating physiological chemical gradients using growth factors. [38] Few studies have specifically attempted to replicate limbal niche structures, [39,40] and have not replicated its precise dimensions. [19,20] Ortega et al., who reported that rabbit limbal cells populated ≈300 µm "micropockets," did not assess phenotypic expression [39] and have not used a clinically relevant cell type, which is vital since 3D morphology of limbal crypt structures varies between species. [41] In another study, the open ridge-like structures under investigation included stromal fibroblasts, and therefore the phenotypic response they reported in the limbal epithelial cell population cannot be attributed to geometry alone. [40] While limited investigation has been carried out into the importance of stem cell size and shape in 3D, [42] a geometrically induced, without the presence of a chemical gradient, differentiation gradient in a biologically relevant architecture has not been achieved, which is the significant finding in this study. This will have substantial impact on the design of in vitro stem cell niche models and novel "microniche-based" scaffolds for epithelial tissue regeneration.
In conclusion, we have demonstrated that mimicking the precise geometry of the epithelial stem cell niche impacts epithelial stem cell fate without the use of exogenous cues. This was achieved by producing microniches replicating in vivo architecture using 2PP, which allows the freedom to recreate structures at length scales appropriate for the production of stem cell microniches. The techniques described herein provide a tailorable platform for the development of synthetic microniches, allowing fine control over cell microenvironments. While this study provides the foundation for recreating epithelial stem cell crypts, further studies are necessary to build up the complexity of the constructed microniches and recreate cues from ECM and associated cells. It also provides a strategy to facilitate epithelial stem cell transplantation, whereby cells can be confined within implantable constructs that mimic characteristics of the stem cell microenvironment. This method is flexible and can be applied to other epithelial tissues, and as such represents a significant advance in the field of regenerative medicine.

Experimental Section
Chemicals were purchased from Sigma-Aldrich (UK) unless otherwise stated.
GelMA Synthesis: GelMA was synthesized as described previously. [43] Porcine skin gelatin (10% w/v type A, 300 bloom) was dissolved in phosphate-buffered saline (PBS; Gibco, UK) at 80 °C. Methacrylic anhydride (8% v/v) was added at 0.5 mL min -1 and stirred at 60 °C for 3 h. To quench the reaction, the solution was diluted 5× with PBS and stirred for a further 30 min. The solution was dialyzed against distilled water using 8 kDa cut-off dialysis tubing (BioDesignDialysis Tubing, USA) for 1 week at 50 °C. Dialysis water was changed three times per day. The solution was filtered through a 40 µm filter and lyophilized for 3 days. Freeze-dried GelMA was stored at -20 °C until use.
Quantification of Degree of Crosslinking: Percentage of methacrylation was quantified using 1 H NMR (400 MHz; Bruker UK Ltd., UK) at room temperature (RT). Phase and baseline corrections were applied before obtaining integrals of peaks of interest. MestReNova software (v6.0.2) was used to analyze spectra (64 scans). GelMA and gelatin were dissolved in deuterium oxide (10 mg mL -1 ). Presence of a double bond signal (5.35-5.60 ppm) indicated addition of the methacrylate vinyl group. Percentage of methacrylation was calculated using the method described by Hoch et al. [44] The integral of the methylene peak (d = 2.7-2.9 ppm) was used for quantification of the lysine signal. Degree of substitution was determined by comparing lysine signals of modified and unmodified gelatin (Equation (1)). Intensity of the aromatic region (7-7.5 ppm) was used as reference (1) Degree of methacrylation was checked using the 2,4,6-TNBS assay. Gelatin and GelMA (4 mg) were dissolved in 0.25% v/v TNBS (in 0.01 M sodium hydrogen carbonate solution) and incubated at 40 °C for 3 h. Hydrochloric acid (3 mL 6N, Alfa Aesar, UK) was added and the solution was incubated at 80 °C for 1 h. After cooling, 4 mL deionized water was added and absorbance measured at 345 nm (Tecan Infinite M200 microplate reader). Degree of crosslinking was calculated according to Equation (2) 2PP of GelMA to Create Crypt Structures: Polymeric solutions were prepared by mixing GelMA foam with the photoinitiator P2CK to obtain a suitable formulation. P2CK was prepared according a published protocol [45] or by a modified procedure. GelMA (15% w/v) was dissolved in PBS, and P2CK was added to a final concentration of 0.3% w/v. To fabricate microniches, a 2PP system (Nanoscribe Photonic Professional, GmbH, Germany) with a femtosecond laser emitting at 780 nm in galvo scan mode was used. The structure was developed using CAD 3D Autodesk Inventor software and printing parameters were set by the Describe software. The laser beam was focused using a conventional 63× microscope objective (Zeiss, NA 1.4 with oil immersion media) mounted on a linear stage for vertical positioning. The laser writing system was controlled by Nanowrite software. A drop of gel was cast between a 22 × 22 mm glass slide (thickness n.1.5), and an 18 × 18 mm www.adv-biosys.com www.advancedsciencenews.com cover glass (thickness n.1.5), separated by a 300 µm spacer ( Figure 1E). The printing process was achieved by tuning laser power, scan speed, and hatching distance. After printing, structures were "developed" by immersing samples in PBS (40 °C for 1 h) to remove nonpolymerized hydrogel; samples were then left at 4 °C overnight. Irgacure 2959 (0.1% w/v) was added and scaffolds were UV-cured (365 nm at 15 W) for 10 min to increase stability.
