Sequential transcriptional waves direct the differentiation of newborn neurons in the mouse neocortex

See allHide authors and affiliations

Science  25 Mar 2016:
Vol. 351, Issue 6280, pp. 1443-1446
DOI: 10.1126/science.aad8361

Tracking neuronal transcriptional programs

Early in brain development, cortical neurons are born near the ventricles, then migrate to their functional destinations. Telley et al. used a fluorescent labeling technique to see what transcripts characterize these earliest stages of neural development. Waves of transcriptional programs are initiated, then passed by as the neuron progresses from proliferative to migratory and finally to connectivity phases.

Science, this issue p. 1443


During corticogenesis, excitatory neurons are born from progenitors located in the ventricular zone (VZ), from where they migrate to assemble into circuits. How neuronal identity is dynamically specified upon progenitor division is unknown. Here, we study this process using a high-temporal-resolution technology allowing fluorescent tagging of isochronic cohorts of newborn VZ cells. By combining this in vivo approach with single-cell transcriptomics in mice, we identify and functionally characterize neuron-specific primordial transcriptional programs as they dynamically unfold. Our results reveal early transcriptional waves that instruct the sequence and pace of neuronal differentiation events, guiding newborn neurons toward their final fate, and contribute to a road map for the reverse engineering of specific classes of cortical neurons from undifferentiated cells.

During neocortical development, distinct classes of neurons assemble to form local and long-range circuits. Although class-specific genes and features identify cortical neuron types relatively late in differentiation (15), early postmitotic fate specification programs have been inaccessible. Here, we describe the dynamic transcriptional activity controlling layer 4 (L4) excitatory neuron birth and differentiation in the mouse neocortex.

Mammalian cortical progenitor cells in the ventricular zone (VZ) undergo DNA synthesis [S-phase, susceptible to bromodeoxyuridine (BrdU) labeling] at the basal border of the VZ and mitosis (M-phase, lasting about an hour at midcorticogenesis in mice) when their soma is apically located, adjacent to the ventricular space (6, 7). At this location, mitotic cells are susceptible to labeling by intraventricular injection of carboxyfluorescein esters [“FlashTag” (FT)], which bind to and fluorescently label intracellular proteins (8). The short extracellular half-life of FT in the mouse ventricular space ensures effective pulse-labeling of juxtaventricular dividing cells (Fig. 1A and fig. S1). Intracellularly, FT is linearly diluted at each mitosis, such that fluorescence reflects the number of cell divisions that have occurred since the time of labeling (fig. S1, D and E, and movie S1) (8). FT+ newborn cells synchronously moved away from the ventricular wall within 3 hours of labeling (Fig. 1A, bottom), reached the subventricular zone (SVZ) within 12 hours, and entered the cortical plate (CP) 24 to 48 hours after mitosis (Fig. 1B). Isochronic cohorts of VZ cells born at the time of injection can thus be specifically identified and tracked during their initial differentiation.

Fig. 1 FT labels time-locked cohorts of newborn VZ cells during corticogenesis.

(A) (Top) Schematic representation of the FT labeling principle. (Bottom) Pulse-labeling of isochronic mitotic cells using FT at E14.5. PH3, phospho-histone 3, an M-phase marker. (B) Isochronic cohorts of FT+ cells radially migrate from the VZ to the CP. PAX6 and TBR2 delineate the VZ and SVZ. (C) E14.5 FT labeling identifies a subset of L4 neurons at P7. (D) E14 (FTV+) and E14.5 (FTG+) VZ-born neurons occupy distinct sublaminae within L4. Cx, cortex; IZ, intermediate zone. See also figs. S1 and S2.

The laminar fate of FT+ neurons was linked to the day of FT injection at all ages examined [embryonic day (E) 11.5 to 17.5] (fig. S2 and Fig. 1C). At postnatal day (P) 7, when neuronal migration is complete, E14.5-labeled FT+ neurons were restricted to a sublamina of L4 (Fig. 1C). These neurons were born at the time of the FT pulse, not later, because they mostly remained unlabeled after continuous BrdU administration beginning at the time of the FT pulse (fig. S1, B to D). Injection of FT at E14 and E14.5 using two dye colors in the same embryo showed two distinct populations of labeled neurons within L4 at P7, revealing a tight relationship between time of birth and final radial location, even within a single layer (Fig. 1D). Thus, we used E14.5 FT injections to label L4 neurons in vivo from the time of mitosis in the VZ and track their early molecular differentiation.

