Partial dedifferentiation of murine radial glia

Cellular reprogramming represents a promising strategy for both, deriving a patient-specific source for biomedical applications and cellular models with enhanced plasticity for developmental studies. Overexpression of Oct4, Sox2, Klf4, and c-Myc in murine fibroblasts yields induced pluripotent stem cells (iPSCs) that show functional equivalence to embryonic stem cells (ESCs) and can thus be differentiated into all cell types of the adult organism1,2. Recent publications suggest that the transient overexpression of the “Yamanaka factors” can induce destabilization of somatic cells and as a result, cell fate can be converted to other lineages as well by help of extrinsic stimuli. By this strategy fibroblasts have been directly converted into cardiomyocyte-like cells3, hepatocyte-like cells4 as well as neural progenitors5. Besides the Yamanaka factors, the overexpression of neurogenic transcription factors have been shown to efficiently modulate cell fate towards neural states. Vierbuchen et al. reported that the overexpression of transcription factors Ascl1, Brn2 and Myt1l results in the generation of 20% neurons from murine fibroblasts6. The same three transcription factors in combination with overexpression of NeuroD1enable transdifferentiation of human fibroblasts into neurons7. Human induced neurons could also be generated by overexpression of Ascl1 and Ngn2 in combination with pharmacological intervention. i.e. by dual SMAD- and GSK3β-inhibition8. Notably, direct conversion of fibroblasts into induced neurons generates a non-proliferative, post-mitotic cell population. More recently, direct neural conversion strategies have been implemented that allow derivation of stably expandable induced neural stem cells (iNSCs)9,10. Our group employed constitutive overexpression of Sox2, Klf4, and c-Myc together with strictly controlled Oct4 activation for 5 days only. With this approach we could generate stably proliferating iNSCs from murine embryonic and adult fibroblasts that are virtually indistinguishable from their NSC counterparts derived from either primary tissue or pluripotent cells. In particular iNSC exhibit complete silencing of the reprogramming factors. Moreover, they can be differentiated into neurons, astrocytes and oligodendrocytes9. Han et al. reported a slightly different conversion protocol involving overexpression of Brn4, Sox2, c-Myc and Klf4 to achieve iNSC derivation from murine fibroblasts10. A more recent publication describes the generation of neuronal restricted progenitors from human fetal fibroblasts by overexpression of Sox2, c-Myc and either Brn2 or Brn411. Oct4, Brn2 and Brn4 belong to the POU transcription factor family that participate in several developmental and cellular processes, such as regulation of the pluripotency state as well as neural determination corroborating the central role of POU factors for cellular reprogramming in general and transdifferentiation in particular12.

In the mouse, directly converted iNSCs represent the radial glia type of neural stem cells (RG-NSCs)13. Up to date, RG-NSCs are the only mouse NSC population that can be stably maintained in culture. RG-NSCs self-renew indefinitely and can, although they exhibit a glial bias, differentiate into astrocytes, neurons and oligodendrocytes. They have a restricted neuronal differentiation potential as they give rise to mostly GABAergic neurons14. Additionally, depending on the way how they were generated RG-NSC are patterned and thus exhibit diverse, partly mixed regionalization profiles15. Thus, their neurogenic differentiation potential is limited. The potential of RG-NSCs to dedifferentiate into a more immature NSC population is poorly investigated. In this study we set out to explore strategies to reprogram mouse RG- NSCs into a more plastic neuroepithelial state by infection with the POU domain transcription factor Brn2 together with c-Myc. We report dedifferentiation of RG-NSCstowards an early neuroepithelial progenitor cell state as judged by a rapid down-regulation of the radial glia markers Olig2 and Vimentin and upregulation of neuroepithelial markers Dach1 and Sox1. The differentiation of such converted cells resulted in smooth muscle actin- and Peripherin-positive cells in addition to the neuronal marker TUJ1 and the glial marker GFAP.

