Radial Glial from mammalian developing neocortex can perform symmetric proliferative divisions in vitro

Radial Glial progenitors in the mammalian developing neocortex have been shown to follow a deterministic differentiation program restricted to an asymmetric only mode of division. This feature contrasts with previous studies and with other developmental systems, such as the developing spinal cord, the retina, epidermis, airway epithelium, germline, and the intestine, where differentiation takes place based on probabilities that can change overtime and other modes of division are possible. Here, we combine experimental, computational and theoretical tools to show that Radial Glial cultured in vitro can divide symmetrically, and that the balance between the different modes of division, as well as the cell cycle length, can be modulated by external signals, such as Fibroblast Growth Factor. Our results suggest that the constraint of deterministic and asymmetric mode of division that Radial Glia exhibit in vivo may not be an inherent property of this particular cell type, but a feature induced by the complex organized pseudo-stratified structure of the mammalian developing neocortex. Significance Statement It is well established that, during the neurogenesis of the mammalian neocortex, Radial Glial can only divide asymmetrically. To understand the features that set this important restriction, we analyze in vitro cultures from mouse developing neocortex during the differentiation of Radial Glial into terminally differentiated neurons. Our results show that Radial Glial cultured in vitro also divide symmetrically. Moreover, the balance between the division modes can be modulated by external signals, such as Fibroblast Growth Factor. In conclusion, our results suggest that the constraint of deterministic and asymmetric mode of division of Radial Glial is not an inherent property of this particular cell type, but a feature induced by the complex organized pseudo-stratified structure of the mammalian developing neocortex.

cell is receiving. In this context, the fate at the single cell level 23 is unpredictable and the balance between proliferation and 24 differentiation is regulated at the level of the population (10). 25 On the other hand, deterministic models of stem cell differ-26 entiation assume that the fate of the progeny is fixed and, 27 therefore, the correct balance between the numbers of different 28 types of neurons is achieved at the single cell level (11). 29 The dynamics of differentiation is often characterized based 30 on the fate of the two daughter cells of a cell division relative 31 to each other (12). This way, proliferating progenitors can per-32 form pp (progenitor-progenitor), pd (progenitor-differentiated) 33 and dd (differentiated-differentiated) divisions (13). 34 In this context, differentiation in the developing chick spinal

Significance Statement
It is well established that, during the neurogenesis of the mammalian neocortex, Radial Glial can only divide asymmetrically. To understand the features that set this important restriction, we analyze in vitro cultures from mouse developing neocortex during the differentiation of Radial Glial into terminally differentiated neurons. Our results show that Radial Glial cultured in vitro also divide symmetrically. Moreover, the balance between the division modes can be modulated by external signals, such as Fibroblast Growth Factor. In conclusion, our results suggest that the constraint of deterministic and asymmetric mode of division of Radial Glial is not an inherent property of this particular cell type, but a feature induced by the complex organized pseudo-stratified structure of the mammalian developing neocortex.
MLT performed research and wrote the manuscript, NPC, perfomed reserach, DGM designed research, performed research and wrote the manuscript.
Authors declare no competing interests.   BrdU (27), EdU (28, 29) and other thymidine analogs 106 constitute the most used tool to estimate the cell cycle length 107 of cells in many contexts (30). The methodology is based 108 on the replacement of endogenous thymidine during DNA 109 synthesis with traceable compounds (31, 32). The length of 110 the average cell cycle is then inferred from the dynamics of 111 the incorporation of these compounds into the DNA of cycling 112 cells using well established methods (33).

113
To estimate the average cell cycle length of the population, 114 samples are cultured in the presence of EdU and then fixed at 115 different time points (corresponding to different times of EdU 116 incorporation). Combined nuclear Hoechst staining with EdU 117 detection assay and immunostaining against Sox2 is used to 118 identify all progenitors that have passed through S-phase for 119 each EdU incubation time.

120
The cell cycle length T and the growth fraction γ are 121 calculated using the standard cumulative curve methodology 122 based on linear regression (see Methods). Results are shown in 123 Fig. 2. The data (Figs. 2C,D) reveals that γ remains at around 124 72% for both conditions tested, while T depends strongly on 125 the culture conditions (T=35.2 ± 3.5 hours for SC, T= 24.7 126 ± 2.0 hours for SC+FGF). In conclusion, our results show 127 that FGF stimulation shortens the average cell cycle length in 128 cultures of RG in vitro.

