Retinal Multipotent Stem-Cell Derived “MiEye” Spheroid 3D Culture Model for Preclinical Screening of Non-viral Gene Delivery Systems
Ding-Wen Chen and Marianna Foldvari
School of Pharmacy, University of Waterloo, 10A Victoria St. S., Kitchener, ON N2G 1C5, Canada.
Submitted: July 1, 2018; Accepted: July 11, 2018; Posted July 16, 2018
Non-viral retinal gene therapy is a promising therapeutic approach towards the management of retinal degenerative diseases especially glaucoma. Current methods of in vitro preclinical screening of candidate nanoparticle systems in monolayer cell cultures are not reliable in predicting in vivo performance. In this paper we describe the development of a multipotent stem-cell derived three-dimensional “mini-retina” culture model (MiEye) that aims to simulate an in vivo clinical model for more reliable gene delivery system screening. Through utilization of multiplex gene expression profiling, we have shown that retinal stem cells can be differentiated in 3D culture to generate retinal neurospheres comprising of multiple retinal cell types. The 3D cell culture model combined with confocal microscopy imaging and fluorescence profiling techniques is a powerful tool as a retinal gene and drug delivery screening model.
- cMFI combined mean fluorescence intensity
- CM culture medium
- DM differentiation medium
- MM maturation medium
- K2-NP K2 nanoparticle
- PLO poly-L-ornithine
- RGC retinal ganglion cell
- RPE Retinal pigmented epithelium
- RSC retinal stem cells
- TE Transfection Efficiency
- ULA ultra-low attachment
Spheroids, retina, neurosphere, stem cell, 3D culture, gene delivery, non-viral, nanoparticle, glaucoma, retinal ganglion cell, gene expression profiling
Rationale and Purpose
In a non-viral approach to delivering therapeutic nucleic acids for glaucoma management, a significant challenge lies in efficiently delivering genes into target retinal cells. The majority of non-viral gene delivery systems that have shown highly effective gene transfer capacity in 2D monolayer transfection have failed to translate to comparable levels of gene delivery efficiency in vivo. The gap, in part, is a consequence of the lack of correlation between the results generated from the current standards of transfection efficiency (TE) screening to in vivo feasibility. Recognizing such deficiency, we aimed to bridge this gap through the development of a “mini-retina” neurosphere model composed of major retinal cell types arranged in a three-dimensional (3D) tissue-like architecture. The MiEye retinal neurospheres aim to serve as a retinal model for the rapid screening of transfection capacity of nanoparticle (NP) systems to retinal cells, towards the generation of in vitro results with higher in vivo translatability.
Glaucoma is a neurodegenerative disease of the retina characterized by multifactorial pathophysiology that ultimately results in the loss of retinal ganglion cells (RGCs). A gene therapy approach such as neurotrophic factor gene therapy has the potential to provide extrinsic neurotrophic support that can rescue and protect glaucomatous stressed RGCs.[2,3] Non-viral approaches to deliver therapeutic genes to the retinal cells to carry out therapeutic action for stressed RGCs provides many advantages compared to viral approaches such as lower risk of immunogenicity and mutagenesis, as well as better patient compliance and large-scale manufacturing feasibility.
Despite the large number of NP systems that have demonstrated a high level of TE in vitro, very few demonstrated effectiveness in vivo. An aspect of this is a disconnect between in vitro and in vivo TE screening, suggesting that current standards of in vitro models do not sufficiently predict in vivo performance and feasibility of NPs. The current standards of in vitro NP gene delivery evaluation rely on the assessment of TE in cells cultured in 2D monolayers, which include two major pitfalls in cellular modeling from a gene delivery perspective: (1) a lack of multicellular tissue spatial architecture; and (2) suboptimal representation of NP biodistribution and kinetics.
In retinal degenerative diseases such as glaucoma, cell-cell interactions play an important role in the pathogenesis through various mechanical and molecular factors near the trabecular meshwork and optic nerve head.[5–8] Thus, tissue models such as organotypic explant cultures could greatly benefit the understanding of glaucoma pathogenesis and facilitate gene therapy development using representative models in vivo. Organotypic explants of the retina have been explored by various groups to study pathogenesis and also for drug development purposes.[9–11] However, preparation of retinal explant culture is difficult. While cultivatable for up to two weeks ex vivo, retinal explants spontaneously undergo degeneration upon isolation. 3D retinal models generated from stem cells can yield various retinal cell types along with cellular arrangements similar to the retina [12, 13]. The ability of stem cells to differentiate autonomously and form the optic cup structure in 3D culture models was first demonstrated by Eiraku et al. using mouse embryonic stem cells. [14, 15] Following that, there have also been various attempts to generate retinal organotypic cultures.[16–18] Maekawa et al. have generated retinal organotypic cultures from mouse embryonic stem cells utilizing mouse and human embryonic stem cells.
