Generation of Retinal Ganglion Cells With Functional Axons From Mouse Embryonic Stem Cells and Induced Pluripotent Stem Cells

Abstract

Purpose: We previously generated self-induced retinal ganglion cells (RGCs) with functional axons from human induced pluripotent stem cells (hiPSCs). We investigated whether self-induced RGCs from mouse embryonic stem cells (mESCs) and induced pluripotent stem cells (miPSCs) are realized by the similar induction protocol.

Methods: Retinal ganglion cells were induced using a protocol in which floating embryoid bodies (EBs) were differentiated into a retinal cell lineage in three-dimensional culture and subsequently attached to two-dimensional culture dishes with brain-derived neurotrophic factor (BDNF) supplementation.

Results: Retinal ganglion cells developed in an attached clump of cells originating from the optic vesicle, and most axons grew from RGC cell bodies at the margins of the clump. The differentiation of RGCs was confirmed by the expression of specific markers, including Brn3a and Math5. The axons contained neurofilament subtypes and tau, and manifested axonal transport and sodium-dependent action potentials. The RGCs derived from mESCs and miPSCs generally showed similar profiles, including RNA and protein expression levels and function.

Conclusions: Retinal ganglion cells generated from mESCs and miPSCs, especially the latter, may contribute to research associated with RGCs and to in vitro analyses of genetically modified mice.

The optic nerve transfers electrical signals for vision from the neural retina to the brain. Numerous genetic and etiologic factors contribute to the pathogenesis of optic nerve diseases, including glaucoma,¹ optic neuritis,² congenital anomalies,^3,4 hereditary diseases,^5,6 trauma, and drug-induced neuropathy, all of which significantly threaten vision. Glaucoma, the leading cause of irreversible blindness,¹ results from injury of the axons and gradual loss of retinal ganglion cells (RGCs) associated with high IOP.^7,8 This disease has been modeled in several animals using corticosteroid eye drops,⁹ microsphere injections into the outflow passage,¹⁰ or obstruction of the episcleral vein, the extraocular outflow passage,¹¹ to increase IOP. In addition, genetically engineered animals have been generated, such as transgenic mice with the MYOC gene mutation¹² and glutamate/aspartate transporter-deficient mice.¹³ Optic neuritis, another important cause of blindness, originates as an axon injury caused by infection, and autoimmunity against myelin or aquaporin 4, which triggers inflammation and other insults.^14,15 This disease has been studied using autoimmune encephalomyelitis (EAE) models.¹⁶ Congenital genetic anomalies of the optic nerve, represented by optic nerve hypoplasia, occur due to abnormalities in the morphogenesis genes, including HESX1,¹⁷ SOX2,¹⁸ and Pax6.¹⁹ Autosomal dominant optic atrophy (ADOA) and Leber hereditary optic neuropathy, other major hereditary optic neuropathies that occur after birth,⁶ are associated with mutations in the OPA1 gene²⁰ and mitochondria DNA,^21,22 both of which result in dysfunction of the mitochondria in axons. Further, optic nerve trauma has been mimicked by mechanically crushing the optic nerves, and toxic drugs, including monosodium glutamate (MSG) and N-methyl-d-aspartate (NMDA), have been injected into the vitreous cavity to affect the surface of the retina.^23,24 In most of the diseases described above, the axons are either injured or exhibit developmental abnormalities. 

Such animal models have been beneficial in investigating pathophysiologic mechanisms in vivo, particularly for elucidating signaling interactions between multiple cell types in tissues and their extracellular matrix. In vitro, purified RGCs and a mixed retinal culture system have been used to screen drug therapies.^25–27 Although the isolated RGCs²⁸ exhibit neurites, including dendrites and axons, which are an important characteristic of RGCs, these neurites of living cells are indistinguishable unless being fixed and stained by MAP2 for dendrites and TAU for axons, vary in length, and can be sustained for only a few weeks, making in vitro experiments on the relationship between axonal stress and apoptosis of RGCs difficult to evaluate. 

