Pyramidal neurons (PyNs) constitute the large majority of nerve cells in the cerebral cortex and mediate all of the inter-areal processing streams and output channels1,2,4. Traditionally, PyNs have been classified into several major classes according to their laminar location and broad axon projection targets, such as intratelencephalic (IT) and extratelencephalic (ET or corticofugal), which further comprises subcerebral (including pyramidal tract; PT) and corticothalamic (CT) PyNs1. Within these classes, subsets of PyNs form specific local and long-range connectivity, linking discrete microcircuits to cortical subnetworks and output channels1,5. Single-cell transcriptome analysis suggests that there are over fifty PyN transcriptomic types6. However, genetic tools and strategies for experimentally accessing PyN subpopulations are limited.

All PyNs are generated from neural progenitors in the embryonic dorsal telencephalon, where regionally differentiated radial glial progenitors (RGs) undergo asymmetric divisions, giving rise to radial clones of PyNs that migrate to the cortex in an inside-out order7. RGs generate PyNs either directly or indirectly through intermediate progenitors (IPs), which divide symmetrically to generate pairs of PyNs8. A set of temporal patterning genes drive lineage progression in RGs, which unfold a conserved differentiation program in successively generated postmitotic neurons3,4,9. Resolving the lineage organization of diverse progenitors and their relationship to projection-defined PyN subpopulations requires fate-mapping tools with cell type and temporal resolution2.

Here we present strategies and a genetic toolkit in the mouse for targeting PyN subpopulations and progenitors guided by knowledge of their developmental programs. We leverage gene expression patterns of the cell-type specification and differentiation programs to target biologically significant progenitor subsets, PyN subpopulations and their developmental trajectories (Fig. 1a–c, Extended Data Table 1). These tools and strategies provide a roadmap for accessing hierarchically organized PyN types at progressively finer resolution. They will facilitate the tracking of developmental trajectories of PyNs for elucidating the organization and assembly of neural circuits of the cerebral hemisphere, including the cortex, hippocampus and basolateral amygdala.

Fig. 1: Strategies and drivers to target PyN types and fate-map progenitors.
figure 1

a, Major PyN projection classes mediating intratelencephalic streams (IT, red) and cortical output channels (PT, blue; CT, purple) in a sagittal brain section. Pn, pons; SC, superior colliculus; Str, striatum; Th, thalamus; Spd, spinal cord. b, PyN developmental trajectory. RGs undergo direct and indirect (IP-derived) neurogenesis, producing all laminar and projection types. The listed genes are expressed in progenitor and PyN subpopulations. SVZ, subventricular zone; VZ, ventricular zone. c, Temporal expression patterns of genes used for generating knock-in drivers across PyN development. Colours correspond to projection class; intensity gradients depict expression levels. d, E12.5 tamoxifen pulse-chase in Lhx2 embryos labelled RGsLhx2+ with a medialhigh to laterallow gradient along the dorsal neuroepithelium, ending at the cortex–hem boundary. The magnified views in d, f, j show RGs at multiple cell-cycle stages, with end-feet (arrows) and dividing soma (arrowheads) at the ventricle wall (dashed line). e, E12.5 RGsLhx2+ produced PyNs across cortical layers. f, E12.5 tamoxifen pulse-chase in Fezf2 embryos labelled RGsFezf2+ with a gradient distribution similar to that in d but at a lower density. g, RGsFezf2+ produced PyNs across layers. h, Top, distribution of RGsLhx2+ (red) and RGsFezf2+ (pink) across cortical neuroepithelium divided into medial (M), dorsal (D) and lateral (L) bins. Bottom, laminar distribution of fate-mapped PyNs. i, Fate-mapping scheme using an IS reporter with Lhx2-CreER and Fezf2-Flp: RGsLhx2+Fezf2 express tdTomato/RFP by ‘Cre-NOT-Flp’ subtraction; RGsLhx2+Fezf2+ express EGFP by ‘Cre-AND-Flp’ intersection. j, E12.5 tamoxifen 24-hour pulse-chase revealed RGsLhx2+Fezf2 and RGsLhx2+Fezf2+ throughout the cortical primordium. k, Top, the labelled number of RGsLhx2+Fezf2+ is half that of RGsLhx2+Fezf2. Bottom, the number of RGsLhx2+Fezf2 versus RGsLhx2+Fezf2+ at rostral (R), mid-level (M) and caudal (C) sections. Data in h, k are mean ± s.e.m.; see ‘Quantification and statistics related to progenitor fate-mapping’ in the Methods for statistical details. l–n, RGLhx2+Fezf2-derived PyNs (red) project to the corpus callosum (arrowheads, m) without subcortical branches; RGLhx2+Fezf2+-derived PyNs project to the thalamus (arrowheads, n) without callosal branches. DAPI (blue). Scale bars, 20 µm (d, f, j insets); 100 µm (all other panels).

