Analysis of Mll1-deficiency identifies neurogenic transcriptional modules and Brn4 as a factor for direct astrocyte-to-neuron reprogramming

Abstract

BACKGROUND

Mixed lineage leukemia-1 (Mll1) epigenetically regulates gene expression patterns that specify cellular identity in both embryonic development and adult stem cell populations. In the adult mouse brain, multipotent neural stem cells (NSCs) in the subventricular zone (SVZ) generate new neurons throughout life and Mll1 is required for this postnatal neurogenesis but not for glial cell differentiation. Analysis of Mll1-dependent transcription may identify neurogenic genes useful for the direct reprogramming of astrocytes into neurons.

OBJECTIVE

To identify Mll1-dependent transcriptional modules and determine whether genes in the neurogenic modules can be used to directly reprogram astrocytes into neurons.

METHODS

We performed gene coexpression module analysis on microarray data from differentiating wild-type and Mll1-deleted SVZ NSCs. Key developmental regulators belonging to the neurogenic modules were overexpressed in Mll1-deleted cells and cultured cortical astrocytes, and cell phenotypes were analyzed by immunocytochemistry and electrophysiology.

RESULTS

Transcriptional modules that correspond to neurogenesis were identified in wild-type NSCs. Modules related to astrocytes and oligodendrocytes were enriched in Mll1-deleted NSCs, consistent with their gliogenic potential. Overexpression of genes selected from the neurogenic modules enhanced the production of neurons from Mll1-deleted cells, and the overexpression of Brn4 (Pou3f4) in non-neurogenic cortical astroglia induced their transdifferentiation into electrophysiologically active neurons.

CONCLUSIONS

Our results demonstrate that Mll1 is required for the expression of neurogenic – but not gliogenic – transcriptional modules in a multipotent NSC population and further indicate that specific Mll1-dependent genes may be useful for direct reprogramming strategies.

INTRODUCTION

In the adult mammalian brain, neural stem cells (NSCs) in the lateral ventricle subventricular zone (SVZ) continue to generate new neurons and glia throughout life.^1–6 In mice, SVZ NSCs are a specialized type of astrocyte that can produce both neuronal and glial cell lineages.⁷ Mixed lineage leukemia-1 (Mll1) encodes an evolutionarily conserved chromatin-modifying factor that is required to maintain lineage-specific gene expression in both embryonic and postnatal stem cell populations.^8,9 In SVZ NSCs, Mll1 is required for neurogenesis but not for glial cell differentiation.^7,10 Thus, Mll1-dependent gene expression may in part account for the neurogenic potential of this specialized population of astrocytes in the adult mouse SVZ.

Interestingly, astrocytes derived from non-neurogenic brain regions such as the cerebral cortex can be directly converted into functional neurons via the overexpression of specific transcription factors.^1,3,5,6,11 For instance, enforced expression of Dlx2, Mash1 (also known as Ascl1), or Pax6 in cultured postnatal mouse cortical astroglia results in neuronal transdifferentiation.^1,3,5,6 Furthermore, emerging evidence indicates that direct astrocyte-to-neuron reprogramming can also be achieved in vivo,^12,13 as is seen in striatal astrocytes where lentiviral transduction of Sox2 along with the expression of specific extrinsic signals results in the production of mature neurons.¹² Given the astrocytic identity of SVZ NSCs, we hypothesized that characterizing their neurogenic gene expression profile may reveal new factors useful for reprogramming astrocytes into neurons.

Genome-wide analysis of Mll1-dependent transcription in SVZ NSCs provides an excellent opportunity to distinguish neurogenic gene expression from that required for glial cell differentiation. While wild-type SVZ NSCs produce all three major neural cell types – neurons, astrocytes, and oligodendrocytes – Mll1-deficient SVZ cells efficiently generate only the glial lineages.¹⁰ However, our understanding of Mll1-dependent gene expression has been limited. Proneural Mash1 and gliogenic Olig2 are expressed in Mll1-deficient SVZ cells during differentiation but Dlx2 is not expressed, correlating with the impaired neurogenesis.¹⁰

In this study, we used microarrays and gene coexpression analysis to identify “modules” of transcribed genes that distinguish neurogenesis from gliogenesis. In contrast to differential gene expression analysis, which compares the mean expression level of individual genes between two or more groups, gene coexpression analysis seeks to identify groups of genes (i.e., transcriptional modules) that are expressed in similar patterns across biological samples.^14–17 Such approaches enable the discovery of cell type-specific gene expression despite cellular heterogeneity within each sample.^14,16

Our coexpression analysis generated a resource that distinguishes gene expression modules related to SVZ neurogenesis from those related to gliogenesis and cell proliferation. To evaluate the neurogenic potential of genes identified by this method, we used lentiviruses to overexpress select module members in Mll1-deficient SVZ cells. Using this approach, we identified genes such as Brn4 (also known as Pou3f4) that belonged to neurogenic modules and partially rescued neuronal differentiation. Finally, to explore whether Mll1-dependent genes in SVZ NSCs can be used to transdifferentiate non-neurogenic astrocytes into neurons, we overexpressed Brn4 in astrocytes cultured from the cerebral cortex. We observed that the overexpression of Brn4 in cortical astrocytes resulted in the production of electrophysiologically active neurons. Taken together, these studies provide insight into the key role that Mll1 plays in postnatal neurogenesis and further indicate that neurogenic factors expressed in SVZ NSCs may be useful for astrocyte-to-neuron reprogramming strategies.