The generation of RGCs from embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs) may solve the problem described above. Although RGCs with characteristic axons derived from mouse ESCs (mESCs) or iPSCs (miPSCs) have been generated, the RGCs were not self-induced but rather induced by transfer of exogenous genes, such as Rx²⁹ and Math5.³⁰ Regeneration of self-induced whole retinal layers, including RGCs from human and mouse stem cells, has been achieved, but the RGCs in these studies did not exhibit characteristic axons,^31,32 probably because the generated primordium structures of the retina in these experiments did not have the optic stalk into which RGCs elongate their axons. Retinal ganglion cells with functioning axons have been derived from human iPSCs (hiPSCs), but the axons were limited and only 30% of the cell population differentiated into RGCs.³³ Recently, we generated RGCs with functioning axons from hiPSCs³⁴ by using a three-dimensional (3D) retinogenesis system from stem cells³² and subsequent conversion to a two-dimensional (2D) system. Retinal ganglion cells were developed from the margin of the optic vesicles (OVs), the primordial structures of the retina. The RGCs had characteristic axons and physiologic functions, such as axonal transport and action potential.³⁴ The RGCs emerged from the OVs at a rate of 90% and survived for approximately 50 days, suggesting that RGC generation is a promising tool in the in vitro study of the pathophysiology of optic nerve diseases. Retinal ganglion cells generated from hiPSCs may be valuable for investigating the pathophysiologic mechanisms of human diseases and evaluating the effect and safety of drugs, especially through the use of disease-specific iPSCs generated from the patients' own cells. On the other hand, RGCs with functioning axons generated from animal ESCs or iPSCs might help to establish an in vitro experimental system, specifically when using iPSCs generated from genetically engineered animals. Therefore, we here investigated the generation of RGCs from mESCs and miPSCs. 

Materials and Methods

Mouse Stem Cell Culture

Mouse ESCs were purchased from MTI-GlobalStem (Gaithersburg, MD, USA), and miPSCs (APS0007_iPS-Hep-FB/Ng/gfp-103C-1 cell passage 10) were obtained from RIKEN BRC (Ibaraki, Japan).³⁵ 

Mouse ESCs were maintained in 15% KnockOut Dulbecco's modified Eagle's medium (DMEM) containing Glutamax (Thermo Fisher Scientific, Waltham, MA, USA), nonessential amino acids, 0.5 mM 2-mercaptoethanol, 100 U/mL penicillin, and 100 μg/mL streptomycin on feeder layers of mouse embryonic fibroblast (MEF) cells (REPROCELL, Kanagawa, Japan). Mouse iPSCs were maintained in DMEM containing 15% fetal bovine serum (FBS), nonessential amino acids, 1 mM sodium pyruvate, 0.5 mM 2-mercaptoethanol, 100 U/mL penicillin, and 100 μg/mL streptomycin, on MEF cells. 

For passaging mESCs and miPSCs, colonies were detached and recovered from the feeder layer by treatment with StemPro Accutase (Life Technologies, Waltham, MA, USA) at 37°C for 3 minutes. The detached cell clumps were dissociated into single cells by pipetting. Cell passaging was performed at a 1:20 to 1:50 split ratio. Each passage was performed every 3 days. 

Induction of Differentiation to RGCs

The procedure used for the induction of mESCs and miPSC differentiation was based on SFEB methods.³⁶ After dissociation into single cells with 0.25% trypsin-EDTA, the cells were resuspended in retinal differentiation medium (RDM; G-MEM supplemented with 1.5% knockout serum replacement [KSR], 0.1 mM nonessential amino acids, 1 mM pyruvate, 0.1 mM 2-mercaptoethanol, 100 U/mL penicillin, and 100 μg/mL streptomycin). After separation from the feeder cells (the feeder cells adhered to the gelatin-coated bottom of the dish) by decantation, the floating cells were collected from the medium and seeded into V-bottomed low cell–adhesion 96-well plates (Sumitomo Bakelite Co., Ltd, Tokyo, Japan) at 4,000 cells per well using a Countess Automated Cell Counter (Invitrogen, Carlsbad, CA, USA). Matrigel (growth factor-reduced; BD Bioscience, San Jose, CA, USA) was added at a final concentration of 1% in the medium on day (D) 1; the day the suspension culture was initiated was defined as D0. On D7, the aggregates were transferred to retinal maturation medium (RMM; DMEM/F12-Glutamax medium containing N2 supplement, 100 U/mL penicillin, and 100 μg/mL streptomycin) and then were cultured in the absence of FBS. The adhesion culture started on D14 when the aggregates were transferred to poly-d-lysine/laminin–coated 24-well plates (BD Bioscience) in RMM containing FBS (1%–10%) and 100 ng/mL brain-derived neurotrophic factor (BDNF; R&D Systems, Minneapolis, MN, USA). Retinoic acid (0.5 μM, Sigma-Aldrich, Corp., St. Louis, MO, USA) and 1% FBS were added to RMM 4 days before the start of adhesion culture. 