Fate-mapping PyN progenitors


The transcription factors LHX2 and FEZF2 act at multiple stages throughout corticogenesis10,11,12. The fate potential of and relationship between Lhx2+ RGs (RGsLhx2) and Fezf2+ RGs (RGsFezf2) are largely unknown. We generated Lhx2-CreER, Fezf2-CreER and Fezf2-Flp driver lines and performed a series of fate-mapping experiments at multiple embryonic stages to reveal these progenitors and their lineage progression, as well as their PyN progeny in the mature cortex (Fig. 1d–h, Extended Data Figs. 1, 2).

At embryonic day (E) 10.5, a 24-hour tamoxifen pulse-chase in Lhx2-CreER;Ai14 embryos resulted in dense labelling of neuroepithelial cells and RGs in the dorsal pallium, with a sharp border at the cortex–hem boundary (Extended Data Fig. 1a). E12.5–E13.5 pulse-chase revealed a prominent medialhigh to laterallow gradient of RGsLhx2 (Fig. 1d), suggesting differentiation of the earlier RGs. E13.5–E14.5 pulse-chase showed a similar gradient pattern at a lower cell density (Extended Data Fig. 1e). Fate-mapping from E10.5–P30, E12.5–P30 and E14.5–P30 labelled PyN progeny across cortical layers (Fig. 1e, Extended Data Fig. 1b–f, p), suggesting multipotency of RGsLhx2 at these stages. During postnatal development, the expression of Lhx2 became postmitotic: pulse-chase in P5 labelled largely IT PyNs across layers and in the second postnatal week labelled more astrocytes (around 60%) than PyNs across layers (Extended Data Fig. 1q–r).

Similar fate-mapping experiments using the Fezf2-CreER driver yielded contrasting results. At E10.5, short pulse-chase labelled only a sparse set of pallial RGs, ending at the cortex–hem boundary (Extended Data Fig. 1g). E12.5–E13.5 pulse-chase labelled a larger set of RGsFezf2 with a similar medialhigh to laterallow gradient as RGsLhx2, with a notably lower density (Fig. 1f). E13.5–E14.5 pulse-chase labelled few RGs, primarily in the medial region, and otherwise postmitotic PyNs (Extended Data Fig. 1k). Fate-mapping from E10.5–P30 and E12.5–P30 labelled PyNs across cortical layers, suggesting multipotent RGsFezf2 (Fig. 1g, Extended Data Fig. 1h, j). After E13.5, the expression of Fezf2 largely shifted to postmitotic layer (L) 5 and 6 (L5/6) corticofugal PyNs (Extended Data Fig. 1l). Both the Lhx2-CreER and the Fezf2-CreER drivers recapitulate endogenous expression across developmental stages (Extended Data Fig. 2), thus providing fate-mapping tools for these progenitor pools.

To probe the relationship between RGsLhx2 and RGsFezf2, we designed an intersection–subtraction (IS) strategy. Combining Lhx2-CreER and Fezf2-Flp with an IS reporter13, we differentially labelled RGsLhx2+/Fezf2 and RGsLhx2+/Fezf2+ (Fig. 1i–n, Extended Data Fig. 1s, t). E11.5–E12.5 and E12.5–E13.5 pulse-chase revealed two distinct RG subpopulations intermixed across the dorsal pallium, with RGsLhx2+/Fezf2 more than twice as abundant as RGsLhx2+/Fezf2+. Both subpopulations distributed in a medialhigh to laterallow and rostralhigh to caudallow gradient, consistent with the patterns of Lhx2 and Fezf2 expression (Fig. 1j, k). Although most RGsFezf2 expressed LHX2, approximately 10% did not (Extended Data Fig. 2e, i), suggesting that there are three distinct RG subpopulations distinguished by differential expression of Lhx2 and Fezf2. Notably, long pulse-chase revealed that whereas RGsLhx2+/Fezf2-derived PyNs extended callosal but no subcortical axons—the IT type—RGsLhx2+/Fezf2+-derived PyNs extended subcortical but no callosal axons—the ET type (Fig. 1l–n). This result suggested fate-restricted RG lineages that produce categorically distinct PyN projection classes.

