NeuroGT: A brain atlas of neurogenic tagging CreER drivers for birthdate-based classification and manipulation of mouse neurons

Summary Neuronal birthdate is one of the major determinants of neuronal phenotypes. However, most birthdating methods are retrospective in nature, allowing very little experimental access to the classified neuronal subsets. Here, we introduce four neurogenic tagging mouse lines, which can assign CreER-loxP recombination to neuron subsets that share the same differentiation timing in living animals and enable various experimental manipulations of the classified subsets. We constructed a brain atlas of the neurogenic tagging mouse lines (NeuroGT), which includes holistic image data of the loxP-recombined neurons and their processes across the entire brain that were tagged on each single day during the neurodevelopmental period. This image database, which is open to the public, offers investigators the opportunity to find specific neurogenic tagging driver lines and the stages of tagging appropriate for their own research purposes.

Correspondence tathirat@nig.ac.jp In brief Hirata et al. develop neurogenic tagging CreER driver mice and the NeuroGT database to showcase the mouse lines. The resource can be used to classify neurons on the basis of their generation timing and manipulate the classified neuron subsets for research purposes.

INTRODUCTION
During development, neurons are generated over a protracted time window. Mounting evidence shows that neurogenic timing, which is often referred to as the neuronal birthdate, has immense impacts on neuronal phenotypes. In the neocortex, the birthdate determines layer positioning, connection patterns, and molecular and physiological properties of neurons (McConnell, 1989;Lodato and Arlotta, 2015). The chronological specification of neuronal fates is not a unique property of the neocortex but rather a generally conserved strategy in various nervous systems for generating neuronal diversity (Suzuki and Hirata, 2013). Even without affecting the intrinsic molecular differences, the birth timing itself could cause differential neuronal phenotypes because neurons with different birthdates are influenced by a changing environment (Hirata and Iwai, 2019). Because of its intimate relationships with various neuronal phenotypes, neuronal birthdating has been a reliable standard for the classification of MOTIVATION Neuronal birthdates are commonly used to classify neurons. However, the current nucleotide-based birthdating method is retrospective in nature and does not allow the manipulation of classified neurons. To overcome this limitation, we developed neurogenic tagging mouse lines that can induce CreER-loxP recombination in neurons with similar birthdates, and the recombination tag can be subsequently used for experimental manipulation of the classified neuron subsets. To encourage the use of this resource, we launched the NeuroGT database, which contains image data of the tagged neurons across the entire brain.
neurons in various nervous systems for over half a century (Bayer and Altman, 1987;Govindan et al., 2018).
Traditionally, neuronal birthdating has been performed by using nucleotide analogs, such as radioactive thymidine and, more recently, bromodeoxyuridine (BrdU) or 5-ethyynyl uridine (EdU), which are incorporated into the DNA during the S phase to mark the final round of the cell cycle. Although these techniques have substantially improved, they are still descriptive histological techniques to determine neuronal birthdates in sacrificed animals. Recently, we developed a neurogenic tagging method by which tamoxifen (TM)-dependent CreER (Feil et al., 1997) recombination was induced in olfactory bulb neurons in a birthdate-dependent manner . Given that this technique is based on irreversible loxP recombination, the tag can be used later to explore the birthdate-classified neuronal subset in various ways.
This neurogenic tagging method uses driver mouse lines in which CreER is expressed only transiently within a short time window immediately after neuronal fates are committed (Figure 1A). Consequently, a single administration of TM at a certain developmental stage induces the recombination of loxP sequences only in the cells that express CreER. In the previous study, we used the enhancer of the neurog2 gene to achieve the transient expression of CreER in a bacterial artificial chromosome (BAC) transgenic mouse. Although the biological principle underlying this method differs from that of nucleotide-based birthdating, the CreER driver that we developed achieved birthdate-dependent neuron tagging in the brain regions such as the olfactory bulb  and cerebellum Zhang et al., 2020), mirroring the endogenous expression patterns of neurog2.