MATERIALS AND METHODS

Tissue preparation and culture of primary SVZ NSCs and cortical astrocytes

All use of animals was performed in accordance with guidelines established by the Institutional Animal Care and Use Committee at the University of California, San Francisco. Mll1-deleted SVZ NSCs were obtained from hGFAP-Cre;Mll1^f/f mice, which exhibit Cre-mediated recombination of the Mll1^f/f alleles in cells expressing the hGFAP-Cre transgene. hGFAP-Cre;Mll1^+/+ litter mates were used as wild-type controls. The brains from mice aged postnatal days 5–7 were removed from the skull and placed in ice-cold Leibovitz’s L15 media (Gibco, Cat. No. 11415-064). Coronal slabs containing the SVZ were obtained and the meninges removed. The lateral SVZ and cortical gray matter were then microdissected with care taken to prevent cross-contamination. Dissected tissue was then dissociated with 0.25% trypsin as previously described.¹⁰ The suspension of cortical astrocytes was also treated with DNase I (Roche Diagnostics, Cat. No. ROCZR139). SVZ single cell suspensions were plated in “N5” medium¹⁸ containing DMEM/F12 with Glutamax (Gibco, Cat. No. 10565-018), 5% fetal bovine serum (FBS; Hyclone, Cat. No. SH30070.03), N2 supplement (Gibco, Cat. No. 17502-048), 35μg/ml bovine pituitary extract (BPE; Gibco, Cat. No. 13028.014), 20ng/ml recombinant human epidermal growth factor (EGF; Peprotech, Cat. No. AF-100-15), 20ng/ml recombinant human fibroblast growth factor (FGF; Peprotech, Cat. No. 100-18B-1mg), and antibiotic-antimycotic (Gibco, Cat. No. 15240-062). SVZ cultures were maintained with serial passages as previously described.¹⁸ At the first or second passage, SVZ cells were seeded onto 16-well chamber slides and immunostained for mouse MLL1 (1:500; Millipore, Cat. No. 05-764) to confirm ≥90% deletion before proceeding with RNA isolation for subsequent experiments. Cortical astrocyte single cell suspensions were resuspended in medium containing DMEM/F12, 10% FBS, 5% horse serum (UCSF Cell Culture Facility, Cat. No. CCFAW001), B27 supplement (Gibco, Cat. No. 17504-044), 3.5mM glucose, 10ng/ml EGF, 10ng/ml FGF, and antibiotic-antimycotic. Astrocytes were then plated on culture dishes pre-coated with 0.1mg/ml poly-D-lysine (PDK; Sigma-Aldrich, Cat. No. P6407) and 2μg/ml laminin (Life Technologies, Cat. No. 23017-015) and allowed to reach confluence.

SVZ cultures were differentiated after 5–7 passages. For differentiation, the medium was changed to a serum-free differentiation medium consisting of DMEM/F12 with Glutamax, N2 supplement, BPE, and antibiotic-antimycotic.¹⁸ Astrocyte differentiation was performed after the first passage by changing to a medium consisting of DMEM/F12, 3.5mM glucose, B27 supplement, antibiotic-antimycotic, and 20ng/ml of brain derived neurotrophic factor (BDNF; Peprotech, Cat. NO. 450-02).

RNA isolation and microarray analysis

Microarray analysis of wild-type and Mll1-deleted SVZ NSC cultures was performed pre-differentiation (0 days) and at 1, 2, and 4 days after differentiation (biologic triplicates at each time point). SVZ cultures were homogenized using QIAshredder columns (QIAgen, Cat. No. 79654) and RNA was extracted using QIAgen RNeasy Plus Mini Kit (Cat. No. 74134). RNA quality and integrity were assessed with the Agilent BioAnalyzer 2100 before samples were converted to cDNA. Affymetrix Mouse Gene 1.0 ST microarrays were used to measure gene expression levels and scanned by the UCSF Genomics Core. Raw data (.CEL files) were pre-processed using the Affymetrix Power Tools (APT) software package (http://www.affymetrix.com/partners_programs/programs/developer/tools/powertools.affx). Expression data were background-corrected, summarized, and quantile-normalized using the robust multi-array average (RMA) algorithm¹⁹ in APT. Sample heterogeneity was assessed using the SampleNetwork R function;²⁰ the mean inter-sample adjacency was 0.968 and no outliers (Z.K < −3) were present.

