Dual SMAD inhibition and Wnt inhibition enable efficient and reproducible differentiations of induced pluripotent stem cells into retinal ganglion cells

Glaucoma is a group of progressive optic neuropathies that share common biological and clinical characteristics including irreversible changes to the optic nerve and visual field loss caused by the death of retinal ganglion cells (RGCs). The loss of RGCs manifests as characteristic cupping or optic nerve degeneration, resulting in visual field loss in patients with Glaucoma. Published studies on in vitro RGC differentiation from stem cells utilized classical RGC signaling pathways mimicking retinal development in vivo. Although many strategies allowed for the generation of RGCs, increased variability between experiments and lower yield hampered the cross comparison between individual lines and between experiments. To address this critical need, we developed a reproducible chemically defined in vitro methodology for generating retinal progenitor cell (RPC) populations from iPSCs, that are efficiently directed towards RGC lineage. Using this method, we reproducibly differentiated iPSCs into RGCs with greater than 80% purity, without any genetic modifications. We used small molecules and peptide modulators to inhibit BMP, TGF-β (SMAD), and canonical Wnt pathways that reduced variability between iPSC lines and yielded functional and mature iPSC-RGCs. Using CD90.2 antibody and Magnetic Activated Cell Sorter (MACS) technique, we successfully purified Thy-1 positive RGCs with nearly 95% purity.


Methods
Human iPSC culture. Undifferentiated iPS cells, Control 1 (CHOPWT8), Control 2 (CHOPWT9), and Control 3 (CHOPWT10) were derived and characterized as previously published showing a complete analysis of iPSC characteristics [34][35][36] . All human sample collection protocols were approved by the University of Pennsylvania and Children's Hospital of Philadelphia Human Subjects Research Institutional Review Board following Declaration of Helsinki. All methods were performed in accordance with the relevant research guidelines and regulations of University of Pennsylvania. Written informed consent was obtained from all human cell donors. The iPSC cells were maintained in iPSC medium (Dulbecco's modified essential medium/Ham's F12 nutrient media; DMEM/F12 (50:50; Corning)) containing 1% Glutamax, 1% penicillin/streptomycin (PS), 15% Knockout serum replacement (KSR), 1% non-essential amino acids (NEAA), 0.1 mM β-Mercaptoethanol (2-ME) (Life Technologies, CA), and 5 ng/mL of basic fibroblast growth factor (bFGF; R&D Systems) on 0.1% gelatin coated dishes with irradiated mouse embryonic fibroblast (iMEFs).
Voltage recordings. The circular cover glass coated with 1:100 growth factor reduced matrigel containing adherent cells was transferred to the recording chamber filled with Neurobasal media supplemented with 2% B27 without vitamin-A and 1% Glutamax. The chamber was placed on the microscope stage (Olympus BX-61 microscope) and perfused with oxygenated Ames's solution (Sigma-Aldrich) resulting in gradual replacement of the growth medium with the Ames medium. The temperature of the solution inside the chamber was gradually increased to 37 °C prior to imaging and recording session using Warner Instruments TC 344B. Confocal images of the cells were acquired with Olympus Fluoview 1,000 MPE system and cells demonstrating strong GFP expression were selected for patch-clamp recording. Whole cell configuration was achieved in the voltageclamp mode at − 60 mV holding potential. After achieving whole-cell configuration, cells were either maintained in the current clamp mode at zero holding current and depolarized with calibrated current steps, or in the voltage-clamp mode at − 60 mV holding potential from which they were depolarized with voltage steps. www.nature.com/scientificreports/ www.nature.com/scientificreports/ and housekeeping-TET) provided in Supplementary Table S4 and TaqMan Fast advanced Master Mix (Ther-moFisher) on a 7,900 Fast Real-time PCR system (Applied Biosystems). All results were normalized to a B2M housekeeping control and were from 3 technical replicates of 3 independent biological samples for each timepoint and experimental condition.