Determining Surface Elastic Modulus Using AFM: A thin cantilever (nominal spring constant between 0.07 and 0.025 N m -1 ) with a silicon nitride tip (MSNL-10, probe A; Bruker Nano, UK) was used. The cantilever was positioned on the construct's rim and on a printed flat scaffold (height = 20 µm) to emulate the base of the scaffold. Force graphs were recorded using an MFP-1D AFM instrument (Asylum Research, USA) and elastic moduli (E) were calculated from the approaching force curves based on the Hertz model, [46] according to Equation (3) where F is the applied force, R is the probe radius, δ is the indentation of the samples, E is the elastic modulus, and µ is the Poisson's ratio. Three experiments were conducted for each condition (A, B, and C). Numbers of modulus measurements in experiments A, B, and C were 278, 92, and 310 for the top and 158, 28, and 211 for the base, respectively. Cell Culture of Xeno-Free Human Limbal Stem Cells: Commercial primary hLESCs were obtained from CELLnTEC Advanced Cell Systems (P2, single-donor; Switzerland) and cultured in serum-free epithelial culture media CnT-Prime (CnT-PR; CELLnTEC Advanced Cell Systems) with addition of 100 U mL -1 penicillin, 0.1 mg mL -1 streptomycin, and 0.25 µg mL -1 amphotericin B, and cultured at 37 °C and 5% CO 2 . Cells were cultured at 4000 cells cm -2 .
Cell Seeding of the Microniche Constructs: hLESCs and td_LESCs suspensions (150 000 cells in 50 µL) were seeded onto the constructs. Afterwards, media was added to cover the entire surface of the scaffolds. For td_hLESCs, the feeder layer was seeded on the scaffold 4-24 h before adding keratinocytes. At 10 days post-seeding, cells were fixed for analysis.
Immunocytochemistry: Cells were dual-stained with either CK14 or p63 (limbal progenitor and stem cell markers, respectively) and CK3/12 (differentiated corneal epithelial cell marker). Following fixation, permeabilization was carried out using 0.1% v/v Triton X-100. Samples were exposed to blocking solution for 1 h (1% w/v bovine serum albumin (BSA; BDH Prolabo, USA) in PBS, 0.3 M glycine, and 3% v/v goat serum). Antibodies (Table S2, Supporting Information) were diluted in wash buffer (1% w/v BSA and 0.3 M glycine in PBS). Samples were incubated in primary antibodies overnight at 4 °C followed by secondary antibodies for 1 h at RT. Samples were washed and stained with Hoechst 33258. Fluorescent mounting medium (Dako, VWR, UK) was added prior to analysis by confocal microscopy (Zeiss Elyra PS1 microscope with ZEN software, Carl Zeiss, Germany).
Statistical Analysis: For the statistical analysis, GraphPad Prism 6 (v6.01) was used. For the AFM analysis, three experiments were conducted for each condition. Two-tailed Mann-Whitney U tests were performed to investigate differences in elastic moduli of the rim and base. In addition, mean AFM measurements of each sample were obtained and a two-sided paired t-test was performed (n = 3; error bars represent ±SD). All differences were considered statistically significant when p ≤ 0.05. For the assessment of stem cell differentiation on GelMA constructs of different stiffness, a two-way ANOVA (area, two levels: top and bottom; GelMA concentration, five levels: 5, 15, 25, 40% w/v and control) with post hoc and Bonferroni correction was performed (n = 2; error bars represent ±SD). All differences were considered statistically significant when p ≤ 0.05.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.