bioRxiv preprint doi: https://doi.org/10.1101/351957; this version posted June 20, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 Single cell expression analysis uncouples transdifferentiation and 2 reprogramming 3 Mirko Francesconi1*, Bruno Di Stefano3,4,5*, Clara Berenguer3, Marisa de Andres3, Maria 4 Mendez Lago6,7, Amy Guillaumet-Adkins6,8, Gustavo Rodriguez-Esteban6, Marta Gut2,6, 5 Ivo G. Gut2,6, Holger Heyn2,6, Ben Lehner1,2,9,10 and Thomas Graf2,3,10 6 7 1Systems Biology Program, Centre for Genomic Regulation (CRG), The Barcelona 8 Institute of Science and Technology (BIST), Dr. Aiguader 88, Barcelona 08003, Spain; 9 2Universitat Pompeu Fabra (UPF), Barcelona 08003, Spain. 3Gene Regulation, Stem 10 Cells and Cancer Program, CRG, BIST, UPF, Barcelona, 08003 Spain. 4. Massachusetts 11 General Hospital Department of Molecular Biology, Harvard Medical School, 185 12 Cambridge Street, Boston, MA 02114, USA. 5. Department of Stem Cell and 13 Regenerative Biology, Harvard University, Cambridge, MA 02138, USA. 6Centro 14 Nacional de Análisis Genómico, Centre for Genomic Regulation (CNAG-CRG); BIST, 15 08028 Barcelona, Spain. 9Institució Catalana de Recerca i Estudis Avançats (ICREA), 16 Barcelona 08010, Spain. 17 18 * These authors contributed equally to the work 19 7Present address: Institute of Molecular Biology (IMB), Mainz, Germany 20 8Present address: Department of Pediatrics, Dana Farber Cancer Institute, Harvard 21 Medical School, Boston MA 02115 22 10Corresponding authors: [email protected], [email protected] 23 24 25 Abstract 26 Many somatic cell types are plastic, having the capacity to convert into other 27 specialized cells (transdifferentiation)(1) or into induced pluripotent stem cells (iPSCs, 28 reprogramming)(2) in response to transcription factor over-expression. To explore 29 what makes a cell plastic and whether these different cell conversion processes are 30 coupled, we exposed bone marrow derived pre-B cells to two different transcription 31 factor overexpression protocols that efficiently convert them either into macrophages 32 or iPSCs and monitored the two processes over time using single cell gene expression 33 analysis. We found that even in these highly efficient cell fate conversion systems, 34 cells differ in both their speed and path of transdifferentiation and reprogramming. 35 This heterogeneity originates in two starting pre-B cell subpopulations, large pre-BII 36 and the small pre-BII cells they normally differentiate into. The large cells 37 transdifferentiate slowly but exhibit a high efficiency of iPSC reprogramming. In 38 contrast, the small cells transdifferentiate rapidly but are highly resistant to 39 reprogramming. Moreover, the large B cells induce a stronger transient 40 granulocyte/macrophage progenitor (GMP)-like state, while the small B cells undergo a 41 more direct conversion to the macrophage fate. The large cells are cycling and exhibit 42 high Myc activity whereas the small cells are Myc low and mostly quiescent. The 43 observed heterogeneity of the two cell conversion processes can therefore be traced 44 to two closely related cell types in the starting population that exhibit different types 45 of plasticity. These data show that a somatic cell’s propensity for either 46 transdifferentiation and reprogramming can be uncoupled. bioRxiv preprint doi: https://doi.org/10.1101/351957; this version posted June 20, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 47 48 One sentence summary: Single cell transcriptomics of cell conversions 49 Main Text: 50 C/EBPα is a master regulator of myelopoiesis(3). When overexpressed in B cell 51 precursors, it induces their efficient transdifferentiation into macrophages(1), and 52 when transiently overexpressed, it poises them for rapid and highly efficient 53 reprogramming into iPSCs in response to induction of Oct4, Sox2, Klf4 and Myc 54 (OSKM)(4). The combination of these two systems gives us the unique opportunity to 55 study the determinants of both types of cell conversion by following gene expression 56 in single cells starting from the same cell population. 