Abstract Ascl1 and Ngn2, closely related proneural transcription factors, are able to convert mouse embryonic stem cells into induced neurons. Despite their similarities, these factors elicit only partially overlapping transcriptional programs, and it remains unknown whether cells are converted via distinct mechanisms. Here we show that Ascl1 and Ngn2 induce mutually exclusive side populations by binding and activating distinct lineage drivers. Furthermore, Ascl1 rapidly dismantles the pluripotency network and installs neuronal and trophoblast cell fates, while Ngn2 generates a neural stem cell-like intermediate supported by incomplete shutdown of the pluripotency network. Using CRISPR-Cas9 knockout screening, we find that Ascl1 relies more on factors regulating pluripotency and the cell cycle, such as Tcf7l1. In the absence of Tcf7l1, Ascl1 still represses core pluripotency genes but fails to exit the cell cycle. However, overexpression of Cdkn1c induces cell cycle exit and restores the generation of neurons. These findings highlight that cell type conversion can occur through two distinct mechanistic paths, even when induced by closely related transcription factors. Subject terms: Reprogramming, Transdifferentiation, Functional genomics __________________________________________________________________ Expression of transcription factors can convert one cell type to another beyond developmental paths. Here, the authors show that cells can take two mechanistically distinct paths in the same transition paradigm when driven by the similar proneural factors Ascl1 and Ngn2. Introduction Cell identity in development is typically crafted by an interplay of multiple transcriptional activators and inhibitors successively restricting the lineage. Cells gradually rewire their gene transcriptional network, retarget multiple transcriptional factors to new loci, reshape their epigenetic landscape, and establish epigenetic barriers to lock the next developmental state^[40]1–[41]5. In contrast, directed lineage conversions rely on the expression of transcription factors (TF) in alien cellular contexts that can be affected by different distributions of epigenetic marks^[42]6,[43]7, posttranslational modifications^[44]8,[45]9, interaction partners^[46]10–[47]13, and other factors^[48]14,[49]15. This can affect the binding and activity of the TF. Conflicting or incomplete cues of lineage formation allow the formation of alternative lineages during directed lineage conversion regimes^[50]16–[51]19. These conversions do not necessarily follow developmental trajectories, and can also skip intermediate cell states^[52]20,[53]21. As such, they provide an opportunity to study alternative inroads to cell states. Thus, cellular reprogramming can be used as a powerful tool to address fundamental principles underlying lineage specifications and generate models for a study of various diseases as well as potential therapeutic modalities^[54]22,[55]23. Basic helix loop helix (bHLH) TFs have proven to be efficient factors for lineage conversion. Various proneural factors, such as Ascl1, Ngn1, Ngn2, Neurod1 and Neurod4 have been utilized to convert different cell types to functioning neurons^[56]24–[57]28. Among those, Ascl1 and Ngn2 (encoded by the Neurog2 gene) have emerged as preferred tools for in vitro neuronal reprogramming^[58]24. Both are master regulators of neuronal fate in the developing central nervous system. Expression of Ascl1 in ventral telencephalon progenitor cells generates inhibitory GABAergic interneurons, whereas Ngn2 directs dorsally situated progenitors toward excitatory neurons with glutamatergic identity. Although, Ascl1 and Ngn2 give rise to different subtypes in the telencephalon, expression of Ngn2 in the ventral telencephalon can rescue Ascl1 null mice^[59]29. Interestingly, while expression of Ascl1 and Ngn2 in astrocytes can recapitulate developmental neuronal subtype specification to GABAergic or glutamatergic neurons, respectively, expression of Ascl1 in mouse embryonic fibroblasts (MEF), as well as expression of Ascl1 or Ngn2 in mouse embryonic stem cells (ESC), lead to primarily glutamatergic neurons^[60]27,[61]28,[62]30. Thus, the subtype specification depends on the initial cell type. It is thus vital to understand how ectopically expressed TFs interact with the initial cellular context to define the outcome of conversions. The ability to transition toward the same states using different transcription factors gives an opportunity to directly contrast differentiation mechanisms. In an elegant study, Aydin et al. showed that Ascl1 and Ngn2 in ESC prefer binding distinct E-box motifs^[63]6. This, in turn, initiates different accessibility and gene expression patterns, influencing downstream