We observed that newborn cells sequentially expressed PAX6, a VZ marker, TBR2 a SVZ marker, and the early neuronal protein TBR1 (9, 10) within the first 48 hours after mitosis (fig. S3). This reveals a highly dynamic cellular process characterized by overlapping signature shifts in protein expression. For an unbiased account of the transcriptional programs active just after cell birth in single cells, we isolated E14.5-born FT+ cells 6, 12, 24, and 48 hours after mitosis by using cortical microdissection followed by fluorescence-activated cell sorting (FACS). We characterized transcriptional activity using single-cell RNA sequencing in microfluidically isolated single cells (fig. S4, A and B) (1, 11, 12).

To determine the sequence and pace of early differentiation processes, we first examined the expression dynamics of a core set of genes involved in proliferation, neurogenesis (i.e., which promote differentiative divisions), and neuronal differentiation. We found that proliferative (P), neurogenic (Ng), and neuronal (N) transcripts were sequentially expressed: All P transcripts were repressed first, Ng transcripts showed delayed repression, and N transcripts were induced after cell division (see fig. S4C and table S1). The closely timed repression of P and Ng transcripts reveals that exit from the cell cycle and initial postmitotic specification are partially overlapping rather than strictly sequential processes. We used these program-specific dynamics to identify a broader set of proliferative-type (Ptype), neurogenic-type (Ngtype), and neuronal-type (Ntype) transcripts (fig. S4D and data table S1). The functional relevance of these three programs was supported by the enrichment of Ptype, Ngtype, and Ntype transcripts in the VZ, SVZ, and CP, respectively; differential enrichment in specific gene ontology terms; and sequential expression in single cells (fig. S4, D and E, and fig. S5). These findings reveal the highly dynamic unfolding of proliferative, neurogenic, and neuronal programs after mitosis in vivo.

Two main classes of juxtaventricular cells are initially labeled by FT in the VZ: (i) progenitor cells and (ii) newborn neurons (Fig. 1A and fig. S1D). We sought to identify neuron-specific transcriptional programs by distinguishing neurons from progenitors. For this purpose, we used a machine-learning approach to cluster cells based on transcriptional expression signatures (13).

This approach delineated distinct groups of cells, which were identified as progenitors [genuine apical progenitors and daughter basal progenitors (14)] and neurons (early and late populations) (Fig. 2A). Apical progenitors and daughter basal progenitors were distinguished based on differential expression of markers such as Eomes and Btg2 [which are enriched in basal progenitors (see references in table S1)] (Fig. 2B) and differential enrichment in Ptype genes (apical > basal progenitors), including Nes and Sox2 (Fig. 2C). Accordingly, cells in the apical progenitor cluster were mostly 6 hours old, whereas newborn basal progenitor identity was more distinct after 12 hours (Fig. 2A, top right). Neurons expressed core neuronal genes such as Tbr1 and Mapt, and were enriched in Ntype genes (Fig. 2, B and C). With apical progenitors and their daughter neurons now distinguishable, we identified cell-type specific, stage-specific transcripts by comparing gene expression at each developmental age (data table S2 and fig. S8). Consistent with this classification, apical progenitor genes were predominantly expressed in the VZ, basal progenitor genes extended into the SVZ, and neuronal transcripts showed stage-specific sequential expression in the VZ, SVZ, and CP (Fig. 2D and fig. S9). Hierarchical relationship analysis revealed that apical progenitors are clearly distinct from daughter basal progenitors and neurons (Fig. 2A, bottom right), further supporting the lineage relationships identified above. Segregation of type-specific transcripts between newborn neurons and their progenitors was detected as early as 6 hours after mitosis (Fig. 2E). This suggests that type-specific transcripts can be premitotically segregated or differentially stabilized in nascent postmitotic neurons versus progenitors. Together, these data identify progenitor and neuron-specific transcripts activated after cell division and reveal rapid cell-type specific segregation and regulation of transcripts after mitosis.

Fig. 2 Identification of newborn cortical neurons.