To cover other brain regions, we developed three neurogenic tagging driver lines by using the neuronal differentiation genes neurog1, neurod1, and neurod4 ( Figures 1B and 1C), all of which are basic-helix-loop-helix transcription factors that are transiently expressed during the maturation phase of neurons (Kim et al., 2011;Sudarov et al., 2011;Aprea et al., 2014). Together with the previous line developed by using neurog2, the collection of driver lines covers most of the mouse nervous systems. Our aim is to contribute this resource to the scientific community. To encourage the use of this resource, we have developed a brain atlas of neurogenic tagging mouse lines (NeuroGT). This database contains section images of the entire brain visualized for tagged neurons as well as their processes; these images were obtained from postnatal mice that received a single TM injection on each day during the neurogenetic period. Researchers interested in particular brain regions can find appropriate mouse lines and tagging stages for their research purposes.

RESULTS
NeuroGT and neurogenic tagging driver mice The neurogenic tagging drivers Neurog2 CreER (G2A), Neuro-g1 CreER (G1C), Neurod1 CreER (D1B), and Neurod4 CreER (D4A) were developed as described in STAR Methods and are deposited at the RIKEN Center for Biosystems Dynamics Research (RIKEN BDR) ( Figures 1B and 1C). To showcase the tagged neuron images comprehensively, each neurogenic tagging line was crossed with Tau mGFP-nLacZ mouse (Hippenmeyer et al., 2005), which is a global neuronal Cre reporter that expresses dual nucleus-and membrane-localized reporters under a constitutive neuronal promoter after the excision of the loxP-STOP- Figure 1. Neurogenic tagging and the driver mouse lines (A) In neurogenic tagging mouse lines, CreER is transiently expressed within a short neurogenic time window in various neurons. TM injection on a single day during E9.5-E18.5 induces loxP recombination only in cells expressing CreER. The mouse embryos were raised up to P7, and the brains were sampled for visualization of tagged neurons. (B) The BAC genomic constructs recombined with the CreER cassette that were used to generate the four neurogenic tagging driver lines. Boxes show the exons, in which the shaded part represents originally protein-coding sequences and the white part untranslated sequences. The approximate lengths of the genome upstream and downstream of the CreER cassette are indicated under each construct. The bottom panel shows a schematic of the gene structure of the Tau mGFP-nLacZ reporter. Both reporter proteins, mGFP and nls-bGAL, were used to visualize tagged neurons. (C) Four neurogenic tagging mouse lines were characterized in this study. The mice can be obtained from RIKEN BDR (http://www2.clst.riken.jp/ arg/TG%20mutant%20mice%20list.html).
2 Cell Reports Methods 1, 100012, July 26, 2021 Article ll OPEN ACCESS loxP cassette ( Figure 1B). Staged pregnant mice were then intraperitoneally injected with TM only once during the gestation stages at embryonic day 9.5 (E9.5)-E18.5 (hereafter called TM9.5-TM18.5, referring to TM injection stages), and the tagged offspring were delivered and raised up to postnatal day 7 (P7). Figure 2 shows X-gal-stained images of whole-mount brains tagged by the four drivers at different TM stages. The tagged neurons stained blue by the nucleus-localized (nls)-bGAL reporter were distributed in different brain regions depending on the TM injection stages and on the driver lines. Some brain regions were tagged by multiple drivers, whereas others were tagged specifically only by a single driver ( Figure 2 and Table  1). From the overall external appearance, a neurogenic wave was observed to migrate from posterior regions, such as the hindbrain, to the more anterior regions according to the TM injection stages ( Figure 2). In the brain regions tagged by multiple drivers, the spatiotemporal patterns of tagging were basically similar (Table 1), indicating that all the driver lines capture a similar neurogenetic time window in neurons.
To expose the internal structures, we collected and coronally sectioned brains tagged at different TM stages by individual drivers. The interspaced serial sections were antibody stained with diaminobenzidine (DAB) for either the nls-bGAL or the membrane-localized mGFP reporter. All these sections were converted into high-resolution digital images (e.g., Figures 6B-6D) and subgrouped into datasets according to three categories, namely the driver line, TM stage, and stained reporter (nls-bGAL or mGFP). Each dataset defined by the combination of the three categories contained 142-180 interspaced section images, which were aligned along the entire anterocaudal axis of the brain (approximately 1.2 mm in length). This coverage is sufficiently high to identify even a small nucleus or structure in the brain (Table 1).