Real-Time Reverse Transcription-Polymerase Chain Reaction (RT-PCR)

Real-time RT-PCR was performed as described previously.³⁴ Briefly, total RNA was extracted from cells using an RNeasy Mini Kit (QIAGEN, Hilden, Germany), and RT-PCR was performed using a One Step SYBR PrimeScript PLUS RT-PCR Kit (TaKaRa Bio, Inc., Shiga, Japan) and StepONE Sequence Detection System (Applied Biosystems, Foster City, CA, USA). The primers used in this study are listed in Supplementary Table S1. The amount of mRNA relative to that specific to each of the target genes at D0 was calculated using the 2^–ΔΔCT method. Error bars indicate ± SD. Each column shows an average value for samples. The sample size for all mRNA data was three (n = 3). 

Immunohistochemistry and Histology

Immunostaining was performed as described previously³⁴ using frozen sections or whole fixed cells. The primary antibodies used in this study and their dilutions are listed in Supplementary Table S2. Specimens mounted with Pro Gold Antifade Reagent with DAPI (Life Technologies) were viewed with an IX71 inverted research microscope (Olympus, Tokyo, Japan) and DeltaVision ELITE (CORNES Technologies Ltd, Tokyo, Japan). Frozen sections were prepared as described previously.³⁴ Hematoxylin and eosin staining of 3-μm thick specimens was performed as described previously.³⁴ 

Transmission Electron Microscopy

Specimens were prepared as described previously.³⁴ Ultrathin sections of representative areas were stained with uranyl acetate and lead citrate, and viewed using a JEM-1200EX electron microscope (Japan Electron Optics Laboratory, Tokyo, Japan). 

Axonal Flow Observation

The time series of anterograde axonal flow was conducted by injection of Alexa Fluor 555-conjugated cholera toxin subunit B (Life Technologies) into the RGC region. A time-lapse analysis was performed immediately after the injection of cholera toxin B with an IX71 inverted research microscope (Olympus) and DeltaVision ELITE (CORNES Technologies). 

Electrophysiologic Recording

Whole-cell patch-clamp recording was performed as described previously.³⁴ Briefly, colonies were cultured on mixed cellulose ester filter paper (0.2-μm pore size; ADVANTEC, Tokyo, Japan) for 7 to 10 days. Whole-cell patch-clamp recordings were made from the RGCs, which exhibited repetitive action potentials, located on the outer perimeter of the cultured colony. The recording pipettes (6–8 MΩ) were filled with an intracellular solution described previously.³⁴ Morphologies of the recorded cells were visualized by Lucifer Yellow (LY), which is a fluorescent biological dye. Liquid junction potentials (−11 mV) were corrected. R_s was compensated at 40%. Data with R_s > 50 MΩ were excluded from analyses. The average membrane capacitance during the recordings was 12 ± 5.7 pF in mESC-derived RGCs (n = 3) and 11 ± 9.2 pF in miPSC-derived RGCs (n = 3). All data are presented as mean ± SD. Statistical analysis was performed using Student's t-test (unpaired). 

Statistical Analysis

Values are expressed as mean ± SD. All statistical analysis were performed using PASW statistics 18 (SPSS, Inc., Chicago, IL, USA). The statistical significance of difference was determined by 1-way ANOVA followed by Tukey's test or Games-Howell test for normally distributed variables and by the Kruskal-Wallis test followed by the Steel-Dwass test for variables that did not have normal distribution. Probability values less than 5% were considered significant. 

Results

Protocol for Self-Induction of RGCs From mESCs and miPSCs

For generating RGCs from mouse stem cells, we modified the 3D retinal regeneration protocol,³⁷ in which OVs from mESCs and miPSCs developed 7 days after induction. 

We adopted the protocol of mouse OV formation for generating retinal ganglion progenitor cells in suspension culture with minor modifications (Fig. 1). In this protocol, which is the same as our RGC generation protocol for hiPSCs,³³ the concentrations of Matrigel were halved (i.e., 1.0%) during suspension culture, compared to previous reports.^36,37 Embryoid bodies showed OV formation on D7, in agreement with a previous report,³⁷ and the culture conditions were changed to serum-free medium on the same day. To obtain RGCs with axons, we changed the OV culture conditions from floating (3D) to dish (2D) to promote attachment of cells on D14, similar to the procedure described earlier for the generation of RGCs from hiPSCs.³⁴ We directly attached embryoid bodies (EBs) with OVs to a dish coated with poly-d-lysine/laminin. The attached OVs were differentiated into an extruded clump of cells during supplementation of BDNF (100 ng/mL) to promote neuronal survival. After 2 days of attachment, axonal growth from the margin of the clump was observed. The axonal growth settled approximately 1 week after attachment. Retinal ganglion cells and their axons were stabilized on D30, whose length was 1723 ± 333 μm (n = 10) in mESC-derived RGCs and 1624 ± 371 μm (n = 10) in miPSC-derived RGCs. 

