Heterogeneous Fates of Simultaneously-Born Neurons During Early Corticogenesis

13 Neocortical excitatory neurons belong to diverse cell types, which can be distinguished by their 14 dates of birth, laminar location, connectivity and molecular identities. During embryogenesis, apical 15 progenitors (APs) located in the ventricular zone first give birth to deep-layer neurons, and next to 16 superficial-layer neurons. While the overall sequential construction of neocortical layers is well17 established, whether multiple neuron types are produced by APs at single time points of corticogenesis 18 is unknown. To address this question, here we used FlashTag to fate-map simultaneously-born (i.e. 19 isochronic) cohorts of neurons at successive stages of corticogenesis. We reveal that early in 20 corticogenesis, isochronic neurons differentiate into heterogeneous laminar, hodological and molecular 21 cell types. Later on, instead, simultaneously-born neurons have more homogeneous fates. Using single22 cell gene expression analyses, we identify an early postmitotic surge in the molecular heterogeneity of 23 nascent neurons during which some early-born neurons initiate and partially execute late-born neuron 24 transcriptional programs. Together, these findings suggest that as corticogenesis unfolds, mechanisms 25 allowing increased homogeneity in neuronal output are progressively implemented, resulting in 26 progressively more predictable neuronal identities. 27


Introduction 30
The neocortex is a six-layered structure containing a large diversity of neuronal cell types, which can 31 be defined by the combination of their birth date, laminar position, connectivity, electrophysiology, and 32 molecular identity 1-3 . Glutamatergic cortical neurons can be divided into two main classes of cells based 33 on their laminar position: deep-layer (DL) neurons (i.e. neurons which are located in layer (L)6 and 34 L5), predominantly subcortically-projecting and of which about 20 transcriptionally distinct subtypes 35 have been described, and superficial layer (SL) neurons (L4 and L2/3), predominantly intracortically-36 projecting, of which about 5 transcriptional subtypes have been described 4-7 . These neurons are born 37 from progenitors located in dorsal germinal zones below the developing neocortex, from where they 38 migrate radially to their final laminar position to differentiate. During corticogenesis, DL neurons are 39 born early (i.e. between embryonic days (E) 11.5 and E13.5), while SL neurons are born later (i.e. 40 between E14.5 and E16.5), in a so-called "inside-out" process of neuronal production [8][9][10][11] . Of note, these 41 neurons can be born either directly from apical progenitors (APs) located in the ventricular zone, or 42 indirectly via basal progenitors (BPs), which are transit-amplifying cells located in the juxtaposed 43 subventricular zone, and whose number increases as corticogenesis proceeds 5,6 . 44 While the overall sequential generation of DL and SL neurons is well established, our 45 understanding of the temporal dynamics of this process is still partial: at a given time point in 46 corticogenesis, are single neuronal subtypes produced, or is there heterogeneity within successive 47 generations of isochronic daughter neurons? In particular, given that in mice a roughly equal amount of 48 time is devoted to the generation of DL and SL neurons (3-4 days) despite a seemingly broader diversity 49 of molecular subtypes of DL neurons 2,7 , could distinct subtypes be simultaneously produced early in 50 corticogenesis? This question has been difficult to address using traditional birth dating approaches 51 such as thymidine analog pulse-labeling, since this method labels progenitors over the several hours of 52 duration of the S phase and is not selective for APs vs. BPs 12,13 . This lack of temporal precisions is an 53 obstacle to link date of birth with final fate. To circumvent these limitations, we recently developed the 54 FlashTag (FT) fate-mapping approach, which labels M-phase APs and their nascent progeny with a 55 temporal resolution of about 2h 13,14 . Using FT, here we reveal a dynamic regulation in the diversity of 56 neurons that are simultaneously produced by APs at single time points of development. At early stages 57 of corticogenesis, as DL neurons are being generated, we reveal a broad heterogeneity in the final 58 identities of simultaneously-born neurons, as assessed by a variety of laminar, connectivity and 59 molecular features. Later in corticogenesis, instead, APs give birth to neurons with more homogenous 60 features that are tightly linked to their date of birth. Using single-cell gene expression analyses, we find 61 that molecular heterogeneity across early-born isochronic neurons is already present within 24 hours of 62 birth, revealing an early-onset diversification process. This initially large neuronal fate heterogeneity 63 early in corticogenesis then narrows down as corticogenesis proceeds, suggesting the progressive 64 implementation of mechanisms controlling the fidelity of neuronal differentiation. 65