Another approach in the construction of 3D models has been explored using bioprinting technologies, in which cells are printed based on the desired cellular arrangement. Lorber et al. have demonstrated that retinal stem cells (RSCs) and glial cells of the retina can be printed using 3D printing technology while maintaining viability and neurite outgrowth capacity for both cell types. More recently, a study conducted by Kador et al. demonstrated that by combining 3D printing with radial electrospun scaffolds, they were able to 3D print RGCs onto electrospun surfaces with precise distribution and positioning of RGCs as found in the retina. From a spatial arrangement standpoint, 3D printing-based techniques theoretically possess a comparative advantage as they allow for more control in dictating cellular location on a tissue-like arrangement by “printing” specific cells at specific locations.
While the spatial arrangement of the cells cultured in 3D culture is either dependent on the resultant conformation from self-aggregation or the conformation dictated by the scaffold onto which the cells adhere. While these various advancements in different tissue modeling approaches are pivotal in developing in vivo-like models for many biological and physiological studies, a relatively simpler 3D model is needed for rapid and more predictive screening of non-viral gene delivery systems.
The mouse multipotent retinal stem cells (CD1-4 RSCs) used in this project have previously been described to differentiate into all major retinal cell types in 2D monolayers. In this paper we describe the engineering of 3D spheroids, termed “MiEye” under different culture conditions using the CD1-4 RSCs. The RSCs can be seeded into 96-well plates and differentiated into multiple retinal cell types while forming spheroids providing an easily scalable model for retinal gene delivery assessment.
The main objective in this investigation was to develop a 3D “mini-retina” model that can be utilized as an in vitro model for the screening of non-viral gene delivery systems. The experimental design focused on two major aspects. First, the generation and optimization of retinal neurospheres from CD-1 RSCs was performed through investigation of various combinations of conditions with the goal of generating structurally stable retinal neurospheres that would contain the most diverse retinal cell types.
The retinal neurospheres (MiEye 1-11) were evaluated using a combination of light microscopic imaging and multiplexed gene expression profiling techniques and the neurosphere with the most optimal properties was selected for screening studies. Second, screening of transfection characteristics of a panel of nine different compositions of a commercially available NP system, K2® (K2-NPs), was conducted in the selected MiEye8 neurosphere and in representative retinal monolayer cells by confocal laser scanning microscopy.
CD1-4 RSC culture
CD1-4 RSCs isolated from 4 to 8-weeks old CD-1 mice were kindly gifted by Dr. Ting Xie, Stower Institute for Medical Research and University of Kansas. The purified CD1-4 RSCs were cultured in 12-well plates coated with 0.2% gelatin for 3 hours at 37°C and maintained in RSCCM media. Culturing medium for CD1-4 (RSCCM) was composed of advanced DMEM/F12 (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) media with addition of 1.0 g/L glucose (Sigma-Aldrich, St. Louis, MO, USA), 1.0 g/L lactose (Sigma-Aldrich), 1.0 g/L BSA (Sigma-Aldrich), 0.045 g/L proline (Sigma-Aldrich), 2 mM nicotinamide (Sigma-Aldrich), 2 mM L-glutamate (Gibco), 1% penicillin/streptomycin (Hyclone), 1% insulin-transferrin-supplement (Gibco), 5% knockout serum replacement (Gibco), 20 ng/mL EGF (Gibco), and 20 ng/mL bFGF protein (Gibco). Differentiation medium (RSCDM-1) was composed of advanced DMEM/F12 (Gibco) with 2% B-27 supplement (Gibco), 1% L-Glutamine (Gibco) and 10 ng/mL bFGF protein (Gibco). RSCDM-1 containing 1% N-2 supplement (Gibco) was named RSCDM-2. Maturation medium was formulated with 2 mM L-Glutamine (Gibco), 2% B-27 supplement (Gibco), 1% N-2 supplement (Gibco), 10 ng/mL bFGF protein (Gibco), and 100 ng/mL BDNF protein (Peprotech), 10 ng/mL glial-derived neurotrophic factor protein (Gibco), and 10 ng/mL insulin growth factor protein (Gibco) in advanced DMEM/F12, with (RGCMM-3) or without (RGCMM-4), 2% fetal bovine serum (Thermo Fisher Scientific).
