TEMPO: A system to sequentially label and genetically manipulate vertebrate cell lineages

During development, regulatory factors appear in a precise order to determine cell fates over time. To investigate complex tissue development, one should not just label cell lineages but further visualize and manipulate cells with temporal control. Current strategies for tracing vertebrate cell lineages lack genetic access to sequentially produced cells. Here we present TEMPO (Temporal Encoding and Manipulation in a Predefined Order), an imaging-readable genetic tool allowing differential labelling and manipulation of consecutive cell generations in vertebrates. TEMPO is based on CRISPR and powered by a cascade of gRNAs that drive orderly activation/inactivation of reporters/effectors. Using TEMPO to visualize zebrafish and mouse neurogenesis, we recapitulated birth-order-dependent neuronal fates. Temporally manipulating cell-cycle regulators in mouse cortex progenitors altered the proportion and distribution of neurons and glia, revealing the effects of temporal gene perturbation on serial cell fates. Thus, TEMPO enables sequential manipulation of molecular factors, crucial to study cell-type specification. One-Sentence Summary Gaining sequential genetic access to vertebrate cell lineages.

activation/inactivation of reporters/effectors. Using TEMPO to visualize zebrafish and mouse 23 neurogenesis, we recapitulated birth-order-dependent neuronal fates. Temporally manipulating 24 cell-cycle regulators in mouse cortex progenitors altered the proportion and distribution of neurons 25 and glia, revealing the effects of temporal gene perturbation on serial cell fates. Thus, TEMPO 26 enables sequential manipulation of molecular factors, crucial to study cell-type specification.

27
One-Sentence Summary: Gaining sequential genetic access to vertebrate cell lineages.

29
Main Text: Cell specification during tissue and organ formation depends on spatial position, 30 temporal patterning and cell-lineage relationships (1-3). In many biological systems the timing of 31 proliferative events and the order of expression of differentiation promoting factors determine the 32 cell's anatomical distribution and identity. A well-studied example is the development of neuronal 33 circuits in Drosophila, formed by multiple neuronal types assembled in a precise spatial and 34 temporal manner. Neural progenitors express a cascade of transcription factors and differentiate 35 following an invariant order, determining temporal patterns of functional circuit assembly (4). In 36 more complex organisms, such as vertebrates, emergence of cell diversity during tissue formation 37 is also regulated by spatial and temporal patterning. For example, gene expression cascades are 38 essential for the progressive differentiation of pancreatic cell types, and for the sequential 39 emergence of neuronal subtypes in the central nervous system (5-6). While the transcriptional 40 timing of many differentiation genes has been established, the temporal boundaries in which these 41 factors can promote cell fate determination have not been defined. This limits our ability to predict, 42 and potentially correct, developmental defects caused by misregulation of cell fate regulatory 43 genes. Thus, establishing the cellular origins and spatio-temporal interactions underlying complex 44 tissue formation is crucial to determine the mechanisms of cell specification in development and 45 disease and to eventually produce cell types at will for cell replacement therapy (7).

47
Genetic labeling and manipulation strategies which preserve the native tissue context are necessary 48 to establish the link between cell origins, identity and spatial distribution. Existing technologies 49 require live imaging to reconstruct cell histories at single-cell resolution (8-9). However, in most To build such a versatile system in vertebrates, we repurposed the recently developed CaSSA 97 switches (20). These genetic switches are disrupted reporter genes which become activated by a 98 CRISPR/Cas9 double-strand break (DSB) between two direct repeats (homologous sequences), 99 followed by single-strand annealing (SSA) repair mechanism that recombines both repeats 100 resulting in a scarless sequence (Fig.1A) (21). In contrast to non-homologous repair mechanisms CaSSA reporter (OFF). After Cas9 editing and SSA repair, the CaSSA reporter is activated 106 resulting in a frame shift that inactivates the downstream reporter. To activate several of these 107 reporters in a predefined order, we deployed a conditional guide RNA (gRNA) switch (19). Similar 108 to CaSSA switches, a gRNA switch is a disrupted gRNA sequence that gets activated by another 109 gRNA and Cas9 editing followed by SSA repair (Fig.1C). With these elements we engineered 110 TEMPO, a compact tri-cistronic transgene containing three fluorescent protein sequences in three 111 different frames sequentially activated by a parallel cascade of gRNAs and Cas9 nuclease. The 112 reporter cascade starts with a preactivated CFP reporter in the first temporal window (T1), 113 followed by the activation of RFP CaSSA reporter by gRNA-1 (T2), which also activates a gRNA-114 2 switch, to subsequently drive YFP CaSSA reporter activation (T3). Importantly, only one 115 reporter is expressed at a time, each step in the cascade is irreversible and only repair through SSA 116 allows reporter activation (Fig.1D). This design maximizes reporter diversity while reducing the 117 number of transgenes needed to allow genetic access to spatio-temporal cell relationships (Fig.1E).