Results 67
We used FT pulse-labeling to determine the fate of simultaneously-born (i.e. isochronic) neurons on 68 sequential embryonic days (E) between E11.5 and E16.5 in the mouse primary somatosensory cortex. 69 This period includes the time of generation of DL neurons (E11.5-E13.5) and SL neurons (E14.5-70 E16.5). FT + neurons overwhelmingly correspond to directly AP-born daughter cells 13,14 ; here, to obtain 71 a better temporal resolution, in most experiments we combined this approach with the chronic delivery 72 of BrdU via an intraperitoneal osmotic pump. Using this approach, neurons born directly from APs 73 were identified as FT + BrdUcells (i.e. neurons which have not undergone intercurrent divisions 74 following FT labelling) (Fig. 1a and Fig. S1) 13,14 . 75 We first focused on the laminar fate of isochronic neurons by assessing their radial position at 76 P7, once migration is complete ( Fig. 1b and Fig. S1a). In addition to an overall inside-out lamination 77 of neurons throughout corticogenesis, this approach revealed that neurons born at early stages of 78 corticogenesis (E11.5-E13.5) distribute broadly within deep cortical layers, whereas at later stages 79 (E14.5-E16.5) neurons were laminarly compact and their radial location closely corresponded to date 80 of birth (Fig. 1b,c and Fig. S1a,c;L1 neurons,which are generated essentially at E11.5 and E12.5, were 81 not included in these analyses). Unbiased cluster analysis of isochronic early-born and late-born neuron 82 position confirmed the greater heterogeneity in radial position of early-born neurons and overall 83 bimodal pattern of distributions (Fig. 1c). Thus, while date of birth accurately predicts the final laminar 84 position of late-born neurons, this is not the case for early-born neurons (Fig. 1d), suggesting that date 85 of birth is not a stringent determinant of laminar fate early in corticogenesis. 86 These data suggest that sequential generations of neurons born at a fixed interval may have 87 overlapping laminar positions if born early, but distinct distributions if born late. To test this possibility, 88 we labelled two sequentially-generated cohorts of neurons with two sequential pulses of distinct-colored 89 FT administered at a 6-hour interval in single embryos (Fig. S2). Supporting this hypothesis, when 90 examined at P7, sequentially-born neurons had overlapping laminar distributions when born at E13.5 91 or E13.75 (Fig. S2a), whereas E15.5-and E15.75-born neurons had distinct laminar positions (Fig.  92 S2b). Thus, during early corticogenesis, sequentially-born neurons have overlapping laminar fates, 93 whereas later on, neurons with a similar birthdate interval have distinct laminar fates. Hence, date of 94 birth more tightly determines radial position as corticogenesis proceeds. 95 One explanation for the greater radial dispersion of early-born neurons could be that these 96 neurons are initially laminarly compact, but are subsequently shuffled by the migration of successive 97 incoming waves of later-born neurons. We thus examined when laminar heterogeneities first appear 98 during differentiation of early-born neurons. For this purpose, we FT pulse-labelled neurons at either 99 E13.5 or E15.5 and tracked the radial location of FT + cells at 24-hour intervals throughout 100 corticogenesis (Fig. 2). We focused our analysis on CP-located neurons to determine radial cortical 101 location. E13.5-born neurons reached the cortical plate within 48 hours of their birth, and their radial 102 distribution was broad since the onset, i.e. we did not observe an increasing dispersion of these cells 103 over time (Fig. 2a,c, Fig. S2c). E15.5-born neurons took 72 hours to reach a now expanded cortex and 104 instead progressively aligned to form a compact, homogeneous layer (Fig. 2b,c). Thus, the broad radial 105 dispersion of early-born neurons is not secondary to the subsequent arrival of later-born neurons, but 106 instead is the direct consequence of migration to a broader diversity of laminar targets. 