Gene coexpression analysis

Gene coexpression modules were identified using a four-step approach. First, pairwise Pearson correlation coefficients (cor) were calculated across all samples for all transcripts (n = 25,232) that were perfectly and uniquely targeted by their microarray probe sequence (Affymetrix ‘CrossHyb Type’ = 1). Second, transcripts were clustered using the flashClust¹⁷ implementation of a hierarchical clustering procedure with complete linkage and 1 – cor as a distance measure. The resulting dendrogram was cut at a static height of ~0.258, corresponding to the top 1% of pairwise correlations for the entire dataset. Third, all clusters consisting of at least 15 members were identified and summarized by their module eigengene (i.e., the first principal component obtained via singular value decomposition).^14,15 Fourth, highly similar modules were merged if the Pearson correlation coefficients of their module eigengenes exceeded an arbitrary threshold (0.85). This procedure was performed iteratively such that the pair of modules with the highest correlation > 0.85 was merged, followed by recalculation of all module eigengenes, then followed by recalculation of all correlations, until no pairs of modules exceeded the threshold. Following these steps, 14 coexpression modules were identified. The strength of module membership (kME) for each probe on the microarray was calculated by correlating its expression pattern across all samples with each module eigengene.^15,16

Module enrichment analysis

Module enrichment analysis with curated gene sets was performed using a one-sided Fisher’s exact test for over-representation with gene symbols as a common identifier. Curated gene sets included experimentally validated markers of astrocytes (Table S7 from Bachoo et al.²¹), oligodendrocytes (from Nielsen et al.,²² reanalyzed as previously described;¹⁶ from Cahoy et al.,²³ restricted to genes expressed >10X higher in oligodendrocytes versus other cell types in the adult mouse forebrain; and from Doyle et al.,²⁴ consisting of markers of mouse cerebellar Olig2+ oligodendrocytes), and neurons (from Cahoy et al.,²³ restricted to genes expressed >10X higher in neurons versus other cell types in the adult mouse forebrain). Modules were defined as consisting of all transcripts that were positively correlated with the module eigengene at a significance threshold of P<1.26e⁻⁰⁷. This threshold corresponds to a Bonferroni-corrected P-value of .05/(the total number of probes X the total number of modules). Gene Ontology (GO, http://www.geneontology.org) analysis was performed using DAVID (Database for Annotation, Visualization, and Integrated Discovery, http://david.abcc.ncifcrf.gov/). Probe IDs of module genes were uploaded and converted to gene lists using as a background all transcripts (n = 25,232) that were perfectly and uniquely targeted by their microarray probe sequence (Affymetrix ‘CrossHyb Type’ = 1). To determine if modules were Mll1-dependent or independent, analysis of variance (ANOVA) was performed in R (http://cran.us.r-project.org/) on multivariate linear regression models that related each module eigengene to treatment condition (wild-type versus Mll1-deleted), the number of days in culture, and the interaction between these two terms. Mll1-dependence was defined as P<0.05 for the treatment effect.

Enforced expression of Mll1-dependent factors

cDNA of genes of interest (Dlx2, Brn4, Tlx, Zdhhc23, and Shisa3) was cloned into the lentiviral pUltra vector (Addgene), which contains a green fluorescent protein (GFP) reporter. Lentivirus was then generated from HEK293T cells with 36μg of plasmid containing a given cDNA of interest, 9μg VSVg coat plasmid, 7.5μg pRSV, and 7.5μg pMDL helper plasmids using the FuGENE 6 transfection system (Promega, Cat. No. E2693). One day after transfection, cell culture media was changed to viral production media containing Lonza Ultraculture media (12-1725F), 110ng/mL sodium pyruvate, and 5mM sodium butyrate. Cells were grown in viral production media for 48 hours before pelleting of virus at 22,000 rpm by ultracentrifuge for 2 hours. Viral pellet was resuspended in PBS and stored at −80° Celsius.

Immunocytochemistry

At the time points of interest, cells were fixed in 4% paraformaldehyde for 30 minutes and then washed three times in phosphate buffered saline (PBS). Blocking was performed for 2 hours at room temperature in blocking buffer containing PBS with 1% bovine serum albumin, 5% normal goat serum, 0.3% Triton-X-100 (Sigma), 0.3 M glycine, and 0.03% sodium azide. All antibody incubations were also performed in blocking buffer at room temperature. Primary antibodies included mouse anti-neuronal class III β-tubulin (Tuj1, 1:500; Covance, Cat. No. MMS-435P), mouse anti-microtubule associated protein 2 (MAP2, 1:200; Millipore, Cat. No. MAB364), guinea pig anti-doublecortin (DCX, 1:400; Millipore, Cat. No. AB2253), rabbit anti-Brn4 (1:500; Millipore, Cat. No. AB15456), rabbit anti-glial fibrillary acidic protein (GFAP, 1:500; Dako, Cat. No. Z0334), chicken anti-GFP (1:500; Aves, Cat. No. GFP-1020), mouse anti-O4 (1:100; Millipore, Cat. No. MAB345), rabbit anti-NG2 (1:400; Millipore, Cat. No. AB5320), mouse anti-GAD67 (1:1000; Millipore, Cat. No. MAB5406), and rabbit anti-Calbindin (CalB, 1:200; Swant, Cat. No. CB 38). Appropriate secondary antibody incubations were performed for 30 minutes in blocking buffer at room temperature and included a nuclear counterstain (DAPI). Imaging was performed using a Leica DMA4000 B inverted fluorescent microscope. Cell counts for a given condition were performed on seven non-overlapping regions imaged at 20X. The totals of each region were then summed and averaged over biological triplicates.