Results

Dedifferentiation of RG-NSC by Brn-2 and c-Myc

In order to achieve partial dedifferentiation of RG-NSCs we employed a combined transcription factor and pharmacological intervention strategy. Overexpression of Oct4 in RG-NSCs alone is sufficient to induce iPSC colonies and additional c-Myc infection promotes this process16. We hypothesized that the closely related transcription factor Brn2 might have a similar, but less effective reprogramming ability as shown for the direct conversion of fibroblasts into iNSCs10. Moreover, since Brn2 is expressed in the neural plate17, the physiological correlate of an earlier NSC state, we hypothesized that it might direct the conversion of RG-NSCs into an earlier neuroepithelial fate. We omitted Sox2 and Klf-4 from the reprogramming cocktail since these factors are expressed in RG-NSCs. Moreover, we used a small molecule cocktail recently described by Li et al.18 to support epigenetic rearrangements and to enhance the reprogramming efficiency. As a target cell population we used RG-NSC that carry an eGFP reporter cassette under control of the Oct4 promotor. Therefore eGFP is expected to be expressed as soon as the pluripotent status is reached. In a control experiment we infected RG-NSCs with Oct4 only and 11 days after infection and culture in the presence of CHIR99021, Tranylcypromine, SB431542 and valproic acid (CSTV) we observed green fluorescent colonies (Fig. 1A-C) indicative of the formation of iPSCs as described earlier16. RG-NSCs, infected with both, Brn2 and c-Myc, formed a characteristic different type of colony, being GFP-negative (Fig. 1D-E). In contrast, uninfected cells grown in reprogramming medium either without or with compounds did not give rise to epithelial colonies (Fig. 1F-G). RG-NSCs infected with Brn2 and c-Myc and cultured in the absence of small molecules changed their morphology but did not give rise to neuroepithelial colonies (Fig. 1H). In order to characterize converted cells at molecular level we investigated the expression of the radial glia markers Olig2 and Vimentin, as well as the early neuroepithelial markers

Dach1 and Sox1 together with E- and N-Cadherin by semiquantitative RT-PCR at day 1 and day 6 after infection. RG-NSCs were infected either with Brn2 and c-Myc or with Oct4 as a control (Fig. 2A). As expected the RG marker Olig2 was highly expressed in the starting RG-NSC population both at mRNA (Fig. 2C) as well as protein level (Fig. 3A). Oct4 infection of RG-NSCs resulted in a slight decrease of Olig2, whereas the combination of Brn2 and c-Myc strongly down-regulated Olig2. mRNA encoding for the early neuroepithelial marker Dach1 showed strong upregulation in Brn2/c-Myc-infected cells after 6 days whereas it was not detected in RG-NSCs and Oct4-infected cells (Fig. 2A). A corresponding picture evolved from the analysis of the desmosomal Cadherin proteins. E-Cadherin is more prominent in epithelial tissues like the neuroepithelium and ESC colonies, whereas N-Cadherin is expressed in neurons and NSCs¹⁹. Accordingly, RG-NSCs did not show detectable E-Cadherin mRNA and Oct4 as well as Brn2/c-Myc was able to induce the expression of E-Cadherin additionally to N-Cadherin, which is still expressed by the remaining radial glia cells present in the primary cultures (Fig. 2A). In contrast, N-Cadherin specific mRNA was not detected in ESCs but slight bands appeared in samples of RG-NSCs as well as murine embryonic fibroblast control cells. As judged by semiquantitative RT-PCR all other conditions did not yield marked differences in N-Cadherin levels (Fig. 2A). We decided to assess the cellular conversion more comprehensively by qRT-PCR and with a higher temporal resolution. Daily RNA samples were taken during reprogramming between day 1 and day 6 after infection (Fig. 2B-D). This analysis revealed that RG-NSC markers Olig2 and Vimentin were strongly down-regulated during the reprogramming process (Fig. 2c). In case of Olig2, this change in mRNA turned out to be very quickly, as at the second day after infection the mRNA expression level decreased to an almost undetectable level. By contrast, the neuroepithelial markers Dach1 and Sox1 were both up-regulated during reprogramming beginning at day 3 and 4 (Fig. 2B), respectively, with a fold change of 16 and 58, at day 6, compared to day 1. The expression of both markers was confirmed at protein level by immunofluorescence (Fig. 4A-F). Furthermore, qRT-PCR also recapitulated the neuroepithelial transition from N-Cadherin to E-Cadherin (Fig. 2D). E-Cadherin expression exhibits a marked increase at day 4 of reprogramming, while N-Cadherin expression is decreased more than 3–fold at day 2. Immunofluorescence analyses confirmed N-Cadherin expression in Nestin/Olig2

double-positive RG-NSCs (Fig. 3D-F) and E-Cadherin expression in the Dach1 positive induced neuroepithelial colonies (Fig. 4G-I)