All datasets of the high-resolution section images (835 GB of data for 13,538 images in 84 datasets including TM-negative controls) and whole-mount images shown in Figure 2 are available in the NeuroGT database (https://ssbd.riken.jp/neurogt/). Users can search for these datasets from a web browser by using terms that match the meta-information stored in the database, such as their identifier, driver name, TM stage, or reporter used for staining ( Figure 3A). The search results are displayed as a list of links to the dataset page ( Figure 3B). To enable this search and visualization, development of the NeuroGT was based on the SSBD (Systems Science of Biological Dynamics) database (Tohsato et al., 2016). The dataset pages at the next stage allow users to download the high-resolution section images ( Figure 3C) and interactively view their thumbnails together with the meta-information. The thumbnail images of nls-bGALand mGFP-stained sections are stacked separately in order along the anteroposterior axis ( Figure 3D) and can be viewed sequentially by dragging the slider (Video S1). The search function for anatomical regions will be implemented in the future. Through NeuroGT, researchers can identify the drivers useful for their research and obtain them from RIKEN BDR (http:// www2.clst.riken.jp/arg/TG%20mutant%20mice%20list.html). Table 1 summarizes the brain regions that were tagged by individual drivers. When the tagged stages were compared with the reported birthdates determined in mice by using nucleotide analogs, the ranges were fairly consistent ( Table 1). The exceptions might be the pontine nucleus and dorsal root ganglion. Considering the short time lag often observed between the final S phase and the expression of CreER (Florio et al., 2012;Toma et al., 2014;; see also this paper), the TM stages for labeling these regions slightly preceded the birthdates determined with nucleotide analogs, suggesting that the loxP recombination might have occurred in pre-mitotic neuronal progenitors in these regions. Table 1 lists only limited brain regions from areas containing extensive tagged neurons. The NeuroGT database can be used to access and visualize a specific brain region of interest.
The following sections describe the features of each driver line, focusing on specific regions that can be effectively tagged by a particular line.
Neurog2 CreER (G2A) driver using the neurog2 enhancer The Neurog2 CreER (G2A) line tags the most extensive brain regions. Specifically, at TM10.5-TM12.5, sagittal stripes in the cerebellum were labeled (magenta arrowheads 1 in Figure 2A), reflecting birthdate-dependent generation of cerebellar Purkinje neurons (Hashimoto and Mikoshiba, 2003;Namba et al., 2011). TM injection between TM11.5 and TM17.5 heavily labeled olfactory bulb neurons, as reported previously , and neurons in the central olfactory areas and the amygdala (blue arrowheads 2 in Figure 2A). The dense labeling of the piriform cortex at TM11.5-TM12.5 demarcated a sharp border from the neocortex that was labeled by later TM injections, as will be described in the next paragraph (see also Figure 4C). In the inferior colliculus of the midbrain, a neurogenic gradient from the lateral to the mediocaudal was marked at TM12.5-TM16.5 (brown arrowheads 3 in Figure 2A) as previously reported (Altman and Bayer, 1981). Labeling of the inferior colliculus was also observed in the same spatiotemporal pattern by other neurogenic tagging drivers (brown arrowheads 3 in Figures 2C and 2D; Table 1).
In cross-sections of the neocortex, the inside-out birth-order arrangement of neurons was clearly visible with the nls-bGAL reporter ( Figure 4A). As expected from the neurog2 expression, only excitatory projection neurons, but not GABAergic inhibitory neurons, were tagged in the neocortex (Figures S1A-S1C). The mGFP reporter prominently labeled neuronal fibers, including axons and dendrites ( Figures 4B-4F). In a close-up of the corpus callosum ( Figure 4B), long axons projecting from the tagged neocortical neurons were visualized. Interestingly, the axons of the early-born (TM12.5-TM13.5) and late-born (TM16.5-TM17.5) cortical neurons formed dorsoventrally segregated fascicles within the callosum, presenting a feature that would have been difficult to recognize by nucleotide-based birthdating. Previous studies have reported the dorsoventral segregation of callosal axons from the medial and lateral cortical areas (Piper et al., 2009;Nishikimi et al., 2011;Zhou et al., 2013). The present axon compartments formed by the early-born and late-born callosal neurons appeared to be different from those based on the distinct cortical areas (Figures S1D-S1G), adding complexity to the organization of the corpus callosum.