(A to C) Unbiased clustering delineates neurons from progenitors. (A) Apical progenitors, daughter basal progenitors, and newborn neurons can be distinguished by unbiased clustering (left), temporal distribution (top right), hierarchical clustering (bottom right), and [(B) and (C)] expression of specific markers and Ptype, Ngtype, and Ntype transcripts. (B, bottom right) Schematic in (B) provides examples of type-specific genes presented in figs. S6 to S9. Although basal progenitors eventually give rise to neurons (dotted arrow), this progeny is not included in the current data set because FT+ cells are essentially VZ-born (see fig. S1). (D) Spatial segregation of progenitor and neuron-specific transcripts with in situ hybridization (24). Values represent median expressions for several transcripts. (E) Rapid segregation of cell-type-specific transcripts after cytokinesis. P < 0.0001 for all values compared to 6-hour apical progenitor (AP) values.

To establish a real-time quantitative account of differentiation programs in newborn neurons, we used an unsupervised approach in which single-cell expression profiles are temporally ordered based on distinct intermediate differentiation states (Fig. 3A) (15, 16). This method appropriately ordered neurons along a pseudotime axis, with 6-, 12-, 24-, and 48-hour-old neurons being progressively staggered along this time line (Fig. 3A). This allowed us to reconstruct the expression dynamics of all transcripts across this pseudotime axis and generate a high-resolution transcriptomic atlas of the first 48 hours of L4 cortical neuron development (Fig. 3B) (see for the data set of all transcripts). The expression dynamics of classical P (Sox2), Ng (Neurog2), and N (Tbr1) transcripts were consistent with their function (Fig. 3B). Neurod2 was identified as an early-onset neuronal transcript; accordingly, NEUROD2 protein was detected in newborn apical VZ neurons, whereas this was not the case for TBR1 (Fig. 3B, inset).

Fig. 3 Real-time functional transcriptomics of early postmitotic neurons in vivo.

(A) Neurons are staggered by age along the pseudotime axis. (B) Gene expression dynamics for classical proliferative (Sox2), neurogenic (Neurog2), and neuronal (Tbr1) genes. Neurod2 is expressed more strongly and earlier than Tbr1. QR code,, for access to dynamics of all transcripts. (C) Unbiased clustering of genes based on expression dynamics reveals distinct transcriptional waves with sequential expression peaks (black arrowheads). Illustrative transcription factors are provided for each wave (see also fig. S10 and data table S3). (Right) Chromatin immunoprecipitation sequencing–identified targets of NEUROD2 (18) are enriched in its own wave but also are present across waves (see also fig. S11). (D) Summary of wave dynamics related to developmental time. (E) Gene ontology term–based analysis. Colors correspond to wave numbers. (F) Double-strand DNA breaks are transiently increased in 12-hour-old cells, as indicated by the presence of phosphorylated histone 2AX (γH2AX) (25). **P < 0.001.

Clustering of expressed transcripts based on their expression dynamics showed how transcriptional networks are organized in newborn neurons. Directly after mitosis, waves of transcriptional programs sequentially unfold, each including temporally distinct complements of transcription factors and networks (Fig. 3, C and D, fig. S10, data table S3, and movie S2). To understand the temporal organization of the molecular pathways across differentiation, we focused on the genetic network of Neurod2, a wave 5 transcription factor required for L4 neuron maturation and whose target genes have been identified in the E14.5 neocortex (17, 18). The temporal distribution of NEUROD2 target genes across the distinct waves was not random (Fig. 3C, right, and data table S4). Instead, NEUROD2 targets were strongly enriched in its own wave (e.g., Nrn1 and EphB2), in line with its role in neuritogenesis, but also present across waves, including in wave 1, where targets include cyclins and cyclin-dependent kinases such as Ccnd2, Ccnd3, and Cdk13, which control cell cycle progression. NEUROD2 may therefore act not only on isochronically expressed genes but also across differentiation. Consistent with a repressive action on wave 1 targets, overexpression of NEUROD2 through in utero electroporation into VZ progenitors induced exit from the cell cycle, as indicated by decreased numbers of Ki67+ VZ cells (fig. S11). Single transcription factors can therefore control distinct differentiation events through combinatorial actions on a variety of temporally gated genetic targets and networks.