CD1-4 RSCs were cultured on 0.2% gelatin (Sigma-Aldrich) coated 12-well tissue-culture treated plate for 3 hours at 37 °C. For maturation of retinal neurospheres, culture surfaces were coated with either Matrigel™ (Corning Incorporated, Corning, NY, USA), or combination of poly-L-ornithine (PLO) (EMD Millipore, Billerica, MA, USA) and laminin (EMD Millipore). For PLO-laminin coating, 0.015 mg/mL of PLO were prepared in sterile water and applied on to the glass-bottom surface of multi-well plates overnight at room temperature. PLO was then removed and laminin was coated at a concentration of 2 µg/cm2 overnight at 4 °C. For Matrigel™ coating, 7.6 mg/mL Matrigel™ (Corning) was coated on the culture surface at 100 µL/cm2 volume, and allowed for gel formation for an hour at 37°C.
Generation of MiEye retinal neurospheres
The retinal neurospheres model was established through differentiation of CD1-4 RSCs in commercially available force-floating three-dimensional 96-well ultra-low adhesive (ULA) microplates (Corning). For differentiation of CD1-4 RSCs, 15,000 cells (best seeding density compared to 5,000 and 10,000 cells/well) were seeded into each well of the ULA microplate, and differentiation was carried out using various protocols consisting of different combinations of media, schedules, or surface coating, resulting in the generation of 11 different retinal neurosphere types termed MiEye (1-11) with 2–4 spheres for each. MiEye 1-4 were 24-day differentiated retinal neurospheres, first initiated through differentiation in RSCDM-1 media for 3 days, which was then switched into RSCDM-2 media for 7 more days. Following 10 days of differentiation, spheres were transferred onto either PLO-Laminin or Matrigel™ coated surface for maturation using either RGCMM-3 or RGCMM-4 media. MiEye5 was an 18-day differentiated retinal neurosphere, first initiated through differentiation in RSCDM-1 media for the first 4 days, followed by differentiation in RSCDM-2 for 14 more days. MiEye7 retinal neurosphere was generated by further maturating the 18-day differentiated spheres on PLO-laminin coated surface for 3 days in RGCMM-4. MiEye8 retinal neurospheres were 6-day differentiated spheres, which were initiated by differentiating in RSCDM-1 for 2 days. After 2 days of differentiation, the differentiated spheres were changed to RSCDM-2 media for 4 more days. MiEye10 was generated by subjecting MiEye8 to further maturation on PLO-laminin for 4 days in RGCMM-4 media.
RNA was isolated using a commercially available silica column-based RNA isolation kit, PicoPure RNA Isolation Kit (Arcturus, Thermo Fisher Scientific, Waltham, MA, USA), based on the manufacturer’s suggested protocol. Isolated RNA was subjected to gene expression profiling for retinal cell biomarkers using the NanoString nCounter Analysis system, a highly sensitive and reproducible multiplexed mRNA gene expression quantitation method (NanoString, Seattle, Washington, USA). The technology enables direct hybridization of mRNA to its sequence-specific target reporter and capture probe, which allows for the quantitation of mRNA copy number for multiple (up to 800) mRNA sequences simultaneously with high sensitivity for the generation of a gene expression profile.
Retinal biomarker gene expression analysis was performed using a custom code set containing 25 target genes representative of most retinal cell types and housekeeping genes as controls. The panel of target genes includes biomarkers representing major retinal cell types. ISL-1, Map-2, Nef-H, Pou4f1, Rbpms, Sncg, Thy-1.2, and Tubb3 are representative RGC biomarkers; Aqp4, Prdx6 and Slcla3 are representative astrocyte biomarkers; Abca8a, ald1a1, and Vim are representative Müller glia biomarkers; and Nrl and Rom1 are representative photoreceptor biomarkers. Rpe65, Calb2, Prkca, and Lim2 each individually represent retinal pigment epithelium (RPE), amacrine, bipolar and horizontal cell biomarkers, respectively. Each reaction was analyzed with 100 ng of RNA in a 5µL loading volume, mixed with capture probe and reporter probe. The gene profiling experiment was carried out at the Princess Margaret Genomics Centre (Toronto, Ontario, Canada).
Preparation of K2-NPs
K2-NPs were formulated by complexing the K2 lipopolyamine-based proprietary transfection reagent (Biontex Laboratories GmBH, München, Germany) with gWiz™ green fluorescent protein (GFP) reporter plasmid (Aldevron, Fargo, ND, USA) at various ratios for 30 minutes prior to transfection. K2-DNA v/w ratios from 1:1 to 10:1, DNA dose ranging from 1 to 10g were formulated and evaluated for TE on the retinal neurospheres. Transfection was carried out by first adding 25 µL of K2 Multiplier into each well for 2 hours. After 2 hours, complexed K2-NPs were added into wells. Nine different K2-NPs formulations (F1-F9) composed of various combinations of K2 transfection reagent and DNA concentrations, were prepared for transfection evaluation (Table 1).