TFs, while still resulting in the formation of glutamatergic induced neurons (iN) in both cases^[64]6,[65]27,[66]28,[67]30. However, downstream mechanistic differences, such as the downregulation of the initial pluripotency network, are still not understood. Furthermore, even though Ascl1 and Ngn2 possess ‘pioneering’ TF properties, their binding and expression can be influenced by cellular context, e.g., Ascl1 binds and induces muscle lineage when expressed in MEF, but myoblast induction was not reported in ESC^[68]6,[69]16,[70]19. In this work, we study the mechanistic differences in how the expression of Ascl1 or Ngn2 transitions cells between two identical states. We observe different side lineages forming in parallel to neurons due to differential binding and different strategies for exiting pluripotency upon Ascl1 and Ngn2 induction: Ascl1 rapidly shuts down the pluripotency network and arrests the cell cycle to install neuronal or trophoblast states, while Ngn2 retains Sox2 expression to produce neuron stem cells (NSC). CRISPR/Cas9 forward genetic screens revealed that genes involved in pluripotency regulation and cell cycle control are affecting neuronal reprogramming by Ascl1, but not Ngn2. Our results highlight different mechanistic pathways to iN employed by these bHLH TFs. Results Ascl1 and Ngn2 convert ESC to iN but generate different side lineages Ectopic expression of Ascl1 or Ngn2 in mouse embryonic stem cells (ESCs) is sufficient to induce terminal differentiation into neurons^[71]28. Yet, the differences in transition mechanism toward neurons as well as possible side populations are not well characterized^[72]6,[73]31. To examine this cell type conversion in detail, we generated clonal ESC cell lines expressing rtTA and TetO-Ascl1 or TetO-Ngn2^[74]27,[75]28 (Supplementary Fig. [76]1a). After doxycycline (Dox) addition, ESCs are rapidly converted to induced neurons: Ascl1 and Ngn2 produce cells expressing the neuronal marker TUBB3 and displaying neuronal morphology from day 3 and day 2 onward, respectively (Supplementary Fig. [77]1a). To report neuronal fate in these cell lines we endogenously tagged the pan-neuronal marker gene Mapt on its C-terminus with the fluorescent protein Venus^[78]27 (Supplementary Fig. [79]1b) and performed time-resolved bulk RNAseq upon Dox-induction. Cells were sorted into Venus-positive neurons and Venus-negative cell populations (Fig. [80]1a) from day 3 onward. As reported before^[81]28,[82]30, both Ascl1 and Ngn2 give rise to similar iN cell identities (Fig. [83]1b; Supplementary Fig. [84]1c–e). Thus, initial ESC and terminal iN states are very similar between Ascl1 and Ngn2-induced conversions (Supplementary Fig. [85]1b bottom). This is in line with previous observations that transcriptomes converge to drive iN formation despite differences in the initial transcriptional response^[86]6 and follow an overall similar trajectory between ESC and iN in the PCA analysis (Fig. [87]1c, d; Supplementary Fig. [88]1g). Fig. 1. Ascl1 and Ngn2 induce different alternative lineages. [89]Fig. 1 [90]Open in a new tab a Schematic overview of the experimental design. b Scatter plot comparing gene expression at Day 6 between Ascl1 and Ngn2 Venus-positive cells with various neuronal subtype specific markers indicated in green. c, d Principal component analysis of time-resolved bulk RNAseq after Ascl1 (c) or Ngn2 (d). Each data point corresponds to the single time point replicate. Color intensity shows day post-induction. Shape corresponds the Mapt-Venus reporter upregulation. Arrows show the trajectory cells take after the Ascl1 (c) or Ngn2 induction (d). e, f Vulcano plot comparing gene expression between Venus-positive and negative populations at day 6 post-induction of Ascl1 (e) or Ngn2 (f). Red circles denote top significantly upregulated or downregulated genes as well as example genes marking in trophoblast (e) or NSC lineages (f). g Representative immunostained cells for a trophoblast marker CDX2 and a neuronal marker Map2 at day 6 post-induction of Ascl1 or Ngn2. Trophoblast markers were expressed only after Ascl1 induction, but not Ngn2. h Representative immunostained cells for an NSC marker PAX6 and a neuronal marker TUBB3 at day 6 post-induction of Ascl1 or Ngn2. NSC markers were expressed only after Ngn2 expression, but not Ascl1. i Scatter plot comparing gene expression changes between Ascl1 and Ngn2 at day 1 post-induction. Highlighted circles are example genes that are neuronal markers expressed in both (green), trophoblast Ascl1 specific markers (blue), NSC Ngn2 specific markers (yellow), pluripotency related genes (red). Source data are provided as a Source Data file. To investigate cells that fail to make iN in more detail, we focused on