Ontology term analysis of the transcriptional waves identified successive functional differentiation events in newborn neurons (Fig. 3E). We observed an initial rapid (6 to 12 hours after mitosis) repression of proliferation-associated transcripts (e.g., Arx, Notch1, and Sox9) and a surge in transcripts associated with ribosome biogenesis and translation (e.g., Etf1, Rpl13a, and Rpl12), which might reflect nucleolar reassembly and increased protein synthesis. Transcripts associated with DNA repair (e.g., DNA2, Ddb1, and Exo1) were transiently increased after mitosis, suggesting postmitotic genetic instability. Consistent with this possibility, DNA double-strand breaks were significantly increased in 12-hour-old neurons (Fig. 3F). This reveals a critical period after mitosis during which neocortical neurons are susceptible to somatic mutations and where clonal mosaicism could be generated (19, 20). Twelve-hour-old neurons already initiated differentiation programs related to late-occurring processes such as synaptogenesis, revealing an early transcriptional poise in anticipation of terminal differentiation events. Finally, chemotaxis-associated transcripts (e.g., Ephb1, L1CAM, and Nrp1) peaked around 42 hours after birth, while neurons are reaching the CP, providing a molecular framework for input-dependent differentiation processes (21).

Finally, we examined whether the distribution of transcript expression across waves instructs the sequence and pace of neuronal differentiation events. For this purpose, we prematurely expressed a late-wave transcript, Nrn1, which normally peaks ∼30 hours after mitosis (wave 5, Fig. 4A) and controls L4 neuron maturation through promotion of neuritogenesis (22, 23). We hypothesized that heterochronic expression of this normally late-occurring gene could bypass early processes and accelerate neuronal differentiation. Indeed, in utero electroporation of Nrn1 led to premature transition to neuronal identity and precocious expression of NEUROD2 (Fig. 4, B and C). Premature acquisition of this neuronal trait was detectable as early as 12 hours after cell birth, as revealed by assessing NEUROD2 expression within an isochronic 12-hour-old cohort of FT+ cells with mosaic overexpression of NRN1 (Fig. 4D). Finally, precocious molecular maturation was associated with an early loss of migrational capacity, leading to neuronal mispositioning at birth (Fig. 4E). Therefore, the precise timing of early differentiation programs is critical not only for the execution of single-cell differentiation events but also for the successful organization of the cortical networks to which it belongs. Precise and dynamic temporal control over the expression of even single genes thus controls the sequence and pace of neuronal differentiation, which is essential for circuit assembly.

Fig. 4 Early expression of the late-wave gene Nrn1 induces premature neuronal differentiation.

(A to D) Premature expression of Nrn1 (A) leads to a forward shift in neuronal differentiation by inducing cell-cycle exit (decreased number of Ki67+ progenitors) (B) and premature neuronal differentiation (increased number of ND2+ neurons) (C). This effect occurs within 12 hours of birth, as assessed within an isochronic 12-hour-old cohort of FT+ cells (D). (A, left) In situ hybridization (24). (E) NRN1-overexpressing neurons undergo premature migrational arrest before reaching L4. *P < 0.05; **P < 0.001; ***P < 0.0001. ND2, NEUROD2.

Our data provide a comprehensive transcriptional blueprint outlining the dynamic acquisition of neuronal identity in vivo. We show that early neuronal differentiation is directed by a series of transcriptional waves whose proper sequence is critical for normal progression through development. These waves provide discrete time windows during which specific transcriptional complexes are present simultaneously and can interact. These transient combinatorial transcriptional niches could act as sequential checkpoints during the course of differentiation, combinatorially coding for specific cell fates. These results build a road map for reverse engineering of cortical neuron subtypes from undifferentiated cells and provide a set of genetic targets for identification and directed differentiation of progenitors and nascent neurons.

Supplementary Materials

Materials and Methods

Figs. S1 to S11

Table S1

Movies S1 and S2

Data Tables S1 to S4

Database S1

Reference (26)

References and Notes

  1. Acknowledgments: Annotated data are available at We thank R. Hevner for the gift of the TBR1 antibody; A. Benoit, M. Lanzillo, and the Genomics Platform and FACS Facility of the University of Geneva for technical assistance; and E. Azim, A. Carleton, S. Tole, and the members of the Jabaudon laboratory for comments on the manuscript. Work in the Jabaudon laboratory is supported by the Swiss National Science Foundation (PP00P3_123447), the Synapsis Foundation, and the Brain and Behavior Foundation (NARSAD Grant). S.N. is funded by the Swiss SystemsX Interdisciplinary PhD Grant 51PHI0–141994. S.G. and I.S. are supported by the iGe3 PhD Award. A.D. is funded by the Swiss National Science Foundation and the National Center of Competence in Research (NCCR) Synapsy. The authors declare no competing interests. The supplementary materials contain additional data.
View Abstract

Stay Connected to Science

Navigate This Article