Figure 1

 

Schematic diagram of the protocol for induction of RGCs. This protocol consisted of a suspension culture (3D) phase followed by an adhesive culture (2D) phase, and resulted in axonal outgrowth from RGCs derived from mESCs and miPSCs within 22 days. Retinal differentiation medium was applied until D6 and RMM was used from D7. Optic vesicle–like small extrusions were observed by D14, and EBs, which were formed in 3D culture and accompanied the extrusions, were attached on D14. Radial axons grew from the margin of attached OVs. Scale bar: 500 μm.

Time Course Expression of Transcription Factors and Protein Markers Associated With Retinal Development

Optic vesicle is derived from the eye field in the anterior neural plate that is specified by so-called eye-field transcription factors, including the retinal homeobox gene (Rx) and the paired box gene 6 (Pax6).³⁸ Using mESCs and miPSCs, temporal gene expression analyses of retinal morphogenesis were conducted (Figs. 2A, 2C). In RGCs from mESCs, Rx was continuously upregulated until D22, and its downstream targets Pax6 and Chx10, which specify the sensory retina, were gradually upregulated (P < 0.05). In RGCs from miPSCs, Rx was upregulated until D14 (P < 0.05), and Pax6 and Chx10 were gradually upregulated with peak expression at D14 and D10 (P < 0.05), respectively. Math5, which is controlled by Pax6³⁹ and determines the fate of RGCs,^40–42 was expressed at all time points, consistent with Pax6 expression. Brn3⁴³ has three subtypes (Brn3a, Brn3b, and Brn3c) expressed in mouse RGCs and associated with cell specification and cell survival.^44,45 Although Brn3b expression was relatively low from D7 to D22 in RGCs derived from mESCs and miPSCs, Brn3b and Brn3a tended to be upregulated from D10, following initiation of Math5 expression. The expression of Crx, a cone–rod-encoding gene specific to the photoreceptor lineage,^46,47 was observed at D7 and continued until D22. Amacrine cells start to differentiate immediately after RGCs differentiation.⁴⁸ In the clump of cells derived from mESCs and miPSCs, syntaxin, an amacrine cell marker,^49,50 was upregulated gradually from D7 (P < 0.05). Typically, in the clump of cells derived from mESCs and miPSCs, calbindin, a horizontal cell marker, was expressed gradually from D10 (P < 0.05). Pkcα, a marker of bipolar cells, which develop at later stages of retinogenesis,⁴⁸ was expressed gradually by D22. The gene encoding MITF, the master regulator of melanocyte and RPE development,⁵¹ was seldom detected, suggesting that our protocol induced mESCs and miPSCs to differentiate into the neural retinal lineage but not into the RPE lineage. 

Figure 2

 

Commitment of mESCs and miPSCs to a retinal lineage and generation of RGCs. Time-dependent expression of genes encoding transcription factors involved in retinogenesis was examined during the period of mESCs (A, B) and miPSC (C, D) differentiation toward RGCs. (A, C) Relative mRNA changes during the differentiation were evaluated by quantitative RT-PCR. Total RNA was extracted on D7, D10, D14, and D22. Characteristic genes associated with development of RGCs were analyzed statistically. The relative amount of mRNA was measured based on that specific to each of the target genes at D0. The sample size is three (n = 3) for all columns. *P < 0.05. NS, not significant. Error bars: +SD. (B, D). Immunohistochemistry detecting RX, PAX6, MATH5, BRN3, and CRX in OVs on the floating EB on D14, and in the extruded clump of cells on D22. Scale bar: 100 μm.

Immunohistochemistry revealed time-dependent expression of marker proteins in RGCs derived from mESCs and miPSCs (Figs. 2B, 2D). RX and PAX6 were detected in OVs on D14 and were present until D22. MATH5 and BRN3 found throughout the clump of cells and were most evident on D22; their expression was relatively delayed following mRNA expression of both genes. In the clump of cells from mESCs, the CRX signal was weak on D14 and slightly more apparent on D22, which was consistent with mRNA expression. In the clump of cells from miPSCs, the CRX signal was weak on D14 and on D22, which differed from mRNA expression, indicating possible immature translation. Their distribution was random throughout OVs on D14 and the clump on D22 (Figs. 2B, 2D). 