the Mapt-Venus-negative cells, which could represent incomplete or alternative differentiation outcomes. Mapt-Venus-negative cells generated by Ascl1 cluster closer to the initial ESC populations in PCA plots than Ngn2-induced Mapt-negative cells (Fig. [91]1c). However, this is not due to retaining a population of undifferentiated ESC as only a marginal number of cells express ESC marker NANOG in terminal population (Supplementary Fig. [92]1g), and NANOG and OCT4 are not expressed in Mapt-Venus population (Supplementary Fig. [93]1h). To identify the alternative type of cells generated by Ascl1, we used PanglaoDB^[94]32 using genes differentially expressed between Venus-negative and positive populations at day 6 (Fig. [95]1e, f; Supplementary Fig. [96]2a). Interestingly, Ascl1 produce cells expressing trophoblast markers such as Hand1, Cdx2, Tpbpa, Krt8 (Fig. [97]1e, g; Supplementary Fig. [98]2b–g) with mesenchymal morphology, which were not present in Ngn2-induced cultures (Fig. [99]1g, Supplementary Fig. [100]2b, c, f). We termed these cells induced-Trophoblast-like-cells (iT). Interestingly, many of the Krt8, Cdx2 positive iTs appeared binucleated, which could be a result of multinucleation similar to trophoblast lineage development in vivo (Supplementary Fig. [101]2e)^[102]33. The induction of iT could be due to Ascl1 mimicking the bHLH transcription factor Ascl2, a driver for the trophoblast lineage. Both Ascl1 and Ascl2 are evolutionary close and share near identical DNA binding domains and bind similar E-box motifs (Supplementary Fig. [103]2h–k)^[104]33–[105]37. Lastly, to exclude clonal effects of the cell line used, we repeated these experiments in the background of an alternative mouse ESC line, E14. We introduced rtTA via a piggybac transposon vector and expressed Ascl1 from a Dox-inducible viral vector^[106]27 and generated 24 single-cell derived clones. All the clones showed the formation of both iN and iT, suggesting that the formation of iTs is a reproducible side product of ectopic Ascl1 expression in mouse ESCs (Supplementary Fig. [107]3). In contrast to Ascl1 induction, Ngn2 induces Mapt-negative cells expressing NSC markers, such as Sox2, Pax3, Pax6, Nes (Fig. [108]1f, h; Supplementary Fig. [109]2a, [110]4a, b), as previously described^[111]21,[112]38. Furthermore, Ngn2 reprogramming could be locked in the NSC-like state (iNSC) in the presence of FGF2 and EGF and is dependent on the Notch pathway^[113]39,[114]40 (Supplementary Fig. [115]4c, d). In contrast, we did not observe NSC markers upregulated during Ascl1-induced differentiation (Fig. [116]1h, Supplementary Fig. [117]4b, c). To see if Ngn2 can use iNSC state as proliferative intermediate, we differentiated cells in the presence or absence of cytosine β-D-arabinofuranoside (AraC) from day 4 post-induction to inactivate dividing cells (Supplementary Fig. [118]4e). Indeed, addition of AraC drastically reduces Ngn2-produced iNs, while Ascl1 was insensitive to AraC treatment, suggesting that no continuously proliferative intermediate is present during Ascl1-induced iN reprogramming (Supplementary Fig. [119]4e). In summary, despite Ascl1 and Ngn2 converting mESC to similar iN subtypes, Ascl1 and Ngn2 produce distinct additional alternative cell lineages, suggesting that despite identical initial and terminal populations, differences exist that we sought to understand further (Supplementary Fig. [120]1b bottom). Ascl1 and Ngn2 initiate paths with different transcriptional programs To get a better understanding of different transcriptional response invoked by Ascl1 and Ngn2, we performed bulk RNAseq and ChiPseq on day 1 (Fig. [121]1a). Both Ascl1 and Ngn2 induce general neuronal markers such as Tubb3, Map2 and Onecut2 and downregulate general pluripotency markers like Nanog, Klf4 (Fig. [122]1i). Furthermore, Ascl1 strongly induces downstream targets Tfap2b, Lmx1b, while Ngn2 strongly upregulates Neurod1, Nhlh1 (Fig. [123]1i). In addition, Ascl1 upregulate Trophoblast lineage markers, e.g., Krt7/8, Hand1, while Ngn2 upregulates expression of NSC related genes, like Pax3 and Sox3 (Fig. [124]1i). Interestingly we observed that early in reprogramming cells are positive for both—neuronal and trophoblast markers (Supplementary Fig. [125]5a, b). In addition, we reanalyzed available scRNAseq data^[126]6 for Day 2 of ESC to iN conversion by Ascl1 and Ngn2 and could also observe cells positive for both neuronal and trophoblast markers (Supplementary Fig. [127]5c, d). This suggests that Ascl1 can induce both lineages simultaneously, which later are resolved into iN or iT cells (Supplementary Fig. [128]5b). As reported by Aydin and colleagues^[129]6, Ascl1 and Ngn2 show different preferences for E-box motives, which in turn result in the