Another unique feature of the neocortex was the patchy areal distribution of superficial neurons in the layer II/III tagged around   Lawson and Biscoe, 1979 the late TM17.5 stage (green arrowheads 4 in Figure 2A). Somatosensory and visual areas contained packed X-gal-labeled neurons, whereas frontal, motor, and medial cingulate areas contained less abundant labeled neurons. The biased areal distribution of these superficial neurons was also confirmed in neocortical sections ( Figures 4F and 4F 0 ). A previous nucleotide-based neuronal birthdating study only reported a neurogenetic gradient in the superficial layers (Bayer and Altman, 1991). Perhaps the discrete mosaic distribution of superficial neurons was more readily recognizable in a whole-mount cortical representation by using neurogenic tagging.
In the hippocampus, pyramidal neurons in the CA1-CA3 regions were labeled with embryonic TM injections ( Figure 5A). Interestingly, deep and superficial neurons in the CA1 layer were separately labeled by different TM injection times (Figure 5A 0 ). They appear to be the neuronal subsets in the radial axis that have recently attracted much attention (Slomianka et al., 2011;Soltesz and Losonczy, 2018). The membrane-localized mGFP reporter, on the other hand, visualized the laminar organization of the hippocampus (Forster et al., 2006), where laminar-specific afferent axons from other brain regions and local projections of hippocampal proper neurons were differentially labeled depending on the TM stage ( Figure 5B).
To characterize the timing of individual neurons that underwent TM-induced recombination in relation to their last cell cycle, we conducted a double injection of TM at the fixed TM12.5 stage and of EdU at a certain time point either before or after TM injection ( Figures 5C-5F). Although the time courses for the emergence of double-positive neurons for EdU and the nls-bGAL reporter varied across brain regions, in several brain domains such as the piriform cortex ( Figure 5C), mammillary body ( Figure 5D), and cerebellum ( Figure 5E), TM-induced recombination was most prevalent in neurons 6-12 h after the final DNA synthesis. This time course was similar to that previously observed in olfactory bulb neurons . There were, however, a few exceptions; for example, neurons in the trapezoid body were most frequently double labeled when EdU and TM were co-injected ( Figure 5F), suggesting that TM-induced recombination occurs (B) An example of a search result, which is returned as links to individual dataset pages with whole-mount brain images and selected meta-information.
(C) Meta-information of high-resolution section images and a button that allows users to download the images. (D) Interactive viewer of the image thumbnails. The images of nls-bGAL and mGFP are stacked separately. By dragging the slider, coronal sections can be viewed sequentially along the anteroposterior axis. The sync button automatically matches the section level of the two reporter images. See also Video S1.
6 Cell Reports Methods 1, 100012, July 26, 2021 Article ll in the progenitor stage before the final DNA synthesis. Nonetheless, a peak of double-positive neurons was detected in a specific time interval between EdU and TM injections, implying that the TM-induced recombination marks the neurons only within a short differentiation time window ( Figure 5F). Neurog1 CreER (G1C) driver using the neurog1 enhancer The characteristic regions tagged by the Neurog1 CreER (G1C) driver were the pontine nucleus (magenta arrowheads 1 in Figure 2B) and the superior colliculus in the midbrain (blue arrowheads 2 in Figure 2B), both of which were not labeled in wholemount brain preparations by the other three drivers (Figures  2A-2D and Table 1).
In sections of the superior colliculus, neurons in the superficial sensory layers that receive retinal axons were labeled at TM12.5-TM13.5, and neurons in the deep motor layers were labeled sparsely for a more protracted TM11.5-TM14.5 ( Figures  S2A and S2B). This spatiotemporal pattern of labeling resembles the reported pattern of neurogenesis in the superior colliculus (Edwards et al., 1986), where superficial and deep-layer neurons are generated as distinct compartments following different time courses. Scattered neurons in the deep motor layers were also tagged by the other drivers at the same TM11.5-TM14.5 stages ( Figures S2C and S2D), although they were invisible in the wholemount preparations (Figure 2). The observation that the superficial layer neurons were only tagged by the Neurog1 CreER (G1C) driver seems to be consistent with the idea that the superficial sensory layers and the deep motor layers of the superior colliculus are populated with neuronal populations of distinct origins.