Brn-2/c-Myc-converted cells differentiate in neural crest derivatives

The data presented thus far demonstrates that Brn2/c-Myc infection of RG-NSCs leads to a down-regulation of radial glia markers, concomitant with an up-regulation of neuroepithelial markers and initialization of a neuroepithelial transition. Since this transition of the molecular profile suggests a partial dedifferentiation of RG-NSCs into early neuroepithelial progenitors we next set out to confirm this conversion by cellular differentiation studies. In order to test the differentiation potential of the putative early neuroepithelial population we initiated differentiation by replacing reprogramming media with neurodifferentiation medium containing BDNF, GDNF and BMP4. After 6 days of culture cells were fixed and stained for the following markers: neuronal marker TUJ1; Peripherin, a marker for peripheral neurons; GFAP (glial fibrillic acidic protein), an astrocyte marker; and SMA (smooth muscle actin), a marker for neural crest derived muscle cells²⁰. Already 4 days after differentiation we observed cells positive for both neuronal markers investigated, TUJ1 and Peripherin (Fig. 5A). A substantial fraction of the cells was TUJ1-positive; rare cells showed positive Peripherin staining together with a characteristic elongated morphology. About half of the cells were positive for GFAP, the rest was SMA-positive (Fig. 3B). Notably, there was no SMA- and GFAP-double positive staining observed demonstrating the specificity of the staining. As a control untreated RG-NSCs were grown under the same differentiation conditions, fixed and analysed side-by-side (Fig. 3C-D). In general, RG-NSCs gave rise to much less neurons as compared to the reprogrammed cells as judged by TUJ1 staining (Fig. 3c). Notably, no single cell was found to be positive for Peripherin (Fig. 3c). The analysis of glia-type differentiation revealed that the majority of RG-NSCs gave rise to GFAP positive astrocytes while no SMA positive cells were detected (Fig. 3D).

Discussion

In this study we addressed the question if mouse RG-NSCs, which have a restricted differentiation capacity, might have the potential to dedifferentiate into a

developmentally earlier, more plastic neural cell type. Therefore, we replaced the pluripotency-associated POU domain transcription factor Oct4 with the neural POU domain transcription factor Brn2. Correlating with previous studies¹⁰ RG-NSCs can be fully dedifferentiated into iPSCs by Oct4 and c-Myc. In contrast, we show that the co-infection of Brn2 and c-Myc did not result in pluripotent colonies. Already after 6 days, eGFP-negative colonies formed that did not reach the pluripotent state even after 11 days (data not shown), the same period of time in which Oct4 was able to induce pluripotency in RG-NSCs. Thus, Brn2 like Oct4, belonging to the POU transcription factor family, is insufficient to induce pluripotency in RG-NSCs. Brn2 and c-Myc induced a mesenchymal-to-epithelial-like transition in RG-NSCs indicated by a switch from N- to E-Cadherin expression. The expression of early neuroepithelial markers like Dach1 and Sox1 is indicative for a dedifferentiation into a neuroepithelial progenitor state. This finding is supported by the fact that reprogrammed cell cultures give rise to central and peripheral neurons, respectively, as well as SMA-positive progeny under differentiation conditions. This staining pattern is characteristic for cells originating from the neural crest²⁰. These marker proteins are not expressed by differentiated radial glia cells, representing exclusively the central nervous system. This was confirmed by side-by-side comparative analysis with non-reprogrammed RG-NSCs. Taken together, the marker profile before and after differentiation indicates a neural plate border-like identity of the dedifferentiated RG-NSCs. Our study provides a platform to gain further insights into early neurodevelopmental processes and to explore novel reprogramming pathways for enhancing the developmental plasticity of neural stem cells in culture.