Structure of RGCs Derived From mESCs and miPSCs, Analyzed by Light and Electron Microscopy

After the attachment of EBs with OVs on D14, axons radiating mainly from the outer margin of the clump of cells were observed. In the horizontal section of the hematoxylin and eosin (H&E)-stained samples (Figs. 3A, 3E), the cells in the clump were relatively larger (RGCs derived from mESCs: range 5.6–10.1 μm, mean 7.9 μm, n = 11; RGCs derived from miPSCs: range 7.5–10.8 μm, mean 9.0 μm, n = 11) than the cells in the EB. These findings were identical to those of RGCs in vivo, whose body size ranged from 9.2 to 11.1 μm.⁵² Nissl bodies, which are characteristic components of RGCs, were identified in the cell bodies (Supplementary Fig. S1). The axons were eosinophilic and grew from RGCs. On the vertical section of the H&E-stained samples (Figs. 3C, 3G), axons grew on the surface of the culture dish from the outer margin of the clump. 

Figure 3

 

Light and electron microscopy of induced RGCs derived from mESCs and miPSCs, and their axons. Horizontal (A, E) and vertical (C, G) H&E-stained sections revealed that axons mainly grew from the outer margin of the extruded clump of cells (A, E, G), but sometimes originated directly from the attached EB itself (C). Stomata of new RGCs were distributed throughout the clump of cells, with variations in size and relatively large nuclei compared to EBs. (B) Electron microscopy analysis of the square portion (b) in (A) revealed that the cell bodies of mESC-derived RGCs contained several rough endoplasmic reticula (arrows) and that axons (arrowheads) developed from a portion of the axon hillock. (F) The cell bodies of miPSC-derived RGCs, the square portion (f) in (E), contained prominent mitochondria (*) in the cytoplasm, and the axon (arrowheads) grew from the axon hillock and contained mitochondria. (D) Electron microscopy analysis of the square portion (d) in (C) revealed that straight axons developed from the mESC-derived RGC and contained numerous microtubules. (H) Electron microscopy analysis of the square portion (h) in (G) showed that the axons developed straightly from the miPSC-derived RGC and contained a number of neurotubules. Scale bars: 100 μm, in (A, C, E, G); 2 μm in (B, D, F, (H).

Electron microscopy of D22 specimens of RGCs derived from mESCs revealed that the RGCs contained rough endoplasmic reticula, which represented the Nissl bodies that were observed using the light microscope, in a portion of the axon hillock (Fig. 3B). Electron microscopy of D22 RGCs derived from miPSCs also revealed the presence of several mitochondria in the hillock (Fig. 3F). The axons developed from the top of the axon hillock (Figs. 3B, 3F), and mitochondrial transport also was observed in the axons of RGCs derived from miPSCs (Fig. 3F). The axons were composed of numerous neurotubules. The diameter of the axons varied (RGCs derived from mESCs: range, 0.14–0.24 μm; mean, 0.17 μm; n = 10; RGCs derived from miPSCs: range, 0.13–0.25 μm; mean, 0.20 μm; n = 10), and were not myelinated (Figs. 3D, 3H). These findings were identical to those of axons in the retinal nerve fiber layer and in the optic nerve anterior to the lamina cribrosa, whose diameter ranged from approximately 0.06 to 0.28 μm.⁵³ 

Determination of the RGC Lineage

Reverse transcription-PCR revealed a more than 19-fold and 30-fold increase in the expression of Brn3a in RGCs derived from mESCs and miPSCs, respectively; a 4-fold increase in the expression of Brn3b in both RGCs was observed. Math5 was upregulated 42-fold and 8-fold in RGCs derived from mESCs and miPSCs, respectively; Islet1 expression was increased 55-fold and 18-fold in RGCs derived from mESCs and miPSCs, respectively; and Tuj1 was upregulated 52-fold and 14-fold in RGCs derived from mESCs and miPSCs, respectively; while γ-synuclein (Sncg), another RGC marker,⁵⁴ was upregulated by less than 5-fold. The expression of these markers showed similar patterns in RGCs derived from miPSCs and mESCs (Figs. 4A, 4C). 