Magnetic activated cell sorting (MACS) to purify CD90 + ve RGCs. RGC cells were lifted using Try-pLE (Invitrogen), pelleted by centrifugation at 350×g for 5 min, and total cell number was determined. Cell pellet was resuspended in 90 μL buffer (1 × PBS pH 7.2, 0.5% BSA, and 2 mM EDTA) and 10 μL of CD90.2 microbeads (catalog # 130-121-278, Miltenyi Biotec) per 10 7 total cells. Cell suspension was mixed well and incubated at RT for 15 min in a tube rotator. In the meantime, MS column was placed onto a MACS separator and the column was prepped. Following the 15 min incubation, the cell suspension was applied onto the column. Flow-through from the column represented the unlabeled or CD90.2 -ve cell fraction. The column was washed with appropriate volume of buffer for at least twice. The column was then removed from the separator and placed on a suitable collection tube. Appropriate volume of buffer was added to the column and magnetically labeled CD90.2 + cells were immediately flushed out by firmly pushing the plunger into the column. The cells were plated using RGCs induction media containing 3 μM DAPT and 10 μM ROCK inhibitor.
Statistical analysis. Quantitative data were obtained from three independent experiments per cell line in triplicate. Statistical analysis was performed with Student T-test in Prism. *p-values of < 0.05 were considered statistically significant.

Results
Combined inhibition of Wnt, BMP, TGF-β and nicotinamide contributes to efficient generation of early RPCs. The hiPSC retinal differentiation methods can differ between 2 and 3D culture conditions, and their use of proteins and small molecules to recapitulate processes responsible for vertebrate retina development 26,41 . Previously, the generation of RGCs from hPSCs required that cells be maintained on matrigel as an adherent support to culture cells obtained from 3D cell aggregates and several additive factors like Noggin, DKK-1, IGF-1, bFGF2, and proneural supplements 37,42,43 . Teotia et al., successfully used a modified version of the "Lamba protocol" by differentiating hiPSCs initially into neural rosettes, manual picking, and then induction to functional RGCs, by recapitulating retinal developmental pathways using chemical regulators 26,37 . These conditions require mid-differentiation enrichment steps that involves the establishment of retinal rosettes followed by manual isolation and maturation to generate retinal subtypes. CRISPR engineered iPSC lines containing fluorescent RFP reporter tag in the Brn3B locus, greatly assisted in evaluation of pathways necessary for RGC differentiation and characterization 44 . This methodology provided a protocol which utilized a monolayer cultures with defined factor supplementations; however, the evaluation were only performed using human embryonic stem cells (hESCs) and resulted in proportions of RGCs between 20 and 30% of the overall retinal differentiation. A major challenge in the regenerative medicine and disease modeling field are the reproducibility between experiments, and variation between individual to individual. Therefore, we set out to develop and characterize a modified two-stage protocol that differentiates hiPSCs into an enriched population of retinal progenitor cell (RPC) cultures followed by targeted differentiation to RGCs that is reproducible, efficient, and requires minimal personnel interpretation in RGCs generation and maintenance 26 . To accomplish this, hiPSCs were grown to confluence and subsequently treated with a RPC induction media containing: DMEM/F12 plus N2, B27, XAV939 (WNT inhibitor), SB431542 (TGF-β inhibitor), LDN193189, (BMP inhibitor), nicotinamide, and IGF1 for 4 days (Fig. 1A). The inhibition of Wnt and BMP signaling has been documented to enhance the expression of eye field transcription factors (EFTFs) during retinal differentiations of hPSC 27 . We observed that addition of TGF-β inhibition induced greater EFTFs expression during early retinal differentiation. Nicotinamide was added to the differentiation media (D0-D3) to promote the expression of early eye field markers LHX2 and RAX, as previously published 45 . Nicotinamide has been shown to promote cell expansion and adaptation to a radial/ rosette morphology 46 . Differentiation factors such as IGF-1 and bFGF2 aid in the specification of eye field identity to differentiating retinal progenitors 27 . From Day 4-21, nicotinamide was removed and bFGF was added to RPC induction media. Analysis at day 7 showed an uniform population of SOX2, RAX and PAX6 positive cells (Fig. 1B). The expression of early retinal progenitor markers, LHX2 and RAX, were identified in over 95% of day 7 cultures (Fig. 1C) indicating an efficient and robust generation of RPCs. Quantification of EFTFs, LHX2, RAX, PAX6, SIX6, SIX3, and VSX2, showed significant expression relative to pluripotent stem cells in all three individual cell lines (Fig. 1D).