Electrophysiology

Whole-cell current-clamp recordings²⁵ were made using a Multiclamp 700B amplifier (Molecular Devices, CA, USA). Signals were filtered at 1 kHz and sampled at 10 kHz using a Digidata 1440A analog-to-digital converter (Molecular Devices). All data were recorded and analyzed with pClamp 10 software (Molecular Devices). The liquid junction potential was measured with 3 M KCl and adjusted. Cells on coverslips were placed in a bath solution (containing the following in mM: 135 NaCl, 5 KCl, 2 CaCl2, 1.2 MgCl2, 10 HEPES, 10 glucose at ~300 mOsm [pH 7.4]) and visualized using an Olympus IX71 microscope with differential interface contrast optics at 40X. Patch electrodes (3–4 MΩ) contained (in mM) 125 KCl, 10 Na₂PO₄, 10 HEPES, 2 Mg3ATP, 0.3 Na2GTP at ~ 290 mOsm (pH 7.4). Currents were manually injected to hold the membrane potential around −70 mV to record voltage responses. Tetrodotoxin (TTX, 1 μM; Tocris, Cat. No. 1078) was applied through a gravity-fed perfusion system.

RESULTS

Mll1 is required for the expression of transcriptional modules related to SVZ neurogenesis

In the adult and postnatal mouse brain, SVZ NSCs produce neuroblasts that migrate to the olfactory bulb where they differentiate into interneurons and integrate into local circuits.^26,27 SVZ NSCs can also generate astrocytes in response to local injury²⁸ as well as oligodendrocytes in both normal and pathologic conditions.^29–31 When cultured as monolayers, SVZ NSC cultures express Nestin and Sox2^10,18 and can recapitulate the development of these distinct neural lineages,^10,18,32 with wild-type SVZ NSCs efficiently generating neurons, astrocytes, and oligodendrocytes (Figure 1A). In contrast, Mll1-deleted SVZ NSCs (which are derived from hGFAP-Cre;Mll1^f/f mice and are ~95% deleted for Mll1¹⁰) survive, proliferate, and produce glia, but are defective for neurogenesis (Figure 1B).¹⁰ To identify Mll1-dependent gene expression, we performed microarray analysis of wild-type and Mll1-deleted SVZ cultures in self-renewal conditions and after 1, 2, and 4 days of differentiation (each performed in biological triplicate). By analyzing gene coexpression relationships,^14–16 we identified 14 transcriptional “modules” (labeled by color), each representing groups of genes with expression levels that were highly correlated across the 24 samples (Supplementary Figure 1 and selected modules in Figure 2A). Each module was related to specific biological processes and neurobiological cell types via enrichment analysis for GO terms (Table 1) and curated biological gene sets.

Figure 1. Mll1 is required for neurogenesis but not for gliogenesis.

A) Differentiation of wild-type SVZ NSCs over a 7-day time course. B) Differentiation of Mll1-deleted SVZ NSCs. Tuj1+ neurons are shown in panels i–iv while O4+ and NG2+ oligodendrocytes are shown in panels v and vi. Panels i–iv: GFAP (green), Tuj1 (red), DAPI (blue). Panels v and vi: O4 (green), and NG2 (red). Scale bar, 50μm. Panels Ai–iv were originally published in Gonzalez-Roybal and Lim.⁵⁰

Figure 2. Gene coexpression analysis identifies distinct transcriptional modules.

A) Hierarchical clustering of gene coexpression modules. This dendrogram was produced using average linkage hierarchical clustering with 1 – the Pearson correlation of the module eigengenes¹⁵ (first principal components) as a distance measure. B–C) Select gene coexpression modules showing representative heatmaps and first principal components at 0 (self-renewal conditions), 1, 2, and 4 days of differentiation in wild-type (WT) and Mll1-deleted (Mll1^Δ/Δ) NSCs. B) Non-neurogenic modules identified through gene coexpression analysis. C) Mll1-dependent modules that are enriched for neurogenic gene expression. Lowercase letters in B and C indicate independent biological replicates.

Table 1.

