Fibroblasts do not contain neural progenitor cells or neurons
In order to prove that the cells we used (HFF-1 cell line) were pure fibroblasts and did not contaminate with neural stem cells or neurons, the cells were stained with the fibroblast-specific markers Fibronectin and Vimentin, the neural stem cell-specific marker PAX6, and the neuron-specific markers Tuj1 and MAP2. The fibroblasts were detected to be fibronectin- and Vimentin-positive（Figure 1A）and PAX6/MAP2/Tuj1-negative (Figure 1B). The positive control neural stem cell line NE-4C and primary mouse cortical neurons express PAX6 and Tuj1/MAP2 respectively (Figure 1C and Figure 1D). These indicated that the cells we used were pure and homogeneous fibroblasts.
Expression of miR-124 and let-7 inhibited the proliferation of fibroblasts, promoted the expression of pro-neural genes.
To explore the feasibility of using miR-124 and let-7 expression in neuronal conversion, we prepared adenovirus barbering the miR-124 and let-7 precursors along with an enhanced green fluorescent protein (EGFP) marker (Figure 2A), and infected human neonatal foreskin fibroblasts. The infection efficiency was almost 100% (Figure S1A) with the expression level of these two miRNAs sharply increased 24 hours post infection (Figure S1B, S1C). We examined their early effects on fibroblast proliferation and cell type-specific gene expression. By EDU assay we found that the proliferation of HFF-1 cells was significantly inhibited by introduction of miR-124 and let-7 for 48 hours, with most of the cells exist cell cycle (Figure 2B and Figure 2C).
Next, on days 0, 1, 3, and 5 after adenoviruses infection, we detected the expression of several pro-neural transcription factors, including myt1L, Brn2, Olig2, NeuroD1 and Ascl1, the fibroblast-specific transcription factors, including Zeb1, Thy1, Prrx1, Osr1, and Lhx9. During this period, the expression of pro-neural transcription factors increased 4 to 170 folds (Figure 2D), while the expression of fibroblast-specific transcription factors reduced by 2 to 5 folds (Figure 2E). These results suggest that miR-124 and let-7 expression inhibit the proliferation of fibroblasts, promote the expression of pro-neural transcription factors, and repress the expression of fibroblast specific genes at the early stage of conversion.
Fibroblasts expressing miR-124 and let-7 displayed neuron-like morphologies and expressed neuronal markers
To investigate whether miR-124 and let-7 can convert fibroblasts into neuron-like cells, human fibroblasts infected with adenoviruses were cultured in neuron induction medium for 15~20 days (Figure 3A), then immunofluorescence staining was performed with antibodies against neuronal specific markers. The human fibroblasts infected with the two microRNAs showed obvious neuron-like morphology after differentiation for 15 days, while there were no such changes in the control and miR-124 or let-7 single infection group (Figure 3B). Most of the converted cells expressed the neuronal marker Tuj1 (88%, Figure 3C and Figure 3D) and MAP2 (82%, Figure 3E and Figure 3F) at day 15 post induction. The synaptic marker SYN1 could detect about 20 days after differentiation (Figure 3G). These results suggest that miR-124 and let-7 overexpression can convert human fibroblasts into mature neuron-like cells.
The transdifferentiation process did not go through the neural progenitor cell stage
To demonstrate that our neural conversion process is direct transdifferentiation rather than first passing through a neural progenitor stage, we performed immunofluorescence staining to examine the neural progenitor marker PAX6 on days 2, 6, 10, and 15 during neuron induction. We did not observe any Pax6 positive neural progenitor cells during the process (Figure 4A and Figure 4B), suggesting that our transdifferentiation process is direct, bypassing the neural progenitor stage, which will greatly reduce the potential tumorigenic risk if using our induced cells for cell transplantation in patients.
Induced neurons were GABAergic.
To further characterize the cell type of our induced neurons, we used immunofluorescence experiment to detect the specific markers of the most important and abundant neurons in the brain, that is, glutamatergic neuron marker VGLUT1, GABAergic neuron marker GABA, and dopaminergic neuron marker TH. Strikingly, all the converted cells express GABA (Figure 5A) and contained neither VGLUT1 nor TH positive cells (Figure 5B and Figure 5C). In order to further confirm the induced cells are inhibitory interneurons, we stained the induced cells with GAD67, another marker of GABAergic neurons, and found almost all cells were GAD67 positive (Figure 5D). The control fibroblast cells induced in parallel did not express these neuronal markers (Figure 5E~5H). These data indicate our fibroblast-derived neuron-like cells are GABAergic.
Neuronal induction leads to changes in the overall transcriptional profile
In order to clarify the detailed alteration in the gene expression profile of our induced neurons, we compared the global gene expression pattern of HFF-1 cells and induced neurons using comparative transcriptome analyses, showing that 4500 genes were upregulated and 3260 genes were downregulated more than two-fold after induction (Table S1). The global genome heatmap with hierarchical cluster analysis revealed that the global gene expression profile of the induced neurons from three induction batches at days 15 showed a higher degree of similarity with one another rather than with their fibroblasts of origin (Figure 6A) and as expected, the expression of the previously confirmed target genes which inhibit neural differentiation, including PTBP1, ROCK1 and PPP1R13L of miR-124(17, 23, 24), NRAS, STAT3 and MYC of let-7(25-29), were significantly down-regulated in the induced neurons (Figure 6B). Ontology (GO) function enrichment analysis on genes whose expression changed more than two-fold after conversion showed the genes involved in the biology process related to central nervous system development, synapse activities and neuron morphogenesis were significantly up-regulated in induced cells (Figure 6C), while the expression of genes related to the formation of extracellular matrix components, cytoskeleton regulation and epidermal cell differentiation was significantly down-regulated (Figure 6D). These findings were consistent with the previous study(12). Of note, the expression of many representative genes of the neurons, such as NCAM1, MAP2, TUBB3 (Tuj1), MAPT (Tau), SYT1, NEFL, GAP43 and SNAP25 were dramatically increased. Conversely, genes coding for the fibroblast markers THY1, COL12A1 and TGFB1I1 were significantly repressed (Figure 6E). These results indicate that the reprogramming process removed the lineage gene expression characteristics of the original cells, while specifically induced neuronal phenotype.
In the cerebral cortex, nearly all GABAergic interneurons arise from two progenitor zones in the ventral telencephalon, the Medial ganglionic eminence (MGE) and Caudal ganglionic eminence (CGE). MGE generates about 70% of interneuron progenitors, which give rise to all PV and SST interneurons. Caudal ganglionic eminence generates about 30% interneuron progenitors, which give rise to VIP and RELN interneurons(30). Notably, the transcriptome results revealed a robust increase in the detection of SST gene, the marker of the SST interneurons: its expression level increased more than 100-fold with the transcripts per million reads (TPM) approaching five hundred after conversion; The expression of PV gene, the marker of PV interneurons and VIP gene, the marker of VIP interneurons were also increased, but the TPM were relatively low for PV and negligible for VIP. No significant changes in RELN gene, the marker of the RELN interneuron were detected (Figure 6F).
In order to further determine the cell subtypes of the induced neurons, we performed immunofluorescence staining on the converted cells with antibodies against PV and SST. Nearly all the cells were SST positive while the PV signal was undetectable (Figure 6G). This further clarifies the converted cells we generated are relatively pure SST interneurons. Since many neuropsychiatric disorders, such as temporal lobe epilepsy, depression, Alzheimer's disease and schizophrenia, are closely associated with the loss or dysfunction of SST interneurons, which suggests that our induction strategy, if further improved, may have great therapeutic potential for clinical application.