Table 1. K2-NPs Formulated for Transfection in MiEye8 Retinal Neurosphere Model
Empirical K2:DNA Ratio
K2 Reagent (µL)
Confocal Laser Scanning Microscopy
Retinal neurospheres were counterstained with either 5µM Syto™ 62 nuclear stain (Thermo Fisher Scientific, Waltham, MA, USA) or 5µM DRAQ5™ nuclear stain (Thermo Fisher Scientific) for 30 minutes at 37°C prior to imaging. Neurospheres were imaged using a Zeiss LSM710 confocal laser scanning microscope (Carl Zeiss AG, Oberkochen, Baden-Wü, Germany) with 20× objective. Three-dimensional images of the spheres were captured using the z-stack functionality in the Zen 2009 software, each constituting 20 optical slices of images through the observable thickness of the retinal neurospheres. Transfection was measured by the detection of GFP expression in the retinal neurospheres. The combined mean fluorescence intensity (cMFI) of retinal neurospheres was obtained by calculating the average of all individual MFIs of optical sections of the neurospheres captured using z-stack. The cMFI of transfected neurospheres was normalized to the MFI of the untreated control neurosphere (background fluorescence). Co-localization between GFP expression and nuclei staining were also assessed using Mander’s Overlap Coefficient for each optical slice.
Generation and characterization of MiEye retinal neurospheres
Initial studies showed that the undifferentiated CD1-4 RSCs themselves were able to form stem-cell spheres in 3D culture conditions when seeded at cell densities of 5,000, 10,000, and 15,000 cells in ULA spheroid plates.
Figure 1. Undifferentiated retinal stem cell-derived retinal neurospheres cultured in ultra-low adhesive spheroid plate at day 3 and 12. Images were captured with 10× objective using light microscopy.
Spheroids were formed within 24 hours and their morphology was maintained for at least 12 days (Figure 1).
Using a similar approach, retinal neurospheres were initiated by seeding CD1-4 RSCs suspended in differentiation medium in ULA plates. Eleven different MiEye retinal neurospheres were prepared using various protocols (Table 2). A representative workflow of differentiation and maturation is illustrated in Figure 2A and 2B. After culturing of the RSCs for 6–10 days, the formed differentiated neurospheres were allowed to mature on the coated surface (i.e., PLO-laminin) to evaluate the possibility of inducing the formation of morphological features such as neurite extensions. The effect of various combinations of parameters such as retinal neurosphere population (5,000, 10,000, and 15,000 cells per sphere), maturation medium (serum / no serum), and surface coating (PLO-laminin/Matrigel™) were evaluated on the spheres structural integrity. First, a decrease in seeding cell density from 15,000 cells to 5,000 cells resulted in a decrease in retinal neurosphere structural stability after 2 days of maturation (Figure 3A). Furthermore, the decreased structural stability was more obvious in the presence of Matrigel™ coating and serum. A decrease in structural stability was evident from size reduction and deformation of the spheres, along with increased monolayer cell attachment on the coated surface surrounding the sphere. Retinal neurospheres were least structurally stable when maturation was taking place in serum-containing maturation media, coupled with Matrigel™ coated surface, where retinal neurospheres were completely deformed by day 6 (Figure 3B). Seeding density played an important role in the sphere structural stability as retinal neurospheres constructed with only 5,000 cells exhibited complete structural deformation across all culturing conditions by day 6. The retinal neurospheres made with 10,000 cells matured in serum-containing medium on either PLO-Laminin or Matrigel™ coated surface also showed significant deformation. However, retinal neurospheres matured in medium without serum on either PLO-Laminin or Matrigel™ coated surface showed greater structural integrity. The retinal neurospheres made with 15,000 cells maintained the highest structural integrity after 6 days of maturation on various coated surface and media.
Matrigel™ and PLO-laminin coated surfaces produced significantly different effect on retinal neurosphere integrity profile over 6 days. Maturation on Matrigel™ resulted in 4 out of 6 retinal neurospheres showing significant structural integrity decline with significant deformation and size reduction after 2 days. By the 6th day, 5 out of 6 maturation conditions resulted in near-complete deformation of retinal neurospheres. On the other hand, PLO-laminin coated surfaces support retinal neurosphere structural destabilization with all 6 retinal neurosphere maturation conditions showing no sign of structural destabilization after 2 days.