Sequential EMT-MET induces neuronal conversion through Sox2

Direct neuronal conversion can be achieved with combinations of small-molecule compounds and growth factors. Here, by studying the first or induction phase of the neuronal conversion induced by defined 5C medium, we show that the Sox2-mediated switch from early epithelial–mesenchymal transition (EMT) to late mesenchymal–epithelial transition (MET) within a high proliferation context is essential and sufficient for the conversion from mouse embryonic fibroblasts (MEFs) to TuJ+ cells. At the early stage, insulin and basic fibroblast growth factor (bFGF)-induced cell proliferation, early EMT, the up-regulation of Stat3 and Sox2, and the subsequent activation of neuron projection. Up-regulated Sox2 then induced MET and directed cells towards a neuronal fate at the late stage. Inhibiting either stage of this sequential EMT-MET impaired the conversion. In addition, Sox2 could replace sequential EMT-MET to induce a similar conversion within a high proliferation context, and its functions were confirmed with other neuronal conversion protocols and MEFs reprogramming. Therefore, the critical roles of the sequential EMT-MET were implicated in direct cell fate conversion in addition to reprogramming, embryonic development and cancer progression.

. Gene expression in MEFs was used as control to calculate the log 2 values for the other 8 samples. Although these samples were obtained from different GEO datasets, the log 2 values of genes identified in all eight samples were further normalized to have similar averages and standard deviations. Cells with similar log-ratio expression profiles were grouped together by applying the Hierarchical Clustering method (a). EMT scores (b) and metabolism scores (c) were also calculated.

Induction of shrunken and MEF-like cells require mitosis
(a-c) Shrunken cells could be generated after a particular round of mitosis and continue to undergo mitosis, or via direct conversion (a). The contributions of the two routes to the final number of shrunken cells are summarized in (b). The average conversion times of cells that used these two routes are listed in (c).
(d-f) MEF-like cells could be generated with high mitosis or with low or even no mitosis (d). The contributions of the two routes to the final number of MEF-like cells on Day 14 are summarized in (e). The numbers of cells on Day 0 that used these two routes are listed in (f).
(g-i) The 500 initial MEFs were traced during the 5C-induced conversion. The numbers of neuron-like, shrunken and MEF-like cells that converted with high levels of mitosis are summarized across the conversion (g). The numbers of apoptotic cells during the generation of these three cell types are listed in (h) and normalized to the final numbers of these three types of cells (i).
Two-tailed Student's t-test was used for (c).

Supplementary Figure 4
High initial cell density activates the Notch pathway (e-f) DAPT differentially effected the proliferation of neuron-like cells and other cells, especially at the late stage (n = 6).
(g) The influence of initial MEF density on the final TuJ + percentages is summarized (n = 6).
One-way ANOVA with Dunnett's post-hoc test was used for (c-e and g), and two-way ANOVA with Bonferroni's post-hoc test was used for (f).   (e) Sox2 was overexpressed in MEFs, and the expression of genes related to EMT was determined with qPCR (n = 5).

Sup
(f-g) Sox2 was overexpressed in MEFs. TGFβ and E616452 were also used. Cell migration (f) and cell cycle (g) were assayed on Day 3 with transwell assays and FACS (n = 6).  Supplementary Table 1 Medium and Materials used in the current research Materials like growth factors and small-molecule compounds used in the current research were listed with their catalog numbers, purchased companies, and conventional final concentrations. Different combinations of these materials or medium composition were also listed in this Table. Supplementary Table 2 Genes with significant expression changes were enriched in seven GO clusters Gene expression profiles during the 5C-induced conversion was determined with RNA-seq. Genes with significant expression changes (over 2 fold) in at least one of the four time points were subjected for GO analysis and seven clusters of GO terms were identified. The genes in these seven clusters and their expression changes were listed. The average of log 2 values of expression change of genes in different clusters were also calculated. Table 3 EMT scores calculated basing on RNA-seq results EMT scores were calculated with two different methods as described in Methods.

Supplementary
Briefly, gene expression changes of 130 genes which have been identified as EMT generic signature from RNA-seq data were used. The log2 values of expression changes of genes which normally are up-regulated during EMT were used as positive score, while those of gene that are normally down-regulated during EMT were used as negative score.

Supplementary Table 4 EMT scores calculated based on qPCR results
EMT scores were calculated with two different methods as described in Methods.
Briefly, Gene expression changes of Cdh2, Fn1, Slug, Zeb1, Ep-cam and Ocln form qPCR were used to calculate another set of score. The log2 values of expression changes of genes which normally are up-regulated during EMT were used as positive score, while those of gene that are normally down-regulated during EMT were used as negative score.

Supplementary Table 5 Expression analysis from three GEO datasets for ESCs, NSCs, PNs, MEFs and conversion intermediates
Expression profiles of cells during 5C-induced conversion (Day 5, 10 and 14), MEFs, primary neurons, NSCs, ESCs, and iPSCs (passage 3 and 20) were downloaded from GEO database. Gene expression in MEFs were used as control to calculate the log 2 values for the other 8 samples. Although these samples were obtained from different GEO datasets, the log 2 values of genes identified in all eight samples were further normalized in order to have similar average and standard deviation. Table 6 qPCR results in cell cultured with 5C and other medium MEFs were treated with 5C, 5C', IFB, FB, or IB medium for 5, 10 or 14 days. TGFβ and E6161452 were used in time period as indicated. Expression of Cdh2,Fn1,Slug,Zeb1,Ocln,Sox2,Stat3,Map1b,Reln,Robo1 and Tubbs were determined with qPCR. Three types of scores, EMT, dual-S and neuron scores were calculated from log 2 values of expression change compared to MEFs. Table 7 Enriched binging sites of transcriptional factors via Pscan  Table 8 Expression changes of Sox2 and other genes in different EMT datasets.

Supplementary
The 74 human and 31 mouse EMT microarray results were integrated as described in Methods. The results on Sox2 and Stat3 expression were extracted and subjected for further analysis. The totally 105 microarrays were classified into three groups with strong, medium, and weak EMT. The expression of Sox2 and Stat3 were summarized with in microarrays of these three groups. Table 9 Sox2 binding sites in NSC are close to neuron projection genes Sox2 binging peaks were downloaded from GSE33024. The genes which are closest to the peaks were subjected for GO analysis and neuron projection (GO0048812/0031175) was enriched. The expression changes of these neuron projection genes during the differentiation from NSCs to Neurons were also listed (from GSE24131, Supplementary Table 10).

Supplementary Table 10
EMT scores during NSCs differentiation and reprogramming with OK+M+S Expression profiles during the differentiation from NSCs to Neurons (GSE24131) and during the reprogramming with sequential introduction of Yamanaka factors, OK+M+S (GSE39260) were downloaded from GEO database. EMT scores were calculated as described in Methods.

Supplementary Table 11
Antibodies and qPCR primers used in current research As title.

Supplementary Table 12
Datasets used in current study As title.

Supplementary Table 13
Statistics Information of each figure and table As title.