Gene ontology terms associated with select transcriptional modules

Module	Top five gene ontology (GO) terms	p-value
Turquoise	GO:0007049~cell cycle	6.44E-42
	GO:0022402~cell cycle process	9.04E-39
	GO:0022403~cell cycle phase	1.95E-37
	GO:0000279~M phase	1.51E-34
	GO:0000278~mitotic cell cycle	2.93E-33
Tan	GO:0019725~cellular homeostasis	4.51E-03
	GO:0019228~regulation of action potential in neuron	4.76E-03
	GO:0010975~regulation of neuron projection development	5.39E-03
	GO:0019226~transmission of nerve impulse	7.43E-03
	GO:0001508~regulation of action potential	7.93E-03
Greenyellow	GO:0007272~ensheathment of neurons	1.09E-06
	GO:0008366~axon ensheathment	1.09E-06
	GO:0019228~regulation of action potential in neuron	1.96E-06
	GO:0001508~regulation of action potential	4.11E-06
	GO:0048709~oligodendrocyte differentiation	7.26E-06
Black	GO:0031175~neuron projection development	1.97E-06
	GO:0048666~neuron development	4.01E-06
	GO:0030182~neuron differentiation	1.36E-05
	GO:0030030~cell projection organization	5.68E-05
	GO:0007409~axonogenesis	1.13E-04
Magenta	GO:0030900~forebrain development	1.35E-03
	GO:0021537~telencephalon development	1.93E-03
	GO:0021879~forebrain neuron differentiation	4.60E-03
	GO:0006928~cell motion	5.25E-03
	GO:0021872~generation of neurons in the forebrain	6.01E-03

Seven of the 14 coexpression modules exhibited similar temporal patterns in wild-type and Mll1-deleted cells, indicating that some SVZ gene expression programs do not require Mll1 (Supplementary Table 1). For example, the turquoise module (Figure 2B) consisted of genes that were progressively down-regulated under differentiation conditions in both wild-type and Mll1-deleted cells. This module was highly enriched for GO terms related to cell cycle regulation (Table 1) and these findings are consistent with the observation that Mll1-deleted cultures proliferate at rates similar to wild-type SVZ NSCs.¹⁰ The tan module (Figure 2B) consisted of genes that were quickly up-regulated under differentiation conditions in both wild-type and Mll1-deleted SVZ cells. This module contained Gfap and was enriched for experimentally validated markers of astrocytes²¹ (P=2.76×10⁻⁸); indeed, both Mll1-deleted and wild-type NSCs efficiently generate GFAP+ astrocytes.¹⁰

The greenyellow module consisted of genes with higher expression levels in Mll1-deleted cells versus wild-type cells at the 4 day time point (Figure 2B), and its relationship to oligodendrocytes was revealed by GO analysis (Table 1) and significant enrichment with experimentally validated markers of oligodendrocytes from Cahoy et al.²³ (P=1.11×10⁻²⁹), Nielsen et al.²² (P=3.64 × 10⁻¹³), and Doyle et al.²⁴ (P=3.02 × 10⁻¹⁴). Interestingly, a trend towards increased oligodendrogliogenesis is observed with Mll1-deletion both in vivo and in vitro.¹⁰

Other coexpression modules exhibited temporal patterns of transcriptional upregulation that were strongly Mll1-dependent. For example, the black and magenta modules (Figure 2C) consisted of genes that were strongly upregulated in wild-type, but not Mll1-deleted, SVZ cells after 2 days (black) or 1 day (magenta) of differentiation (Supplementary Table 2). Both modules were enriched for GO terms related to neurogenesis (Table 1) and the black module was strongly significant for experimentally validated markers of mature neurons from Cahoy et al.²³ (P=2.77×10⁻¹⁵). Black module gene members include Dlx5, Dlx6, Sp8, and Sp9 – all of which encode transcription factors expressed in SVZ neuroblasts and olfactory bulb interneurons.^33,34 Furthermore, the presence of Glutamic acid decarboxylase-1 (Gad1) and GABA(A) receptor-associated protein like 1 (Gabarapl1) in the black module suggest a GABAergic identity of neurons within the population of differentiating cells.

The magenta module contained genes that are expressed at earlier stages of neurogenesis (Supplementary Table 2). For instance, the expression of Dlx1 generally precedes that of Dlx5 and Dlx6.³⁵ Consistent with this temporal relationship, Dlx1 was a member of the magenta module, whereas Dlx5 and Dlx6 were members of the later-upregulated black module. Dlx2, which is also temporally expressed with Dlx1, was most strongly associated with the magenta module. Other magenta module genes included Arx, a homeobox transcription factor that regulates the proliferative expansion of immature neuronal progenitor cells in the developing cortex (P=3.45E-08),³⁶ and Tlx (also known as Nr2e1), which is required for SVZ NSC self-renewal.³⁷ Finally, Brn4, which has also been associated with SVZ neurogenesis,³⁸ was also identified in the magenta module.