107 We next assessed whether this laminar diversity was accompanied by a corresponding diversity 108 in the axonal target specificity of isochronic neurons. For this purpose, we used retrograde labeling 109 from distinct subcortical (thalamus and spinal cord) and intracortical (contralateral hemisphere) targets 110 to assess the axonal projections of E13.5 isochronic neurons (Fig. 3a). Depending on their laminar 111 position, isochronic neurons had distinct projections: corticothalamic projection neurons were located 112 in L6 (Fig. 3b), corticospinal projection neurons were confined to L5 (Fig. 3c) and contralaterally 113 projecting neurons were in L2/3 and L5a (Fig. 3d). Thus, the laminar diversity of isochronic early-born 114 neurons is accompanied by a corresponding diversity in their connectivity 115 We next examined whether the laminarly diverse isochronic early-born neurons also showed a 116 corresponding diversity in their molecular identity. For this purpose, we performed single-cell patch 117 RNA sequencing (Patchseq) of 49 E13.5-born neurons at P7 while recording their radial position ( Fig.  118 3e, Fig. S3). This approach revealed that the combinatorial expression of classical laminarly-enriched 119 molecular markers was congruent with their radial position (Fig. 3e). Immunocytochemistry for select 120 markers enriched in DL (TBR1, CTIP2) and SL (SATB2, CUX1) 15-18 confirmed these results (Fig. 3f). 121 Of note, while SATB2 is mostly expressed by SL neurons, it is also expressed in a fraction of DL 122 neurons and this was also the case in E13.5-born neurons, as previously shown 48 . Hence, molecularly 123 distinct types of neurons are simultaneously born during early corticogenesis. 124 Finally, we investigated the transcriptional counterparts of fate heterogeneity in isochronic 125 early-born neurons. Fate heterogeneity could either reflect a pre-mitotic process, in which neuronal 126 diversity reflects a corresponding diversity in progenitor types, or a post-mitotic process, in which 127 neurons emerge from shared progenitor pools but diversify as they differentiate in response to intrinsic 128 or extrinsic cellular processes. To distinguish between these two possibilities, we examined molecular 129 heterogeneity within isochronic cells. We took advantage of a single-cell RNA sequencing resource 130 providing the molecular identities of FT-labelled APs, 1-day-old neurons and 4-day-old neurons born 131 between E12 and E15, as well as adult mouse cortical neurons 19,20,21 . This approach revealed a transient 132 increase in the molecular heterogeneity 24-hours following neuronal birth (Fig. 4a). One-day old early-133 born neurons clustered into 3 distinct molecular types, compared with only 2 types for later-born 134 neurons. Each of the three early-born molecular types had a roughly equal proportion of E12-and E13-135 born neurons, while the two later-born molecular types mostly consisted of isochronic E14-and E15-136 born neurons, respectively (Fig. 4b). Amongst 1-day old early-born neurons, genes with the highest 137 variability were genes which are normally expressed at higher levels in later-born neurons (Fig. 4c). 138 Accordingly, while many early-born neurons did not express classical SL neuron markers, significant 139 subsets of cells did, thereby contributing to early cell-to-cell transcriptional variability (Fig. 4c, d). 140 Hence, SL neuron-like transcriptional programs are, at least partially, precociously executed in subsets 141 of early-born neurons within hours of their birth, thereby likely increasing fate diversity early in 142 corticogenesis (Fig. 4e). 143 144

Discussion 145