Material and Methods

Derivation of RG-NSC and cell culture RG-NSC were isolated from Oct4-GiP mouse embryos E1421 as described previously9. For the cultivation of RG-NSCs plates were coated with poly-ornithine and laminin (PO/Ln) for at least two hours at 37°C. To facilitate attachment to the plates, 1ng/mL laminin was added to the culture medium after seeding of the NSCs. Cells were cultured in NSC cell culture medium14 (98.8% EuroMed, 1% N-2 Supplement, 0.1% EGF, 0.1% FGF), which was changed every other day. On days in between the media change, 10ng/mL of each growth factor EGF and FGF was addeddigestion, 1µg of RNA was used. Furthermore, 1µL of 10x reaction buffer and 1 µL of DNase were added. The reaction solution was then filled up to 10µL with RNase-free water. Digestion took place at 65°C for 30 minutes. Afterwards, 1µL of EDTA was added and the solution was heated to 37°C for 10 minutes. Subsequently, a reverse transcription was performed according to the manufacturer’s protocol following protocol. PCR was performed using GoTaq polymerase.

Table 1: Primers used for amplifying cDNA (5’ to 3’ direction)

Primers for ORFs				Sequence
	Dach1_fwd			CCT GGG AAA CCC GTG TAC TC
Dach1_rev				AGA TCC ACC ATT TTG CAC TCA TT
	GAPDH_fwd			AGG TCG GTG TGA ACG GAT TTG
GAPDH_rev				TGT AGA CCA TGT AGT TGA GGT CA
	E-Cadherin_fwd			CAG GTC TCC TCA TGG CTT TGC
E-Cadherin_rev				CTT CCG AAA AGA AGG CTG TCC
	N-Cadherin_fwd			AGC GCA GTC TTA CCG AAG G
N-Cadherin_rev				TCG CTG CTT TCA TAC TGA ACT TT
	Olig2_fwd			TCC CCA GAA CCC GAT GAT CTT
Olig2_rev				CGT GGA CGA GGA CAC AGT C
	Sox1_fwd			AAG GGA CAC CCG GAT TAC AAG T
Sox1_rev				GTT AGC CCA GCC GTT GAC AT
	Vimentin_fwd			CTG CTT CAA GAC TCG GTG GAC
Vimentin_rev				ATC TCC TCC TCG TAC AGG TCG

Immunohistochemistry

Cells were fixed using 4% PFA for 10 minutes. Afterwards, PFA was removed, cells were washed with PBS two times and a blocking solution was applied for 45 minutes. In the case of staining neurons, PBS was replaced by HBSS buffer, which contains a balanced salt concentration and therefore, does not damage the membrane of the sensitive neurons. Subsequently, primary antibodies ( α-GFAP rabbit IgG (Dako, Glostrup, Denmark), α -Peripherin rabbit IgG (Abcam, Cambridge, UK), α -SMA mouse IgG (Dako, Glostrup, Denmark), α-TUJ1 mouse IgG (Covance, Princeton, NJ), α-Dach1 rabbit IgG (Protein Tech Group, Manchester, UK), α-Nestin mouseIgG    (Merck-Millipore,    Darmstadt,    Germany),     α -E-Cadherin    rat    IgG     (Abcam,

Cambridge, UK),  α-N-Cadherin rabbit IgG (Calbiochem, La Jolla, CA), α-Sox1 goat

IgG (RD-Systems, Minneapolis, MN), ZO-1 rabbit IgG (Invitrogen, Carlsbad, CA)) were dissolved in blocking solution and added to the cells for two hours at room temperature or overnight at 4°C. Primary antibody solution was removed and the cells were washed three times with PBS or HBSS containing 0,1% Triton-X-100 for 5 minutes. Cells were then incubated with secondary antibodies. Fluorescence and phase contrast microscopy pictures were taken using a Leica microscope (DM IL LED) and a confocal laser scanning microscope (Nikon, Tokyo, Japan).