The efficient induction of retinal progenitors is largely dependent on the inhibition of pathways responsible for patterning of diencephalon; however, it is unclear if modification of these inhibitory pathways play a role on retinal ganglion cell generation. Here, we tested whether pharmacological inhibition of Wnt, TGF-β, and BMP can improve RGC induction and subsequently enable the controlled generation of mature iPSC-RGC neurons. We evaluated seven different treatment conditions to determine pathways that enhance or restrict RPC generation from iPSCs. The following 7 conditions were tested from d16-d21: (1) BMP (L), (2) Wnt inhibition (X), (2) late stage RGC differentiation (D22-36) ( Fig. 2A,B) 47 . On day 24, cultures were expanded using a cross-hatching technique known to separate cells into small clusters and SHH, TGF-β, and notch signaling were inhibited to promote RGC maturation using cyclopamine, follistatin, and DAPT, cyclopamine was removed for the following 2 days. To improve survival of RGCs during the expansion 48 , cAMP, BDNF, neurotrophin 4 (NT4) and CNTF, were added to the media to promote the development and survival of cells with neuronal fate 49 . Forskolin and Y-27632 were included as compounds shown to promote RGC neurite growth, survival and lineage commitment 41 . These conditions were established by combining previously described conditions used by Sluch et al., 2015 and Teotia et al., 2017 for hiPSC-RGCs. After 35 days in vitro, cultures were disassociated and flow cytometric analysis were performed on single cell suspension obtained from three independent iPSC lines, and data showed that highest proportions of RGCs were generated from the LSBX conditions, with a range between 40 and 50% Thy-1 (CD90) (Fig. 2C,E), 82-84% BRN3B, and 11-12% RBPMS ( Supplementary Fig. S1). The LX, CHIR, and LSB conditions resulted in 26.7% (± 3.2), 22.5% (± 4.7), and 27.2% Thy-1 positive cells (± 6.2), respectively (Fig. 2C). We observed consistent results across individuals and experiments (Fig. 2D). Additionally, on Day 35 we further concentrated the number of RGCs in our culture by employing the MACS technique and CD90.2 microbeads. Since RPCs are multipotent cells, they have the potential to differentiate into neuronal cell types and one glial cell type called the Müller glial cells 50,51 . Therefore, we utilized the CD90.2 microbeads for the positive selection of RGCs expressing Thy1 cell surface marker and the removal of Müller glial cells from our culture. Following the purification of RGCs using MACS, the cells were analyzed using immunocytochemistry by detecting for the presence of Müller glial cells, astrocytes, and RGCs using CRALBP, GFAP, MAP2, TUJ1, RBPMS and BRN3A markers, respectively (Fig. 3). Our results show that roughly 95% of the cells in our culture were positive for BRN3A with the presence of extended synaptic connections between RGCs. Whereas, about 5% of the cells were positive for astrocytes and we detected no presence of Müller glial cells in our system (Fig. 3). However, the presence of astrocytes is significant since they are essential for the functional activity of the neuronal cell types by ensuring proper synaptic maturation and signaling 52 .
For long-term cultures, we initially used RGC induction supplemented with 1% N2 and 1% B27 for the maintenance of RGC cultures after DIV35; however, this resulted in significant overgrowth of proliferative late RPCs. The removal of N2 supplementation reduced the number of dividing cells in our cultures, providing a culture with predominantly RGCs. We have also cultured RGCs in 1% CultureOne supplement (Life Technologies) in the RGC maturation media, which reduced the number of non-RGC cell growth in day 45 (only in two weeks, from Day 27-40) cultures and beyond. After Day 35 in culture, the iPSC derived RGCs produce appropriate morphological and physiological features of mature RGCs.