197
To validate the full TEMPO reporter cascade, we injected 1-cell-stage zebrafish embryos with two 198 integrative constructs ubiquitously expressing: (1) the TEMPO reporter with the preactivated 199 gRNA-1 and (2) Cas9 and the inactive gRNA-2 switch (Fig. 2A). The efficiency of co-expression 200 of both plasmids was measured to be >88% (Fig. S3). We then monitored the progression of the   sequential labels, while providing genetic access to specific temporal windows. Further, it allows 296 specific spatial labelling when combined with the broadly used UAS/Gal4 system, making it a 297 versatile strategy to reconstruct cell histories which could be applied to any tissue.

TEMPO links birthdate of neurons and glia with layer distribution in the mouse brain 360 361
Existing technologies require live imaging for spatio-temporal reconstruction of cell histories (8-   Between E15.5 and E17.5, TEMPO+ progenitors emigrated from the VZ while later-born RFP+ 385 (at E16.5) and YFP+ (at E17.5) neurons migrated past early-born CFP+ or RFP+ cortical plate 386 neurons, respectively (Fig. 4B-D). To quantify the rate of color transition we focused on TEMPO 387 progression in cortical progenitors. Sox2 expression was used to delineate the region occupied by 388 radial glial cells (RGC) in the VZ (Fig.S4). Quantification of the percentage of color transition (% 389 CT) resulted in 80% progenitor transitions from CFP+ to RFP+ in the first 24h after the 390 electroporation. One day later, from E14.5 to E15.5, we found 31% progenitor transitions from 391 CFP+ to RFP+ and 25% transitions from RFP+ to YFP+ (Fig. 4F, table inset). The lower transition 392 capacity at later stages is consistent with the known decrease in cell division and reduction in the 393 RGC numbers in the VZ during corticogenesis (33). This also explains the dramatic reduction in 394 dividing TEMPO+ progenitors after E13.5 ( Fig. 4F and S5). Overall, the progression of TEMPO 395 in developing mouse cortices recapitulates the temporal dynamics of RGCs cell division capacity 396 and sequential production of their progenies. Furthermore, these results demonstrate the efficacy 397 of TEMPO to reconstruct spatio-temporal cell histories in a mammalian model.

399
We then sought to explore TEMPO color distribution in postnatal mouse brains, 10 days after birth integrate into the genome and propagate throughout development. We found that more than 28% 418 CFP+ neurons in the upper layers co-expressed the episomal plasmid, and thus did not divide much 419 after the electroporation. In contrast, less than 10% of RFP+ neurons and none of the YFP+ 420 neurons co-expressed the episomal plasmid, consistent with their progenitors undergoing more 421 divisions before differentiation (Fig. S6).

423
After neurogenesis is complete (~E16), neural progenitors transition to a gliogenic mode, 424 generating astrocytes and oligodendrocytes (32). Given the timing of the transition from neuro to 425 gliogenesis and that of TEMPO reporters electroporated at E12.5 (Fig. 4B) (Fig. 4L). We still found some CFP+ astrocytes but this was explained by their lack of Cas9 430 expression, which would not allow cascade progression (Fig. S7). In contrast to TEMPO+ neurons 431 whose distribution was highly correlated with the inside-out pattern of layer formation (Fig. 4B,   432 G, K), TEMPO+ astrocytes were scattered along the radial axis ( Fig. 4G-I, N), consistent with 433 previous studies showing that astrocyte spatial localization is highly plastic (34). Interestingly, 434 analysis of TEMPO color distribution in superficial Layer 1 (L1) astrocytes revealed a dramatic 435 reduction in cascade progression in brains electroporated at E16.5 compared to brains 436 electroporated at earlier stages (E12.5-E14.5), suggesting most L1 astrocytes are generated 437 perinatally and do not divide much after that (Fig. S8).