Neurod1 CreER (D1B) driver using the neurod1 enhancer The Neurod1 CreER (D1B) driver tagged only several externally visible structures in the whole-mount brains; some olfactory areas and the amygdala were significantly labeled at TM10.5-TM14.5 (blue arrowheads 2 in Figure 2C). The labeled amygdala nuclei overlapped with, but were distinct from, those labeled by the Neurog2 CreER (G2A) driver (Figure 2A). Neurons in the inferior colliculus (brown arrowheads 3 in Figures 2A, 2C, and 2D) and the cochlear nucleus (green arrowheads 4 in Figures 2B-2D) were labeled following the same spatiotemporal pattern shared  (Table 1). The late labeling of the cochlear nucleus after TM15.5 appeared to correspond with the second wave of neurogenesis in the dorsal cochlear nucleus reported recently (Shepard et al., 2019).
In cross-sections of the neocortex, a tangential band of neurons was labeled at the border between the neocortex and white matter at TM11.5-TM12.5 ( Figure 6A). These cells appeared to be surviving subplate neurons in layer VIb (Friedlander and Torres-Reveron, 2009). The mGFP reporter selectively visualized their characteristic dense basal dendrites at the bottom of the cortical plate and widespread axons in layer 1 ( Figure 6B; Clancy and Cauller, 1999). Tagging at TM13.5-TM14.5 labeled only a few cortical neurons scattered in layers VI and V ( Figure 6A). These neurons were mainly positioned in the medial cortex ( Figures 6C and 6D), and their axons were observed to selectively project to the medial part of the striatum ( Figures 6C and 6D), exhibiting a unique feature reported for neurons in the medial prefrontal area (Gerfen et al., 2013;Mailly et al., 2013). Later-stage TM injections did not label neurons in the upper layers ( Figure 6A), showing a clear contrast to the inside-out pattern throughout the cortical plate induced by the Neurog2 CreER (G2A) driver ( Figure 4A).
Neurod4 CreER (D4A) driver using the neurod4 enhancer The Neurod4 CreER (D4A) driver abundantly tagged neurons in the caudal parts of the brain such as the medulla and the spinal cord (magenta arrowheads 1 in Figure 2D). The spatial pattern and time course of labeling in these caudal regions resembled those of Neurog2 CreER (G2A) (Figure 2A and Table 1), consistent with the idea that neurod4 is a downstream effector of neurog2 (Seo et al., 2007;Masserdotti et al., 2015). However, there were differences between the two drivers; for example, the neocortex and cerebellar Purkinje cells were not significantly labeled by the Neurod4 CreER (D4A) driver ( Figure 2D and Table 1).
This driver notably visualized long projecting axons in several brain regions. For example, mGFP-labeled optic axons from the retinal ganglion cells tagged at TM13.5-TM14.5 were targeted at the superficial layers of the superior colliculus ( Figure S2D). The habenular neurons in the thalamus were generated in a lateralto-medial gradient ( Figure 7A) (Angevine, 1970;Aizawa et al., 2007), whereas their axons exhibited a periphery-to-center arrangement within the fasciculus retroflexus, which is the axon tract formed by habenular neurons ( Figure 7B).
Using this driver, we also characterized the cell-cycle timing of neurons that underwent TM-induced recombination. TM injection at the fixed E14.5 stage labeled neurons in the hippocampal CA3 region ( Figure 7C), habenula ( Figure 7D), supramammillary nucleus ( Figure 7E), and inferior colliculus ( Figure 7F). The time-spaced injection of EdU indicated that neurons were most susceptible to TM-induced recombination at 6-12 h after the final DNA synthesis. Although the time course of double labeling seemed slightly delayed compared with that of Neuro-g2 CreER (G2A) (Figures 5C-5F), such a comparison might not be meaningful, as different neurons and TM stages were involved and we could not quantitatively determine the time lag in the gene expression cascade from neurog2 to neurod4.

Neurogenic tagging resource
In the past few decades, various important discoveries have been made through neuronal birthdating (Bayer and Altman, 1987;Govindan et al., 2018). The neurogenic tagging resource presented in this study resolves some of the limitations of the previous method and forges a link between this traditional histological classification and the new experimental approaches and applications. Specifically, neurogenic tagging can be easily combined with various molecular genetic tools such as optogenetics and chemogenetics (Deisseroth et al., 2006;Alexander et al., 2009), which enable the functional manipulation of birthdate-classified neuronal subsets. We hope that the unique ideas of researchers will lead to important discoveries using this resource, given that the neuronal birthdate is so fundamental to organizing the neural circuitry.