Figure 4

 

Prevalence of characteristic markers of RGCs derived from mESCs and miPSCs. (A, C) The expression of representative RGC markers, Brn3a, Brn3b, Math5, Islet1, γ-synuclein (Sncg), and β3-tubulin (Tuj1). Quantitative RT-PCR analysis of total RNA extracted from mESC-derived RGCs and miPSC-derived RGCs on D22 was performed. The relative amount of mRNA was measured based on that specific to each of the target genes at D0. The sample size is three (n = 3) for all columns. Error bars: +SD. (B, D) Immunohistochemistry detection of BRN3, MATH5, ISLET1, SNCJ, and TUJ1 on D22. Whole-mount staining for TUJ1, serial section staining for the other targets, and double-staining for BRN3b and TUJ1 were performed. Scale bars: 100 μm. Scale bar for the double staining: 10 μm.

In RGCs derived from miPSCs and mESCs, immunohistochemistry revealed positive staining for all the markers associated with RGCs. All markers, except for TUJ1, stained throughout the clump of cells. Double staining of TUJ1 and BRN3 was detected in many cells located at the peripheral margin of the clump (Figs. 4B, 4D). It also was confirmed by triple staining of MATH5, 4′,6-diamidino-2-phenylendole (DAPI), and neuron-specific β-III tubulin antibody (TUJ1; Supplementary Fig. S2). 

RGC Axon-Related Proteins and Determination of Neurofilament Protein Types in Early RGC Genesis

Tau and neurofilaments (NFs) were used as axonal markers. Genes encoding all NF components, that is, NF-light (NFL), NF-medium (NFM), and NF-heavy (NFH),⁵⁵ were studied in parallel (Fig. 5A). Microtubules, NFs, microtubule-associated proteins (MAPs), and actin have crucial roles in axonal development.⁵⁶ Tau is a MAP that contributes to the axonal transport process of RGCs.⁵⁷ Axons of RGCs derived from miPSCs and mESCs were stained for tau (Figs. 5B, 5D). Tau expression was relatively higher in RGCs derived from miPSCs (Fig. 5C). Neurofilaments also have essential roles in radial axonal growth and in the ability to transmit action potentials.⁵⁸ In RGCs from mESCs and iPSCs, NFL and NFM were already expressed at D22, while the expression of NFH was quite low on D22 (Figs. 5A, 5C). Specimens from D22 also showed positive staining for NFL and NFM but weak staining for NFH (Figs. 5B, 5D). Baseline staining of these proteins at D15 was almost negative (Supplementary Fig. S4). Axonal growth cones were identified by GAP43 staining (Supplementary Fig. S3). 

Figure 5

 

The cytoskeleton in axons elongating from RGCs. (A, C) Quantitative RT-PCR analysis was used to investigate genes encoding nerve fiber markers, including Tau, NFL, NFM, and NFH, using total RNA extracted at D22. The relative amount of mRNA was measured based on that specific to each of the target genes at D0. The sample size is three (n = 3) for all columns. Error bars: +SD. (B, D) Immunohistochemistry probing for Tau, NFL, NFM, and NFH in RGCs derived from mESCs and miPSCs at D22. Scale bar: 100 μm.

Axonal Transport in RGCs Derived From mESCs and miPSCs

We tested for axonal transport in the apparent axons on D22. We observed anterograde axonal transport by injecting Alexa Fluor-conjugated cholera toxin B into the central portion of the clump. In RGCs derived from mESCs and iPSCs, the fluorescein stain spread in a time-dependent manner, although the transport seemed immature compared to that of RGCs from hiPSCs.³⁴ Moreover, the transport of RGCs from mESCs was more evident than that from miPSCs (Fig. 6; Supplementary Movie S1: RGCs from mESCs; Supplementary Movie S2: RGCs from miPSCs). 

Figure 6

 

Axonal transport in RGCs. In mESC-derived and miPSC-derived RGCs, axonal transport was assessed to investigate the anterograde flow, by injecting Alexa Fluor 555-conjugated cholera toxin B into the central portion of the clump of cells. Phase-contrast micrographs show the extruded clump of cells with radial growth of its axons. Anterograde transport was observed within 10 minutes, and cholera toxin B spread to the peripheral axons within 30 minutes. Scale bar: 100 μm.

mESC-Derived and iPSC-Derived RGCs show Neuronal Excitability by the Whole-Cell Patch-Clamp Technique

Each RGC had an axon process (arrows in Figs. 7A, 7B). We observed that the recorded cells generated repetitive action potentials in response to current injection from the resting membrane potential under the current-clamp mode (Fig. 7C, 7D). In mESC-derived RGCs (n = 9) and in miPSC-derived RGCs (n = 7), the resting membrane potential was −50 ± 15 mV and −57 ± 20 mV, respectively, and the amplitude of the first action potential was 62 ± 24 mV and 55 ± 17 mV, respectively (Fig. 7G). The cells that generated action potentials exhibited tetrodotoxin (TTX)-sensitive Na currents (Figs. 7E, 7F) with maximum amplitude of 1013 ± 571 pA in mESC-derived RGCs (n = 9) and 1276 ± 592 pA in miPSC-derived RGCs (n = 7), respectively (Fig. 7H). 