D R A F T
corresponding sigmoidal curve fitting is also plotted (green, red, and blue lines for RG, neurons and total cells, respectively). 264 The data shows that an initial regime of reduced change in 265 cell numbers is followed by an increase in both cell types until     the maximum change in the differentiation dynamics occurs 324 around 36-37 hpp, coinciding with the minimum of T of both 325 conditions. In conclusion, quantification of the cell numbers 326 using the branching process framework suggest that the pres-327 ence of FGF in the culture media speeds up the cell cycle 328 time while also reducing the differentiation rate. Interestingly, 329 our data also shows an increase in the growth fraction γ in 330 conditions of SC+FGF compared to SC, when monitoring 331 the cycling progenitors using KI67 staining (Supplementary 332 Figure S2C-D), in agreement with previous studies (63).  Fig. 6. The number of progenitors labeled with 370 EdU does not change significantly in SC conditions, consistent 371 with a large proportion of asymmetric divisions (i.e, one EdU+ 372 RG produces two EdU+ cells: one RG and one neuron, so the 373 amount of EdU+ RG remains constant). On the other hand, 374 in conditions of SC+FGF, we see a statistically significant 375 (P<0.05) increase in the number of EdU+ RG when comparing 376 "pulse" and "chase" time points. This result shows that some 377 of the RG originally labeled with the short EdU pulse, divided 378 and produced more than one RG per division, showing that, 379 as predicted by the branching formalism, RG are capable to 380 undergo symmetric proliferative pp divisions in vitro. situation, the branching process framework is able to estimate 415 the rates of each of the three modes of division (14). This 416 prediction for the case of RG in culture is shown in Fig. 5E-F, 417 where we can see that the predominant mode of division is 418 pp (green). This symmetric mode of division is even more 419 probable in conditions of SC+FGF, to the expenses of a re-420 duction in pd and dd. A scheme that illustrates our findings is 421 shown in Fig 6C. In brief, the single mode of division observed 422 in vivo contrast with the probabilistic scenario observed in 423 vitro, where all modes of division are possible. Changes in 424 the culture conditions can shift the balance between the three 425 modes of division, and increase the rate of pp divisions to the 426 expenses of the other modes.

427
A detailed analysis of the dynamics of vertebrate neuroge-428 nesis involves a careful characterization of the rate of division. 429 The most direct method to measure the cell cycle length re-430 quires to monitor the time between consecutive mitotic evens 431 at single cell resolution (64). Unfortunately, due to the high 432 degree of variability, many cells in a population need to be 433 sampled, segmented and tracked simultaneously to obtain an 434 accurate value, even when dealing with clonal samples (65).

435
Indirect methods based on cumulative incorporation of 436 thymidine analogs perform well in conditions of constant pro-437 liferation and differentiation, but they are not designed to 438 study systems where the cell cycle changes overtime, which 439 is is potentially the case in many developmental systems. In 440 these conditions, the Branching Process formalism and the 441 Pulse-Chase outperform cumulative curve methods. On the 442 other hand, the Pulse-Chase method requires experiments that 443 are longer that cell cycle length. Therefore, the toxic effect of 444 the labeling agent for long periods of time may affect strongly 445 the normal cell cycle progression (34, 35). A clear advantage 446 of the Branching Process is that it does not involve manipula-447 tion of the samples before fixation, so there is no interference 448 with the normal progression of the cell cycle. In addition, the 449 Branching Process formalism also provides the correct value 450 of T with temporal resolution, and the measurement of the 451 average differentiation rate, (also with temporal resolution).

452
Several studies have shown that the length of G1 phase 453 increases progressively when neurogenesis starts, resulting in a 454 overall increase of the cell cycle (48)(49)(50)(51)(52)(53). Alternatively, others 455 studies show that the cell cycle length is shorter in neurogenic 456 divisions, compared to proliferative divisions (46,47,(54)(55)(56), 457 due mainly to a shortening in S-phase. Our results show that 458 FGF promotes pp divisions and shortens cell cycle, consistent 459 with the hypothesis that proliferative divisions have a shorter 460 cell cycle, maybe via a shortening of G1-phase (similarly to 461 insulin-like growth factor (66, 67)). A careful characterization 462 of how FGF affects each phase of the cell cycle it is far from 463 the scope of this contribution.

464
The culture and differentiation of RG cells in vitro provides 465 a very good framework to study basic features that orches-466 trate the formation of the mammalian neocortex. In brief, 467 the system provides a well controlled environment where the 468 effect of signaling molecules and other conditions can be tested 469 reliably, while providing easier manipulation and imaging com-470 pared to studies performed in vivo. We use this framework to 471 study the features that restrict the division mode of RG as 472 deterministic versus the more common probabilistic scenario 473 observed in many other scenarios. Our combined experimen-474 tal/computational/theoretical approach can be also used to 475 test the effect of other signaling networks by quantifying the 476 cell cycle and mode of division after small molecule inhibition 477 and comparison with a control culture.