Interestingly, at the time that the magenta module was first upregulated (1 day of differentiation), both wild-type and Mll1-deleted cultures lacked Tuj1-positive neuroblasts and contained many GFAP-positive astroglia (Figure 1). This early-differentiation expression of the magenta module suggests its relationship to early stages of neuronal lineage commitment in wild-type SVZ NSCs, which at that time still have astroglial characteristics.

Enforced expression of select magenta module genes partially rescues neurogenesis in Mll1-deleted SVZ cells

To evaluate the neurogenic potential of genes associated with the magenta module, we tested their ability to rescue neuronal differentiation of Mll1-deleted SVZ cells. For each of these genes – Dlx2, Tlx, Zdhhc23, and Brn4 – we generated an overexpression lentivirus that coexpresses the green fluorescent protein (GFP) marker (Table 2). To provide a negative control, we produced a lentiviral vector that expresses only the marker proteins GFP and alkaline phosphatase (LV-hPLAP-GFP). Mll1-deleted SVZ cultures were infected with each lentivirus, and after 4 days of differentiation, neuronal differentiation was assessed by immunocytochemistry (ICC, Figure 3A–F) and quantified (Figure 3G). In control virus-infected cultures (Figure 3A), approximately 10% of the GFP-positive cells were immunopositive for the Tuj1 neuronal marker (Figure 3G), representing the lower levels of neurogenesis of Mll1-deleted SVZ cells. As compared to this control, cultures infected with the Dlx2 lentivirus (Figure 3B) resulted in a 5-fold increase in GFP-positive, Tuj1-positive cells (Figure 3G), which is similar to the rescue previously observed with transient Dlx2 overexpression.¹⁰ Individual overexpression of Brn4, Tlx, and Zdhhc23 also increased the production of neurons from Mll1-deleted SVZ cells (Figure 3C–E), albeit with approximately 2-fold lower efficiency than that of Dlx2 (Figure 3G). We also generated a lentivirus that expresses Shisa3, a gene that was not found in the neurogenic modules, and infection with this vector did not significantly increase neuronal differentiation (Figure 3F and 3G). Thus, when overexpressed, select members of the magenta module can induce neurogenesis in Mll1-deleted SVZ cells.

Table 2.

Lentiviral expression vectors

Magenta module genes (neurogenic)

LV-Dlx2-GFP

LV-Brn4-GFP

LV-Tlx-GFP

LV-Zdhhc23-GFP

Mll1-independent gene (non-neurogenic)

LV-Shisa3-GFP

Control vector

LV-hPLAP-GFP

Figure 3. Rescue of neurogenesis by overexpression of select transcription factors.

A–F) Immunocytochemistry of hGFAP-Cre;Mll1^f/f SVZ NSC cultures after 4 days of differentiation and enforced expression with lentiviral constructs containing select Mll1-dependent factors: control (A), Dlx2 (B), Brn4 (C), Tlx (D), Zdhhc23 (E) and Shisa3 (F). GFP (green), Tuj1 (red), DAPI (blue). G) Quantification of infected (GFP+) cells that are also Tuj1+ (n=3 wells). Error bars = standard deviation. One-way ANOVA with post-hoc Dunnett’s test was used to compare each experimental factor to hPLAP control. Scale bar, 50μm.

Enforced expression of Brn4 in cortically-derived primary astroglial cultures induces neurogenesis

Astroglia cultured from the postnatal cortex can be transdifferentiated into neurons via the overexpression of certain single transcription factors including Dlx2 and Pax6^1,3 – both of which are members of the magenta module. To test if another magenta module member could likewise induce neurogenesis from cortical astroglia, we enforced Brn4 expression using LV-Brn4-GFP and analyzed the cell fate of GFP-positive cells after 20 days of differentiation. While Brn4 has been associated with SVZ neurogenesis,³⁸ its ability to transdifferentiate astrocytes into neurons was unknown. As a control, we also followed the fate of astroglial cells infected with LV-hPLAP-GFP. Before Brn4 overexpression, cortical astroglial cultures consisted of GFAP-positive cells (Figure 4A), and no Tuj1-positive neuronal cells were observed (data not shown). After 10 days of differentiation, LV-Brn4-GFP infected cultures contained GFP-positive, Tuj1-positive cells (data not shown). After 20 days of differentiation, these cultures contained GFP-positive, Tuj1-positive cells with neuronal morphology (Figure 4B). As expected, we did not observe any GFP-positive neuronal cells in the control LV-hPLAP-GFP infected cultures (quantified in Figure 4G). LV-Brn4-GFP infection also generated GFP-positive cells that co-expressed MAP2 (Figure 4C) or DCX (Figure 4D), further supporting the neuronal identity of these cells. In contrast, LV-hPLAP-GFP control infection did not produce any GFP-positive cells expressing the DCX and MAP2 neuronal antigens (quantified in Figure 4G). Furthermore, some neuronal cells induced by LV-Brn4-GFP expressed GAD67 (Figure 4E), an enzyme present in GABAergic neurons, and calbindin (Figure 4F), a marker of specific neuron subtypes. This ICC analysis indicates that LV-Brn4-GFP infection induces the neuronal transdifferentiation of cortical astrocytes in vitro.