Figure 7

 

Action potentials of RGCs. mESC-derived RGCs (A) and miPSC-derived RGCs (B) filled with LY under fluorescence illumination. These cells exhibited an axonal process (arrows). P, recording pipette. Whole-cell recordings of the RGCs derived from mESC (C) and miPSC (D) revealed action potentials in current-clamp mode (50-ms current injection of +15 pA from the resting membrane potential). A family of currents was recorded in voltage-clamp mode in response to depolarizing steps from a holding potential of −71 mV to target voltages increasing from −61 to −1 mV in 20-mV increments (upper traces in [E] and [F]). The fast inward currents were blocked by 1 μM TTX (middle traces in [E] and [F]) and were recovered by washing (lower traces in [E] and [F]). Action potentials and inward currents were recorded from the LY-labeled cells shown in (A) and (B), respectively. In mESC-derived RGCs (n = 9) and miPSC-derived RGCs (n = 7), the amplitude of the first action potential was 62 ± 24 mV and 55 ± 17 mV (G), and maximum amplitude of Na current was 1013 ± 571 pA and 1276 ± 592 pA (H), respectively. There was no significant differences in spike amplitude and Na current between these cells (P > 0.05). Scale bar: 30 μm in (A, B).

Discussion

We demonstrated the generation of RGCs with functional axons from mESCs and miPSCs via the same method as reported previously,³³ albeit in a shorter time period. In a series of experiments, retinogenesis, characterized by the expression of retinal progenitor markers, including Rx, Pax6, and Chx10, was induced as OVs formed in a 3D culture system. All extruding vesicles constitute the OV and not the brain primordium, which was confirmed by positive staining for RX and PAX6 (data not shown). Generation of RGC was achieved in OVs, which were attached in the 2D system, during optic cup formation. Among the several cell types in the retina, RGCs begin to develop first from the retinal progenitor cells, just after optic cup formation.⁴⁸ Math5, the most characteristic upstream marker of RGCs, is reported to be upregulated beginning at E11, with its peak between E12.5 and E15.5, which is followed by the upregulation of its supposed downstream targets, Brn3b and Brn3a.⁵⁹ In the current study, Math5 expression was detected at D7, which was earlier than that reported previously, and the Brns tended to be upregulated from D10 until D22 following Math5 expression. Thus, the sequence of expression may mimic that observed in vivo. Hence, the attachment of OVs to the dish at that stage may be suitable for generating RGCs with axons, even in mice. In vertebrates, the eye-formation process through OVs and the subsequent optic cup formation are universal events,⁶⁰ and the sequence of cell genesis, which is initiated by RGCs, is highly conserved.⁴⁸ Hence, our method of RGC genesis may be applicable to other vertebrate species. 

Although generation of RGCs from mESCs and iPSCs was achieved using the same method as that used for the generation of RGCs from hiPSCs,³⁴ several differences were observed in each stem cell type. In RGCs derived from hiPSCs, most RGCs with axons were located at the peripheral margin of the clump of cells that developed from the OV and other retinal cell types, including amacrine cells and photoreceptor precursor cells in the inner part of the clumps.³⁴ However, RGCs derived from mESCs and miPSCs were located not only at the margin of the clumps but also in the inner portion of the clumps, where axons developed only from RGCs at the margin, indicating that RGCs developed more randomly in retinal cell clumps from mouse stem cells. As a marker of RGCs, Brn3b expression was relatively low compared to RGCs derived from hiPSCs; however, expression of Math5 and Brn3a was similar. 