Our findings reveal that early in corticogenesis, simultaneously-produced neurons born from APs have 146 heterogeneous fates, whereas later on, fate control becomes tighter and the final identity of isochronic 147 neurons is more homogeneous. Hence, in the first half of corticogenesis, the correlation between date 148 of birth and final neuronal identity is relatively loose, but then tightens as corticogenesis unfolds. 149 Simultaneous production of neurons with distinct laminar fates has been reported in species 150 with large germinal zones and long neurogenic periods, such as primates 11,21,22 . Fate diversity in these 151 cases may in part reflect the simultaneous production of neurons born from APs and BPs, which have 152 distinct final identities 23,24 , since thymidine analog birthdating used in these studies undistinguishably 153 labels all progenitor types 13 . In contrast, the FT birthdating approach used here (enhanced by mutually 154 exclusive chronic BrdU birthdating) allows specific labeling of isochronic AP progenies 13 , hence 155 revealing a functionally meaningful fate heterogeneity of AP-born neurons early in corticogenesis. 156 Homogenization of AP neuronal output as corticogenesis proceeds suggests a progressive 157 implementation of mechanisms allowing increased fidelity in final neuronal output. Early-born APs are 158 rapidly cycling cells 25-27 with active epigenetic control programs 20,28 . Despite some level of increase in 159 AP transcriptional heterogeneity across corticogenesis, early APs may display primed chromatin states 160 and poised promoters 29,30 , allowing early-born neurons to differentiate across multiple (and even 161 simultaneous) paths and to be refined post-mitotically. Supporting this possibility, 1-day old neurons 162 transiently co-express markers of DL and SL neurons 20,31,32 and components of the polycomb complex 163 2 (PRC2), which acts to regulate access to transcriptional sites, are strongly expressed by early but not 164 late APs 20 . Later in corticogenesis, the progressive implementation of epigenetic gatekeeping 165 mechanisms may allow more robust transcriptional programs to consolidate, giving rise to more 166 standardized, albeit less innately diverse, neuronal cell types 33,34 . Related to this, the significant increase 167 in AP cell cycle length during corticogenesis 25-27 may allow more time to homogenize transcriptional 168 output later in development since short cell-cycle length acts as a transcriptional filter for long 169 transcripts, which may increase inter-cell variability 35 .

Acknowledgments 346
We thank the Imaging Platform of the University of Geneva, A. Benoit for technical assistance and 347 members of the Jabaudon laboratory for constructive comments on the manuscript. The Jabaudon while epifluorescence images were obtained on an Eclipse 90i Nikon epifluorescence microscope with 490 a 10x, 0.44 micron/pixel objective. 491

Image quantifications 492
Photomicrographs quantified in Fig.1 and Fig.S1 were processed for cell detection on FT channel using 493 MetaXpress software (v.5.1.0.41, Molecular Devices). Detected cells properties (x and y position, FT 494 and BrdU intensities and size) were extracted using a custom Matlab script. Colocalization tests were 495 automatically performed with same threshold for all images. FT + BrdUcells were filtered for FT 496 intensity (> median value) and BrdU intensity (<20% of all cells) for each section. High FT signal 497 thresholding allows to select for FT + BrdUneurons, justifying the use of top 10 % FT signal as a way 498 to detect directly born neurons without chronic BrdU (Fig. S1). Radial position was measured by manual 499 determination of the coordinates of pial surface and subplate lower border. To compare radial position 500 across animals, the thickness of the cortex was normalized across animals, as was the thickness of CP 501 for samples of brains before P7. In Fig. 1c, hierarchical clustering was performed with Euclidean 502 distance calculation using centered radial positions data. A two-sample Welch test was performed 503 between E11.5-, E12.5-, and E13.5-injected brains against E14.5-, E15.5-, and E16.5-injected ones. 504 Normalized density distribution was performed with radial position normalized to the mean per pup. 505 For Fig. S1c, curve fitting was performed through Loess method. All other image quantifications were 506 performed using standard Fiji functionalities 51 . Radial position was measured by recording coordinates 507 of manually-counted cells. Colocalization tests were performed by manual analysis of confocal images. 508 The Kolgorov-Smirnov test in Fig. S2 was performed between E13.5, E13.5 + 6h (a), E15.5-P7 and 509 E15.5 + 6h cells (b). In Fig. 2c, the standard deviation of radial position of E13.5-and E15.5-born cells 510 was used to perform linear interpolation over time of collection. All data analysis scripts were custom-511 prepared in R. Packages used: ggplot2 52 , reshape2 53 , stringr, stringi, plyr, XML, SpatialTools, 512