57 58 We isolated CD19+ B cells precursors from the bone marrow of reprogrammable 59 mice(5) and infected them with a retrovirus encoding a hormone inducible form of 60 C/EBPa (Cebpa-ER-hCD4). After expansion in culture, we induced them to either trans- 61 differentiate into macrophages or to reprogram into iPSCs. To induce the macrophage 62 fate, we treated the cells with beta-estradiol (E2) to activate C/EBPα. To induce the 63 iPSC fate, we treated the cells with E2 for 18 hours, washed out the hormone and 64 added doxycycline to induce OSKM(4, 6). For transdifferentiation, we collected cells 65 before (0h) and after 6h, 18h, 42h, 66h and 114h of C/EBPα induction; for 66 reprogramming samples were prepared at days 2, 4, 6 and 8 after OSKM induction of 67 18h C/EBPa-pulsed cells (Fig. 1A). We collected two pools of 192 cells at each time 68 point and sequenced their RNA using MARS-Seq(7). After quality control and filtering 69 (see Methods), we obtained expression profiles for 17,183 genes in 3,152 cells. We 70 then performed dimensionality reduction, corrected for batch effects, and extracted 71 gene expression signatures (See Methods and Fig. S1-3, Table S1-4). Visualizing the 72 data using diffusion maps(8) revealed branching between transdifferentiation and 73 reprogramming at the 18h time-point, with largely synchronous cohorts of cells 74 moving along distinct trajectories and reaching homogenous final cell populations 75 consisting of either induced macrophage (iMac) or iPSC-like cells, respectively (Fig. 1B). 76 We observed no branching into alternative routes, in contrast to what has been 77 described for the transdifferentiation of fibroblasts into neurons(9), muscle cells(10) or 78 iPSCs(11, 12). 79 80 B cell genes become largely silenced after 18h (Fig. 1C and Fig. S4A). During 81 transdifferentiation, there is a transient activation of granulocyte/GMP genes (Fig. 1D, 82 Fig. S4B), followed by activation of monocyte (Fig. 1E, Fig. S4C) and then macrophage 83 genes (Fig. 1F, Fig. S4D). After OSKM induction, endogenous Pou5f1 (Oct4) is activated 84 at day 2, followed by expression of Nanog at day 4 and Sox2 at day6 (Fig. 1G, Fig. E-H), 85 consistent with the high reprogramming efficiency of our system(6, 13). 86 87 Visualizing single cells in the expression space spanned by B cell, monocyte and 88 macrophage programs (Fig. 1H) and B cell, mid and late reprogramming (Fig. 1I), 89 however, reveals a degree of asynchrony. To identify potential causes of this 90 asynchronous behaviour, we determined which independent component analysis 91 (ICA)-derived expression signatures (Fig. S2) best predicted cell progression toward the 92 macrophage state (Fig. 2A) at each time-point (excluding expression signatures directly 93 involved in transdifferentiation, that is the B cell, monocyte, granulocyte, and bioRxiv preprint doi: https://doi.org/10.1101/351957; this version posted June 20, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 94 macrophage programs). We found that a signature highly enriched in Myc target genes 95 (component 5, Fig. S2, Table S4) best predicts and negatively correlates with the extent 96 of transdifferentiation at intermediate time points (Fig. 2B, Fig. S5). Expression of the 97 Myc targets varies extensively across cells within each time point but changes little 98 during transdifferentiation (Fig. 2C). The data therefore suggest that the cells with 99 lower expression of Myc targets transdifferentiate more rapidly into macrophages. 100 101 We next tested how the expression of Myc targets relates to the loss of the B cell state 102 and the acquisition of transient myeloid-like cell states during transdifferentiation. 103 Visualizing similarity to the pre-B cell state shows that low expression of Myc targets is 104 more strongly associated with a rapid gain of the macrophage state than with a rapid 105 loss of the B cell state (Fig. 2D). Moreover, higher Myc target expression is associated 106 with a larger and more persistent induction of a GMP-like state (Fig. 2E). Myc target 107 expression does not associate with the extent of induction of a transient monocyte- 108 (Fig. 2F) or granulocyte-like state (Fig. S6). In conclusion, cells with low expression of 109 Myc targets acquire the macrophage fate more rapidly and transdifferentiate via a less 110 pronounced transient induction of a GMP-like state. 111 112 We similarly searched for expression signatures that predict the progression of 113 individual cells toward pluripotency within each time-point during reprogramming to 114 iPSCs (Figure 2g). The expression of Myc targets was again predictive of cell fate 115 conversion especially at early stages, however, in contrast to what was observed 116 during transdifferentiation, high expression of Myc targets is associated with a more 117 advanced state of reprogramming (Fig. 2H, Fig. S7). Moreover – and also different to 118 what was observed during transdifferentiation – the expression of Myc targets 119 increases during reprogramming (Fig. 2I). Visualizing similarity to pre-B cells, GMPs and 120 monocytes during reprogramming shows that cells with high expression of Myc targets 121 and a transient GMP state are at the forefront of the reprogramming trajectory.