Gene expression profiles during differentiation of iPSC to RPCs and RGCs.
During eye formation in vertebrates, cell intrinsic signals, extrinsic signals and/or transcription factors control the differentiation and fate determination of retinal cells. We evaluated the gene expression profile of three EFTs, RAX, PAX6, and SIX3 that play a role in the anterior neural plate (Fig. 4). The expression of Rx (encoded by RAX gene) was maximum in the RPC inhibited by BMP and Wnt inhibition when compared to the other conditions at DIV23 (Fig. 4, Supplementary Table S5). The RAX expression at DIV35 RGCs was minimal suggesting a commitment to a more differentiated retinal fate, a consequence of retinal progenitor cell (RPC) expansion. PAX6 is expressed in the cornea, lens, ciliary body, and retina through development and plays a role in determining their cell fate. The PAX6 transcript expression was observed in all experimental conditions in RPCs and RGCs; however, predominant expression of PAX6 is detected at DIV23 and DIV35 in the CHIR condition (Fig. 4), which stimulates the canonical Wnt signaling. Our results indicate that prolonged stimulation of RPCs with Wnt restricts their differentiation potential and maintains majority of the cells as multipotent progenitors.
Among other gene transcripts analyzed, we observed an increase in SIX3 expression in both LX and LSBX conditions at DIV23 days, when compared to RPCs stimulated with LSB or CHIR. The expression of SIX3 decreased in RGCs at DIV35 indicating that neurospecification was reaching completion at this stage.
The SOX11 transcript is heavily expressed in developing retina during embryonic stages 53 . It is required for the maintenance of hedgehog signaling and is critical for axonal growth, extension and driving adult neurogenesis [54][55][56] . The expression of SOX11 significantly increased in LSBX and LX conditions when compared to LSB and CHIR in RPCs at DIV23. The expression of SOX11 was significantly reduced in RGCs at DIV35 (Fig. 4). GLI3 has a dual function as a transcriptional activator and a repressor of the sonic hedgehog (Shh) pathway. GLI1 is a simple transcriptional activator encoded by a target gene of Shh signaling. We observed increased expression of GLI3 in RPCs at DIV23 when compared to RGCs at DIV35. Expression of both GLI3 and GLI1 genes demonstrated that sonic hedgehog signaling is important for development of RGCs.
Direction selective RGC (DS-RGC) are subtypes of RGCs that respond to motion of light in different directions can be identified by expression of specific molecular markers, such as CART , CDH6, and FSTL4, among other genes [57][58][59][60] . The CDH6 mediates axon-target matching and promotes wiring specificity that does not lead to image formation in the mammalian visual system. Cadherin mediated cell-cell adhesion ensures precise connectivity of neurons in the eye to target nuclei in the brain. Increased expression of CDH6 was observed in LSBX condition in DIV35 RGCs demonstrating their maturation towards increased specificity for axonal wiring between RGCs. The CARTPT (cocaine-and amphetamine-regulated transcript) is expressed by a major subtype of RGCs, ooDSGCs. In our study, the expression of CARTPT was seen in RGCs matured only in LSB, CHIR and LSBX conditions at DIV35 indicating that these conditions develop specific subtypes of mature RGCs that are known to be markers for ON-OFF direction-selective RGCs 61 . FSTL4 is a gene expressed in ON DS-RGCs and is colocalized with BRN3B in few RGCs 43 . We have seen increased expression of FSTL4 gene transcript in CHIR Scientific RepoRtS | (2020) 10:11828 | https://doi.org/10.1038/s41598-020-68811-8 www.nature.com/scientificreports/ induced RGCs when compared with other conditions at DIV35 indicating the development of various subtypes of mature RGCs in our culture conditions. We did not observe expression of CARTPT or FSTL4 transcripts in RPCs at DIV23 (Fig. 4). www.nature.com/scientificreports/ The