439
Overall, these results demonstrate that TEMPO works in a predefined order and recapitulates the 440 inside-out pattern of cortical layer formation and the transition from neurogenesis to gliogenesis 441 in mouse.

508
Cell specification during tissue and organ formation often relies on temporal patterning: different 509 cell types are produced sequentially when exposed to temporal intrinsic or extrinsic cues. An 510 evolutionary conserved strategy for cell specification relies on deploying cascades of transcription 511 factors in a particular order (4-6). Determining the mechanisms of cell type specification and being 512 able to expand specific cell-types at will, would require tools capable of genetic manipulation in a 513 predefined order, enabling sequential activation and inactivation (temporal "pulses") of effector 514 genes in specific developmental windows. Such tools are not available in vertebrates and we 515 envision TEMPO to overcome that limitation. To this end, the modular design of TEMPO allowed 516 us to express effector genes in different temporal windows by incorporating the effector sequence 517 in frame with the correspondent reporter gene (Fig. 5A-B). Thus, TEMPO-2.0 couples activation 518 and inactivation of a reporter in a particular temporal window with that of an effector gene, leaving 519 the other reporter frames intact and allowing sequential labelling of manipulated and control cell 520 populations.

522
In multiple developmental contexts, cell cycle and cell fate decisions are strongly linked. For

533
To test this hypothesis and demonstrate we could manipulate the spatio-temporal distribution of 534 mouse cortical progenies with TEMPO-2.0, we incorporated sequences of either Cyclin D1 or 535 Cyclin B1 to be expressed in different temporal windows (Fig. 5A, B). We then co-electroporated 536 the resulting plasmid with a Cas9 containing plasmid at E12.5 and analyzed postnatal P10 mouse 537 brains. As a control, we used the original TEMPO construct, in which no perturbation was 538 introduced. A pulse of Cyclin D1 in the first temporal window (T1-Cyclin D1) slightly increased 539 the number of CFP+ lower layer neurons at the expense of CFP+ upper layer neurons. This 540 tendency was also observed in RFP+ and YFP+ neurons (Fig. 5C, D). A pulse of Cyclin D1 in the Contrary to Cyclin D1, Cyclin B1 has been linked with progenitor maintenance and later-born 556 cortical cell generation (37). We thus analyzed the effects of temporal overexpression of Cyclin 557 B1 with TEMPO-2.0, reasoning that this could shift perturbed temporal windows to later lineages, 558 including astrocytes. We found that overexpression of Cyclin B1 in the first temporal window (T1-559 Cyclin B1) increased the number of CFP+ astrocytes but not that of RFP+ astrocytes in the 560 subsequent window. In contrast, overexpression of Cyclin B1 in the second temporal window (T2-561 Cyclin B1) increased RFP+ astrocytes, while CFP+ astrocyte numbers remained comparable to 562 control samples (Fig. 5E, F). These results demonstrate that Cyclin B1 overexpression leads to a 563 shift from neurogenesis to gliogenesis in the specific temporal window where it is activated (T1 564 or T2). Interestingly, early Cyclin B1 overexpression (T1-Cyclin B1) caused a slight increase in 565 late-born YFP+ neurons which could be explained by an increase in progenitor maintenance 566 enhancing cell division capacity, thus providing more opportunities for reporter cascade 567 progression within the neurogenic period. In parallel, we observed a decrease in YFP+ astrocytes, 568 which suggests an early YFP progenitor pool exhaustion during the neurogenic phase (Fig. 5F, G).