This neurogenic tagging method uses the expression timing of neuronal differentiation genes; therefore, the underlying biological principle is significantly different from that of the traditional method using nucleotide analogs. The four neuronal differentiation genes exploited in our resources were selected on the basis of their seemingly transient expression during the maturation phase of neurons as shown in a previous study (Mattar et al., 2008) and databases (Allen Developmental Mouse Brain Atlas, https://developingmouse.brain-map.org; GENSAT: Gene Expression Nervous System Atlas, http://www.gensat.org/ index.html). As expected, the neurons underwent loxP recombination 6-12 h after the final DNA synthesis in different parts of the nervous system ( Figures 5C-5E and 7C-7F), and recapitulated the patterns consistent with those seen in birthdating analyses using nucleotide analogs (Table 1). In some exceptional cases, such as the trapezoid body ( Figure 5F), TM-induced recombination seemed to occur within a short time window during the progenitor stage before the final DNA synthesis. In the pontine nucleus and the dorsal root ganglion (Table 1), the earlier tagging stages, compared with the previously determined birthdates, also suggest that TM-induced recombination occurs pre-mitotically in these regions. Thus, neurogenic tagging using neuronal differentiation genes appears to mark the timing of neuron commitment rather than cell-cycle exit . Recently, increasing evidence has indicated that neuron commitment and cell-cycle exit are dissociable processes Cell Reports Methods 1, 100012, July 26, 2021 9 Article ll OPEN ACCESS (Imayoshi et al., 2013;Hardwick and Philpott, 2014;Oberst et al., 2019). A prime example is the cortical basal progenitor, which is the fate-restricted neural progenitor that undergoes additional cell divisions but only generates neurons (Lodato and Arlotta, 2015;Hevner, 2019). Our ongoing study shows that these cortical basal progenitors are indeed tagged by one of the neurogenic tagging driver lines.
During development, the neurod1 gene is widely expressed in many nervous systems (Miyata et al., 1999;Mattar et al., 2008), and a short enhancer element of this gene is commonly used to study neuronal differentiation in vivo and in vitro (Guerrier et al., 2009). Somewhat contradictory to the general use of this gene, the Neurod1 CreER (D1B) driver tagged only a restricted subset of neurons ( Figures 2C and 6). This might be because we obtained only one mouse line that exhibited significant CreER activity after the intensive production of transgenic mice by using the neurod1 gene. Regardless, the unique characteristics of neuronal tagging by the Neurod1 CreER (D1B) driver shown in this study and NeuroGT demonstrate the usefulness of this driver line for some specific purposes.
The present collection of four neurogenic tagging drivers does not fully cover all brain regions. For example, the majority of neurons in the striatum and hypothalamus were unlabeled by any of the drivers. To complement the current collection a promising candi- date is the Ascl1 gene, which is expressed in the basal plate of the neural tube in a manner complementary to that of neuronal differentiation genes used in this study (Ma et al., 1997;Wilkinson et al., 2013). Furthermore, the mice that express CreER under the Ascl1 gene enhancer have been reported to apparently show birthdate-dependent recombination (Tg(Ascl1-cre/Esr1*)1Jejo, MGI: 3767428, Battiste et al., 2007;Ascl1 tm1.1(cre/ERT2)Jejo , MGI: 4452601, Kim et al., 2008;Sudarov et al., 2011). In the future, we hope to include detailed image data of neurons tagged by Ascl1 CreER mice in the NeuroGT database.
To date, multiple CreER-expressing mice under the control of neuronal differentiation genes have been generated by several groups (Neurog2 tm1(cre/Esr1 * )And , MGI: 2652037, Zirlinger et al., 2002;Tg(Neurog1- Aprea et al., 2014). Although these mice have been primarily used for cell lineage analyses related to the manipulated genes, they are theoretically applicable to birthdate-based classification and manipulation of neurons. Our Nerog2 CreER (G2A) driver mouse generated by BAC transgenesis has a greater CreER activity than that under the endogenous locus in Neurog2-CreER knockin mice (Zirlinger et al., 2002). However, suitable drivers differ depending on the research purpose. It is important to note that information in the NeuroGT database can be useful even for developing a research design that does not use our mouse lines.