RGCs at DIV35 expressed BRN3A transcript predominantly in LSBX growth condition when compared to other conditions. Interestingly, ATOH7 expression was observed early in RPC differentiation at DIV23 and decreased in RGCs by DIV35. We detected low CRX expression levels in DIV35 RGC cultures across all culture www.nature.com/scientificreports/ conditions implying that the RGC differentiation conditions at this stage restrict the photoreceptor precursor cell populations. We observed low expression of RCVRN (expressed by photoreceptors) and MITF in our cultures at DIV23 and 35 indicating that our cultures are differentiated predominantly towards RGC fate, with minimal retinal pigment epithelium cell identity (Fig. 4). Development of few interneurons in mature RGCs (amacrine or horizontal cells) was also observed, as evidenced by expression of CALB2 especially in RGCs with Wnt activation at DIV35. To understand the importance and requirement of BMP, Wnt, TGF-β inhibition in differentiation and maintenance of RPCs, we performed pairwise gene expression analysis of the RPC cultures at DIV23 for the four conditions (L, SB, X and CHIR) in different combinations (Supplementary Table S5). We observed increased expression of RAX, PAX6, LHX2 and SOX11 at DIV23 in L and X when compared to LSB, LSBX and LSB-CHIR suggesting that BMP and Wnt inhibition is required for maintaining majority of cells in retinal progenitor state. High expression of CRX, and RCVRN at DIV23 in L, X and LSB when compared to LSB-CHIR suggests that Wnt signaling may play role in preventing the differentiation of RPCs toward photoreceptors lineage. Decreased expression of CALB2 at DIV23 in X (Wnt inhibitor) condition when compared to LSB, LSBX and LSB-CHIR; while increased expression at DIV35 under CHIR (Wnt agonist) condition suggests that Wnt signaling is crucial for promoting RPC differentiation to RGC. Increased expression of GLI1 and GLI3 transcripts at DIV23 in L ( BMP inhibitor) condition when compared to LSB, LSBX and LSB-CHIR suggests that BMP inhibition is important for promoting SHH signalling (Supplementary Table S5). Two out of the six cells tested demonstrated action potential firing in the current clamp mode in response to the depolarizing currents. The responses of one of these cells (cell a) are illustrated in Fig. 5B. Although cells are not expected to fire action potentials under ideal voltage-clamp conditions, we believe that imperfect clamping in distal processes can lead to such firing observed as illustrated in Fig. 5C for the same cell a from Fig. 5D. Interestingly, all but one cell (6 out of 6 and 4 out of 5 DIV35 old cells) demonstrated reliable firing when depolarized with voltage steps in the voltage-clamp mode. In Fig. 5B, cell b illustrates firing of the older cells and cell c illustrates firing of the younger cells in response to the depolarizing voltage steps. As expected, later stage iPSC-RGCs cultures (D75) produced higher frequency, sustained firing and generated larger spikes when compared to early born iPSC-RGCs.
Cells firing under current-clamp conditions had resting membrane potentials around − 50 mV. In contrast, cells that fired under voltage-clamp but not current-clamp conditions had resting potentials around or above − 30 mV. In voltage clamp mode membrane potential was maintained at − 60 mV allowing more effective recovery of sodium channels from inactivation after depolarizing steps, thus enabling action potential firing in response to depolarization.

Discussion
In the current study, we employed a two-step/stage differentiation to induce RGC differentiation from iPSCs by modification of existing methodologies 26,27,43 . The first stage involved differentiation of iPSCs to RPCs. The RPCs were matured in a stage-specific manner using small molecules and recombinant proteins to modulate SHH pathway, Wnt pathway and Notch signaling to produce abundant RGCs reliably, which stained positive for RGC markers and emulated action potential.