The cell size and axon diameters of RGCs derived from mESc and miPSCs were within the normal range, although the number of axons apparently was lower than the prominent axons of hiPSC-derived RGCs.³⁴ Moreover, the density of Nissl bodies in the cytoplasm of RGCs was low, especially in RGCs derived from miPSCs. Immunohistochemistry also demonstrated that BRN3 and/or MATH5-positive cells were located not only in the peripheral margin of the clump but also in the clump without characteristic axons. Regarding the whole-cell patch-clamp recording technique, RGCs were difficult to approach, because their cell membranes were covered with mucus and premature RGCs were relatively smaller than RGCs derived from hiPSCs. The amplitude of action potentials was relatively low, and the repolarization of miPSC-derived RGCs was incomplete. Although axonal transports were clearly identified, axonal currents also were insufficient. These findings suggest that RGCs derived from mESc and miPSCs were relatively immature compared to hiPSCs-derived cells. Further improvements in the methodology to facilitate generation of more mature RGCs from mESCs and iPSCs may be needed. 

Regarding the differences between RGCs derived from ESCs and iPSCs, researchers have demonstrated significantly different gene expression profiles for these cells in an early gene expression study,⁶¹ while others have suggested that ES and iPS cells are indistinguishable in terms of gene expression.⁶² Recently, the amount and variety of noncoding RNAs (ncRNAs) were significantly different between mESCs and miPSCs derived from B lymphocytes, which may impact their clinical application.⁶³ Thus, it is difficult to simply compare differences in the differentiation of these two stem cells. However, in the experiment described herein, mESCs and miPSCs were induced to differentiate into RGCs using the same protocol. Gene and protein expression levels, morphologies, and functions were similar between RGCs derived from these two stem cell types, whereas RGCs from mouse iPSCs were relatively immature in terms of Rx expression and action potentials. In addition, the expression of characteristic RGC markers, Math5 and Brns, varied slightly between RGCs derived from mESCs and iPSCs, indicating that the total number of differentiated RGCs might differ between those derived from these stem cells. Tau in RGCs from miPSCs also was expressed more highly than that in RGCs from mESCs, which may indicate differences in structural components, development, or maturation processes between RGCs from mESCs and miPSCs. Such differences between the RGCs derived from mESCs and miPSCs should be studied further. 

Generation of RGCs from pluripotent stem cells provides a new in vitro experimental system to investigate the pathophysiologic mechanisms of optic neuropathies, evaluate drug effects, reproducing cell transplantation, and for basic neuroscience studies. In such a series of in vitro experiments, the use of RGCs derived from hiPSCs has advantages over the use of animal cells, especially in terms of investigating the pathophysiologic mechanisms in human disease-specific iPSCs generated from a patient's own cells, and evaluating the effect of drugs on human cells. Human disease-specific iPSCs can be generated not only from a patient's cells but also from normal hiPSCs transformed using a genetic modification system, such as the CRISPR/Cas9 system; however, off-target transfer of mutations is a concern, and disease-specific iPSCs derived from the patient's cells are ideal as an in vitro disease model. 

In contrast, as vast amounts of molecular and cellular biology data already have been compiled using retinal tissues from diseased and normal mice, RGCs derived from mouse stem cells also have great potential, and RGCs generated from miPSCs derived from genetically modified mice¹² enable the analysis of time-dependent molecular changes of diseased cells in vitro. In future reproductive medicine experiments, mouse eyes that are affected by the loss of RGCs may be treatable by using RGCs derived from allogeneically transplanted mESCs, and by RGCs derived from autologously transplanted miPSCs, reducing the rejection response. In basic sciences, including embryology, the developmental process of gene expression cascades has been studied thoroughly in the normal developing mouse retina. Retinal ganglion cells derived from mouse stem cells in vitro can be used for more precise time-series studies of normal development, including the processes of RGC differentiation into different cell types, axonal guidance, and remyelination by mixed culture with glial cells. Thus, RGCs derived from mESCs and miPSCs, with accompanying RGCs derived from human stem cells, are promising tools to elucidate the pathogenesis of optic nerve diseases and, especially, to treat ocular diseases. 

Our study had several limitations. Retinal ganglion cells from mESCs and miPSCs developed not only in the margin of the clump. Those inside the clump appeared to be immature, which may have resulted from less complete development of the neural retina from optic vesicles than that from RGCs generated from hiPSCs. In addition, RGCs were not purified and heterogeneous, and the axons were less mature and uniform with respect to their features than those from hiPSCs. 

Acknowledgments

Supported by a Grant from the Ministry of Health, Labor and Welfare (H24-Nanchi-Ippan-031), JSPS KAKENHI Grant Numbers 22390327 and 24659770, and a Grant of National Center for Child Health and Development 25-7. 

Disclosure: T. Tanaka, None; T. Yokoi, None; F. Tamalu, None; S.-I. Watanabe, None; S. Nishina, None; N. Azuma, None