Features highlighted by the present neurogenic tagging analyses Our original motivation for developing the neurogenic tagging method was to visualize axon projections of birthdate-classified 10 Cell Reports Methods 1, 100012, July 26, 2021 Article ll OPEN ACCESS neuron subsets in the olfactory bulb in adulthood. This method was indeed effective in revealing the birthdate-dependent olfactory axon trajectories . It has also been applied to examine birthdate-dependent compartments in the cerebellum Zhang et al., 2020). The present analyses using the multiple neurogenic tagging drivers added interesting findings, three of which are discussed below.

Corpus callosum
Within the corpus callosum, early-born (TM12.5-TM13.5) and lateborn (TM16.5-TM17.5) cortical neurons formed segregated axon bundles ( Figures 4B and S1D-S1G), consistent with the current knowledge that the callosal projections are formed by a subset of layers II/III and V neurons in the neocortex (Fame et al., 2011). The segregated construction of the callosum by pioneering and following axons has been proposed (Koester and O'Leary, 1994;Rash and Richards, 2001). Although the previous studies have considered the time lag of axon arrivals from the medial and lateral cortical areas (Piper et al., 2009;Nishikimi et al., 2011;Zhou et al., 2013), the present analysis indicates that the time lag between the early-born and late-born callosal neurons in different layers also underlies the organization of this axon bundle, which is the largest in placental mammals (Suarez et al., 2014).

Fasciculus retroflexus
Neurogenic tagging revealed a periphery-to-center axon topology within the fasciculus retroflexus ( Figures 7A and 7B). A previous axon-tracing analysis reported that the medial and lateral habenula nuclei project their axons into the core and shell of this tract, respectively (Herkenham and Nauta, 1979). To explain this organization, the present study provides a hint: late-growing axons might penetrate the center of the bundle pre-formed by early-growing axons. In general, for axon tracts positioned on the brain surface, late-growing axons are sequentially added to the pial superficial space, displacing the pre-existing axons deeper (Walsh et al., 1983;Walsh and Guillery, 1985;Inaki et al., 2004;Yamatani et al., 2004). Although little is known about the development of internally located axon tracts, new axons are found to grow in the center of pre-formed axon bundle in the mushroom body of Drosophila (Kurusu et al., 2002). Within the optic nerve, which is an isolated axon bundle, newly growing axons are positioned near the center (Walsh, 1986).

Hippocampus pyramidal neurons
Although hippocampal neurons in the single pyramidal layer have been regarded as a homogeneous population, recent studies indicate that hippocampal pyramidal neurons are in fact heterogeneous along the radial axis (Slomianka et al., 2011;Soltesz and Losonczy, 2018). The deep and superficial subsets in the layer express different genes, make different connections, and are expected to have distinct functions in different behavioral contexts (Mizuseki et al., 2011;Lee et al., 2014;Danielson et al., 2016). These subsets are assumed to differentiate at different times in an inside-out pattern (Slomianka et al., 2011;Soltesz and Losonczy, 2018). The present neurogenic tagging analysis confirmed this assumption and allowed effective classification of these neuronal subsets (Figures 5A and 5A 0 ). An advantage of this neurogenic tagging is that this classification can now be connected to functional tests. We hope that the neurogenic tagging resource will benefit ingenious ideas that will lead to future scientific discoveries.
Limitations of the study Neurogenic tagging is different from nucleotide-based neuronal birthdating. The timing of recombination in the cell cycle differs among neurons and appears to take place prior to the final mitosis in some neurons. It is important to consider this difference when the neurogenic tagging approach is taken, as this difference might be beneficial in some cases. In the NeuroGT database, the anatomical ontology of tagged brain regions is not yet available but will be included in the future to enable searching by anatomical regions.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to Tatsumi Hirata (tathirat@nig.ac.jp).