This method is a quick and efficient RGC generation protocol without the need for 3D aggregate formation or manual enrichment to initiate RGC differentiation. In our method, the entire hiPSCs monolayer was differentiated to RPCs using a chemically defined medium in 2D cultures. We employed crosshatching technique to generate clumps of cells that underwent stage-specific differentiation to produce functionally mature RGCs by DIV28. This provided an accelerated timeline considering other published methods to date 22,26,44,62,63 . Our methodology involved using a chemically defined media standardized by Teotia et al., with minor modifications 26 , to take advantage of 15-day RGC differentiation timeline and pairing it with our novel RPC generation protocol to produce RGCs. The RPCs generated showed immunoreactivity to RAX and LHX2 in the RPC lines for over 97% of cells in iPSC culture indicating that our protocol committed cells towards RPC lineage. Using our method, we reliably differentiated six normal iPSC lines and multiple clones from those lines (data not shown) to generate mature RGCs that stain positive for Thy1/Tuj1, BRN3A, BRN3B, TUBB3, RBPMS, MAP2 and SNCG markers. Using different chemical conditions, we generated over 80% of pure iPSC-RGC cultures. Using FACS sorting analysis, we quantified the presence of Thy1 positive RGCs (~ 58%), BRN3B positive (~ 84%) and RBPMS positive (~ 12%) iPSC-RGCs in our differentiated cultures. Howerver, there is a discrepancy between the FACS quantification and immunocytochemistry (ICC) results for RBPMS positive RGCs in our cultures that could be primarily due to differences in the antibodies used for FACS and ICC and their antigen recognition abilities. Although we are certain about RBPMS antibody used for ICC as it is well characterized in previous studies 64 . Therefore for FACS analysis we will be testing other RBPMS antibodies in future to improve the efficacy of iPSC-RGC detection by this technique.
In addition to the matured RGCs, we also generated fewer other retinal cell types that express pan-retinal markers and appear to be astrocytes, amacrine and/or bipolar cells. Further characterization of these cells is www.nature.com/scientificreports/  (Fig. 4B, cell c) and maturation DIV75 (Fig. 4B, cells a and b). Blue arrows point to cells selected for patch clamp recordings and that fired action potential upon depolarization. (C) Action potential firing in response to depolarizing voltage step recorded from cell a and cell b (DIV75) and from cell c (DIV35). (D) This trace demonstrates action potential firing in response to the two different depolarizing steps of current recorded from cell a from Fig. 4B. We also observe a spontaneous spike firing at the start of the trace. Same scale bar applies to all images. www.nature.com/scientificreports/ needed to confirm their identity. We did not produce RPE and photoreceptor outer or inner segments in our differentiations.
Here, we also show valuable data representing a heterogenous population of RGCs that can be characterized by cell type specific gene expression. Future work will be to further characterize in vitro conditions that will better segregate RGC cell patterning of these subtypes, and to perform single-cell RNA sequencing to confirm cell identity. This will be extremely useful in providing a runway for identification of possibly new surface markers specific to RGC subtypes, which may allow for selection using magnetic bead isolation, fluorescent activated cell sorting or immunopanning of iPSC derived RGCs.
Furthermore, the RGCs differentiated using the LSBX condition exhibited electrophysiological function, with the ability to conduct sodium and potassium through voltage-dependent channels and fire action potentials. When comparing action potentials obtained from RGCs at day 35, day 75 and day 110, we observed that RGCs were able to fire continuously producing larger spikes with higher frequency, as they matured and aged in culture when compared to younger cells. Based on the nature and type of physiological responses, several RGC subtypes like ON-OFF-and alpha-RGCs were observed. Therefore, our validated methodology can reliably harness iPSC technology as a renewable source of RPCs to efficiently produce highly enriched populations of RGCs for in vitro studies of glaucoma and potential therapeutic modalities for incurable RGC-related diseases.

Conclusion
In this study, we present a reproducible and efficient chemically defined in vitro methodology to generate unprecedented yields of RPC populations from multiple iPSCs, that are then directed toward the RGC lineage. We differentiated multiple iPSC lines into RGCs in a step-wise manner using small molecules and peptide modulators by inhibiting bone morphogenetic protein, TGF-β, and canonical Wnt pathways. Purified populations of these mature iPSC-RGCs have the potential for in vitro studies of glaucoma and for therapeutic purposes for many RGC-related diseases.