570
Taken together, these results demonstrate that TEMPO-2.0 can be used to sequentially label and 571 manipulate neural progenitors within predefined temporal windows and simultaneously analyze 572 the phenotype of perturbed and control neurons and glia (Fig. 5H). This enables the study of 573 temporal effects of molecular factors on cell specification .   574  575  576  577  578  579  580  581  582  583  584  585  586  587  588  589  590  591  592  593  594  595  596  597  598  599  600  601  602  603  604  605  606  607  608  609  610   Establishing the association between a cell's birth timing and tissue distribution is essential to   of spatio-temporal relationships of different sublineages in single samples (Fig. 3). This is in stark 670 contrast with the high number of samples required to obtain the same conclusions with pre-existing 671 techniques (25, 29). In addition to studying neuronal lineages, the conditional TEMPO zebrafish 672 lines can be combined with other drivers to enable temporal labelling of any tissue of interest that 673 contains dividing cells.

675
Temporal access to longer lineages using TEMPO could be useful but would require adding steps 676 to the color cascade. We proved that new gRNA sequences worked robustly and had a similar 677 transition efficiency (>70%) to the one used in this study (Fig. S1, S2) so that adding more steps 678 is feasible. Modifying cascade speed may be desired as well. To that end, coupling Cas9 expression 679 to the overexpression of enzymes in charge of SSA repair may help shift the repair mechanism 680 towards SSA instead of non-homologous end joining (NHEJ) or other competing repair 681 mechanisms (21).

683
A major advantage of TEMPO is that it allows spatio-temporal reconstruction of cell histories 684 retrospectively in single samples without the need to perform live imaging. This is an essential 685 feature to be able to access cell histories in complex organisms with poor imaging accessibility, 686 such as the mouse. We thus implemented TEMPO in mouse, by targeting the transgenes to Previous studies using multicolor clonal labelling concluded that astrocytes, which are produced 708 perinatally, colonize the cortex in a non-ordered fashion where a single progenitor can produce 709 superficial and cortical parenchyma astrocytes (34). However, those studies used clonal labelling 710 techniques which lack temporal resolution and cannot distinguish subsequent generations of 711 astrocytes in the same sample. By electroporating TEMPO constructs at different stages, one can 712 subdivide different developmental processes of interest into temporal windows. As an example, 713 electroporation of TEMPO constructs into E16.5 mouse brains resulted mostly in TEMPO+ 714 astrocyte labelling and only few CFP+ neurons were labelled. This result was expected, given 715 neurogenesis is almost complete at this stage and gliogenesis is underway. Interestingly, TEMPO 716 electroporation at consecutive developmental stages showed a decrease in reporter cascade 717 progression in superficial L1 astrocytes precursors along time (Fig. S8). This suggests that most 718 superficial L1 astrocytes are generated perinatally and do not divide much after that, linking 719 proliferation mode and cell-fate. This further highlights the importance of TEMPO in revealing 720 otherwise unnoticed cellular spatio-temporal relationships.

722
Beyond spatio-temporal cell labelling, the design of TEMPO enables the activation of genetic 723 cascades, ideal for functional analysis in vivo. We demonstrated that TEMPO-2.0 allows 724 perturbation and differential labelling of temporal windows in a precise order and visualization of 725 both control and perturbed progenies in the same sample (Fig. 5). Overexpression of Cyclin B1 or 726 Cyclin D1 in different temporal windows resulted in a fate switch from neurons to glia or a change 727 in neuronal layer distribution from upper to lower layers of the perturbed progeny, respectively 728 (Fig. 5). Interestingly, we show that shifting the commitment of early progenitors to later fates 729 (astrocytes) through Cyclin B1 overexpression, seems to exclusively affect those progenies arising 730 from the perturbed window, while later-born progenitors remain capable of producing neurons and 731 have a tendency towards enhanced neurogenesis (Fig. 5E-H, S9). This suggests a feedback 732 regulation which could be compensating the early generation of astrocytes at the expense of        The following transgenic lines were used in this study: atoh1a:Gal4 (26), vsx2:Gal4 (28). We In utero electroporations were performed in embryonic day E12.5, E14.5 or E16.5 timed-pregnant 993 C57BL/6J mice (Charles River). Mice were anesthetized by using an isoflurane-oxygen mixture 994 [2% (vol/vol) isoflurane in O2]. The uterine horns were exposed and 1 μL of DNA solution was