Data and code availability
All images acquired in this study are available in the NeuroGT database at https://ssbd.riken.jp/neurogt/. Raw images and unprocessed data related to this study and NeuroGT are archived in SSBD:repository (https://doi.org/10.24631/ssbd.repos.2021.03.001).  Hippenmeyer et al., 2005), with a CD-1 mixed background were provided by Dr. Silvia Arber at Friedrich Miescher Institute for Biomedical Research and backcrossed with C57BL/6 wild-type mice for at least four generations before use in this study. All mice were maintained in the animal facility of the National Institute of Genetics or RIKEN BDR. They were group housed in conventional cages under controlled conditions (temperature 23 ± 2 C, humidity 50 ± 10%, 12 h light-dark cycle), and food and water were provided ad libitum. All procedures for the care and treatment of mice were approved by the Institutional Animal Committees and carried out according to their guidelines. Heterozygous neurogenic tagging driver mice were mated with homozygous reporter mice. The day on which a vaginal plug was detected and the day of birth were designated as embryonic day 0.5 (E0.5) and postnatal day 0 (P0), respectively. TM treatment was performed by intraperitoneally injecting a staged pregnant mouse with 250 mL of corn oil (C8267, CAS#8001-30-7, Sigma-Aldrich) containing 9 mM tamoxifen (T5648, CAS#10540-29-1, Sigma-Aldrich) and 5 mM progesterone (161-14531, CAS# 57-83-0, Fujifilm Wako). As TM often delays delivery, when pups were not born by E19.5, they were collected by caesarian delivery and given to ICR foster mothers. Brain sampling was performed indiscriminately once pups grew up to the appropriate age, and only brains of the desired genotype were later selected by PCR genotyping of the reserved tissues for CreER internal sequences. At the time of sampling, the sex of the mice was still obscure and therefore undetermined; for the Nerog2 CreER (G2A) driver, it is highly likely that only male mice were sampled because the transgene seems to be located on the Y chromosome .
The Neurog1 CreER (G1C) line, officially named C57BL/6-Tg(Neurog1-cre/ERT2)G1CTahi, was generated using the genomic BAC clone RP24-347K19 encoding the mouse neurog1, which was obtained from the BACPAC Resource Center (Children's Hospital Oakland Research Institute, Oakland, CA). The entire coding sequence of exon1 was replaced by CreER(T2) (Feil et al., 1997) (gifted by Dr. Pierre Chambon), and the loxP site in the vector backbone (pTARBAC) was deleted by replacement with the zeocin resistance gene in the p24loxZeo (gifted by Dr. Junji Takeda). For the Neurod1 CreER (D1B) line, officially named C57BL/6-Tg(Neurod1-cre/ERT2) D1BTahi, the CreER(T2) coding sequence was inserted at the first ATG sequence of exon2 in the BAC clone RP23-280C11 encoding the mouse neurod1 (BACPAC Resource Center). For the Neurod4 CreER (D4A) line, officially named C57BL/6-Tg(Neurod4-cre/ERT2) D4ATahi, the CreER(T2) coding sequence was inserted at the first ATG sequence of exon2 in the BAC clone RP23-55O18 encoding the mouse neurod4 (BACPAC Resource Center), and the loxP site in the vector backbone (pBACe3.6) was deleted by replacement with the zeocin resistance gene in the p23loxZeo (gifted by Dr. Junji Takeda). Except for the above-mentioned wild-type loxP sequences, the BAC vector backbone sequences, including a few genes, were unmodified.
The constructed BAC recombinants were injected into fertilized eggs with a C57BL/6 background, and the tail genomic DNA of the resulting mice was assayed by PCR for the integration of the transgene. Transgene containment was determined by PCR using internal Cre recombinase primers 5'-TAAAGATATCTCACGTACTGACGGTG-3' and 5'-TCTCTGACCAGAGTCATCCTTAGC-3', resulting in the amplification of 300-bp fragments. In total, eleven, two, and five mice were found to have random integrations of the transgenes for neurog1, neurod1, and neurod4, respectively, including multiple BAC constructs for each gene. These mouse lines were assayed for CreER activity by crossing them with ROSA26R Cre reporter mice, officially named Gt(ROSA)26Sor tm1Sor (RRID: MGI: 1861932, Soriano, 1999). After crossing, pregnant mice were injected with TM solution at E12.5 or E14.5. Embryos were dissected from the dam at E18.5-E19.5, and their isolated brains were whole-mount stained with X-gal (5-bromo-4-chloro-3-indoyl-f3-D-galactopyranoside, 029-15043, CAS#7240-90-6, Fujifilm Wako) as will be described later. Through this screen, Neurog1 CreER (G1C), Neurod1 CreER (D1B), and Neurod4 CreER (D4A) were identified by the highest recombination rate among the transgenic mice for each gene.