Figure 4. Brn4 induces neurogenesis in cortically-derived astroglia.

Immunocytochemistry of cortically-derived astroglial cultures before (A) and 20 days after (B–F) neuronal differentiation and enforced lentiviral expression of Brn4. B–F) Examples of cells with neuronal morphology expressing various neuronal markers: Tuj1 (B), MAP2 (C), DCX (D), GAD67 (E), and CalB (F). GFAP (red, A), DAPI (blue, A–F); GFP (green, B–F). No cells with neuronal morphology expressing these neuronal markers were identified in control-infected cultures. G) Quantification of infected (GFP+) cells that were also Tuj1+, MAP2+, or DCX+ (from B, C, and D, respectively; compared to hPLAP controls; n=3 wells for each). Error bars = standard deviation. Student’s t-test used to compare Brn4 to hPLAP for each stain. Scale bars: A) 50μm; B–F) 25μm.

To further characterize the identity of these LV-Brn4-GFP induced neurons, we performed whole-cell patch clamp recordings (Figure 5). Resting membrane potential was recorded in current clamp mode without injecting current and found to be negative in each cell tested (−41 ± 2 mV, n=3), indicating the expression of ion channels and the regulation of ionic gradients across the cellular membrane. We then depolarized cells by injecting a series of depolarizing current steps and all cells tested fired action potentials that appeared after a threshold membrane potential in an all-or-none fashion (Cm = 28 ± 0.1 pA/pF, n=3, Figure 5B left). Additionally, these action potentials were reversibly eliminated by the addition of TTX, a specific blocker of voltage-gated Na⁺ channels (Figure 5B middle and right). Taken together, these data indicate that neurons transdifferentiated from cortical astroglia via LV-Brn4-GFP infection are electrophysiologically functional.

Figure 5. Electrophysiologic analysis of neurons induced by Brn4 overexpression in cortically-derived astroglia.

A) Phase contrast image of a sample recorded neuron. B) Action potentials were elicited by a series of current injections ranging from −40 to +80 pA in 20 pA increments. Action potentials were reversibly blocked with 1 μM TTX.

DISCUSSION

The differentiation of neurons from multipotent NSCs requires the concerted expression of distinct sets of genes that promote neuronal identity. Such lineage-specific gene expression is in part coordinated by epigenetic regulators such as MLL1, a histone methyltransferase and member of the Trithorax group (TrxG) of chromatin regulators.^8,9 In human patients, MLL1 mutations have been linked to Wiedemann-Steiner syndrome, a rare developmental disorder characterized by intellectual disability including autism.³⁹ We demonstrate that Mll1 is required for the expression of neurogenic transcriptional modules in SVZ NSCs, indicating its central role in the gene expression required for normal neural development. Furthermore, genes selected from the neurogenic modules can promote neuronal differentiation from gliogenic cell populations. Our study thus provides new insights into the role that Mll1 plays in neural development and further suggests that specific neurogenic module genes may be useful for direct astrocyte-to-neuron reprogramming strategies.

Mll1-deletion targeted to SVZ NSCs results in defective neurogenesis, and in a previous study, we found that this lineage-specific deficiency relates to the lack of proper Dlx2 expression.¹⁰ In the current study, we used microarray and gene coexpression analyses to further characterize the Mll1-dependent transcription that corresponds to neurogenesis. We identified two neurogenic gene coexpression modules – black and magenta – and our analysis suggests that the magenta module is involved in earlier stages of neuronal lineage commitment. The magenta module was upregulated in wild-type cultures after just 1 day of differentiation and before the appearance of Tuj1-positive neuroblasts, suggesting that gene expression in this module does not simply reflect transcription related to the presence of neurons. Indeed, Dlx2, which is most closely associated with the magenta module, is first expressed in SVZ transit amplifying cells, before their differentiation into Tuj1-positive migratory neuroblasts.³³ In addition, the magenta module defined 127 other genes with a similar pattern of expression (Supplementary Table 2), providing many novel candidates for the study of genes involved in early stages of neurogenesis.

Previously, we showed that transient overexpression of Dlx2 partially rescues Mll1-dependent neurogenesis. For this study, we used lentiviral expression vectors to enforce the expression of Dlx2, Tlx, Zdhhc23, and Brn4 in Mll1-deleted SVZ NSCs, and the resultant increase in neuronal differentiation induced by each vector demonstrated the neurogenic potential of these magenta module genes. Furthermore, the ability to induce neuronal differentiation of Mll1-deleted cells suggests that this TrxG chromatin regulator is not required for the transcription downstream of key neurogenic genes, at least when they are overexpressed.

Tlx encodes a nuclear receptor that is expressed in SVZ NSCs, and Tlx-deficiency reduces NSC self-renewal and SVZ neurogenesis.^37,40 Interestingly, transgenic overexpression of Tlx in vivo results in the expansion of the NSC population and glioma-like lesions;⁴⁰ it is therefore possible that the increased numbers of neurons observed with Tlx overexpression in Mll1-deleted cultures relates to increased proliferation of neurogenic progenitor cells. Zdhhc23 encodes a zinc-finger-containing palmitoyl-transferase⁴¹ that is associated with the synaptic membrane.⁴² The function of this molecule is poorly understood, and, to our knowledge, it has not been previously associated with neurogenesis or Mll1.

Brn4 is a member of the POU domain transcription factor family. Brn4 knock-out mice exhibit deafness, vestibular abnormalities, and gait impairments.⁴³ Brn4 is expressed in the SVZ of the adult rodent⁴⁴ and results of a recent study suggest that Brn4 is a component of a transcriptional network downstream of Pax6 that drives SVZ neurogenesis.³⁸ The Brn4 locus is a direct target for the PAX6 transcription factor and Pax6 overexpression upregulates Brn4.³⁸ Given this transcriptional interrelationship and the Mll1-dependence of Pax6 expression, it is possible that the activation of the Brn4 locus during neurogenesis is not directly dependent upon MLL1 and instead relates to activation by PAX6. Future studies of MLL1 protein enrichment and related chromatin state changes – as has been performed to demonstrate the role of MLL1 at the Dlx2 locus¹⁰ – will help clarify whether Brn4 and other neurogenic genes are directly regulated by this chromatin-modifying factor.

Our study identified Brn4 as a transcription factor that can induce neuronal transdifferentiation from cultured astrocytes. This finding contributes to the relatively small collection of factors that have been shown to promote the conversion of astrocytes into neurons.^1,3,5,6 Pax6 can induce neurons from astrocytes, and, interestingly, this transcription factor was also a member of the magenta module. Similarly, Dlx2, which also induces neurons from astrocytes, was also most strongly associated with the magenta module. We did not detect Ngn2 in an SVZ neurogenic module; however, the overexpression of Ngn2 induces the production of glutamatergic neurons,³ a cell type that is not commonly produced from SVZ NSCs. We also did not find Mash1 as a member of a neurogenic module. Consistent with previous results, Mash1 expression was not Mll1-dependent.¹⁰ While Mash1 overexpression induces transdifferentiation into GABAergic neurons, this transcription factor also plays a role in normal SVZ gliogenesis.⁴⁵ Recent studies of direct fibroblast-to-neuron reprogramming indicate that Mash1 acts as a “pioneer” factor that promotes the recruitment of other transcription factors.⁴⁶ Indeed, astrocytes with both Mash1 and Dlx2 overexpression generate greater numbers of neurons than either transcription factor alone.³ Future combinatorial strategies (e.g., Mash1, Brn4, and Dlx2 together) may further improve the efficiency of neuronal reprogramming observed both in vitro and in vivo. Importantly, these findings may serve as a basis for future studies of neuronal reprogramming in human cells. Several studies have already demonstrated that human-derived cells can be successfully reprogrammed.^47–49 Corti et al. have even converted human astrocytes into neurons using either OCT4, SOX2, or NANOG.⁶ Interestingly, they also show that this process is epigenetically regulated through DNA methylation. Given the important role of Mll1 in murine SVZ neurogenesis, it will be worthwhile to explore both its role in human neurogenesis and cell reprogramming. Neurosurgeons are in the unique position to do this through access to intraoperatively-derived human tissue.

CONCLUSION

This study provides new insights into the Mll1-dependent expression that underlies postnatal SVZ neurogenesis. Gene coexpression analysis identifies transcriptional modules that distinguish neurogenesis from gliogenesis, and enforced expression of select neurogenic factors can promote neurogenesis from Mll1-deleted SVZ NSCs. Brn4 is identified as a factor for direct astrocyte-to-neuron transdifferentiation. These data thus serve as a resource for future studies of Mll1 in neural development and direct astrocyte-to-neuron reprogramming strategies.

Supplementary Material

Supplementary Figure 1. Supplementary Figure 1. Transcriptional modules identified through gene coexpression analysis.

Gene coexpression analysis was performed on microarray data from wild-type and Mll1-deleted SVZ NSC cultures at 0 (self-renewal conditions), 1, 2, and 4 days after differentiation. Fourteen transcriptional modules were identified and labeled by colors (A–N). Each page shows a snapshot for a given module consisting of a heat map, a barplot of the module eigengene (ME, or first principal component), and the top 15 genes based on the strength of module membership (kME). Where microarray probes did not correspond to a known gene “NA” is listed as the gene name. The presence of a given gene more than once in the top genes by kME indicates that there were multiple microarray probes corresponding to that gene. All samples were analyzed in biological triplicates.