Neurogenic RGs

Early cortical progenitors comprise proliferative and neurogenic subpopulations. Tis21 (also known as Btg2) is a transcription co-regulator that is expressed in both pallium-derived glutamatergic and subpallium-derived GABAergic neurogenic RGs (nRGs)14. E10.5 fate-mapping in the Tis21-CreER driver line labelled columnar clones of PyNs and astrocytes intermixed with subpallium-derived GABAergic interneurons (Extended Data Fig. 3b, f, g). We used Tis21-CreER;Fezf2-Flp;IS mice to restrict fate-mapping to glutamatergic neurogenic RGs (Extended Data Fig. 3h–k). E11.5–E12.5 pulse-chase demonstrated that Tis21Fezf2 intersection specifically labelled a set of pallial nRGFezf2+ with enhanced green fluorescent protein (EGFP), whereas Tis21Fezf2 subtraction labelled pallial and subpallial nRGFezf2 with red fluorescent protein (RFP). Pallial nRGs consisted of both Fezf2+and Fezf2 subpopulations, suggesting heterogeneity. E12.5–P30 fate-mapping in these mice revealed three types of PyN clones (Extended Data Fig. 3i-k). RFP-only clones are likely to have derived from nRGFezf2 in which Tis21-CreER activated RFP expression; they probably consisted of PyNs that did not express Fezf2 at any stage. EGFP-only clones are likely to have derived from nRGFezf2+, in which Tis21-CreER and Fezf2-Flp co-expression activated EGFP in the IS reporter allele. Mixed clones containing both EGFP and RFP cells probably derived from nRGFezf2 in which Tis21-CreER activated RFP expression followed by postmitotic activation of EGFP through Fezf2-Flp. Together, these results indicate the presence of nRGFezf2+ and nRGFezf2, both multipotent in generating PyNs across all cortical layers.


IPs and indirect neurogenesis have evolved largely in the mammalian lineage and have further expanded in primates14,15. Along the neural tube, IP-mediated indirect neurogenesis is restricted to the telencephalon and is thought to contribute to the expansion of cell numbers and diversity in the neocortex. The majority of PyNs in mouse cortex are produced through IPs16,17, but the link between indirect neurogenesis and PyN types remains unclear. The T-box transcription factor Tbr2 (also known as Eomes) is expressed in pallial IPs throughout indirect neurogenesis18. E16.5 pulse-chase in the Tbr2-CreER driver line specifically labelled IPs (Extended Data Fig. 3a, c). E16.5 and E17.5 fate-mapping labelled PyNs in L2/3 and upper L2, respectively (Fig. 2c, d, Extended Data Fig. 3d). Therefore, the Tbr2-CreER driver enables highly restricted laminar targeting of PyN subpopulations in supragranular layers. Furthermore, Tis21-CreER and Tbr2-FlpER intersection enabled specific targeting of neurogenic but not the transit-amplifying IPs (Extended Data Fig. 3a, e). Altogether, these progenitor driver lines facilitate dissecting progenitor diversity and tracking the developmental trajectories of PyNs from their lineage origin to circuit organization.

Fig. 2: Genetic targeting of PyN subpopulations.
figure 2

aj, Driver line recombination patterns visualized through reporter expression (green; background autofluorescence, red). First row, coronal hemisections. TM, tamoxifen. ad, Second row, IT drivers targeting laminar subsets in L2–L5a of somatosensory barrel cortex (SSp-bfd), which project axons across the corpus callosum (Cc) (third row) and to the striatum (Str) (bottom row). Cux1 and Plxnd1 drivers also label subsets of medium spiny neurons in the striatum (arrowheads). E16.5 and E17.5 tamoxifen induction of Tbr2-CreER label L2/3 and L2 PyNs, respectively. eg, Second row, PT drivers label L5B PyNs, which project to numerous subcortical targets, including the thalamus (Thal) (third row) and spinal cord (corticospinal tract; CST) (bottom row). hj, Second row, CT drivers label L6 PyNs, sending axons mainly to different nuclei in the thalamus (third row). Tamoxifen induction times are indicated in the first row. The reporter allele was Ai14, except for Plxnd1-CreER (Snap25-LSL-EGFP) and Foxp2 (systemic injection of AAV9-CAG-DIO-EGFP). White matter (Wm). Scale bars, 100 µm (bottom (CST) panel in g); 1 mm (hemisection in j, which applies to the entire row); 200 µm (all other scale bars). Cell bodies are indicated by arrowheads and axons by arrows.

Targeting PyN subpopulations

We generated driver lines targeting PyN subpopulations and characterized these in comparison to existing lines where feasible (Fig. 2, Extended Data Table 1, Extended Data Fig. 4, Supplementary Tables 1, 2). These tamoxifen-inducible drivers confer temporal control and dose-dependent labelling and manipulation from individual cells to dense populations.

IT drivers

IT PyNs constitute the largest top-level class and mediate intracortical and corticostriatal communication streams1,19,20,21. Cux1 and Cux2 are predominantly expressed in supragranular IT PyNs and their progenitors22,23. In our Cux1-CreER;Ai14 mice, postnatal tamoxifen induction prominently labelled L2–L4 PyNs dorsal to the rhinal fissure as well as a set of hippocampal PyNs, recapitulating the endogenous pattern (Fig. 2a, Extended Data Fig. 4, Supplementary Video 1). Anterograde tracing revealed that PyNsCux1 in somatosensory barrel cortex (SSp-bfd or SSp) projected predominantly to the ipsi- and contralateral cortex, with only very minor branches in the striatum (Fig. 3a, d, Extended Data Fig. 7, Supplementary Video 2). Compared with existing IT drivers (Extended Data Figs. 4, 5, Supplementary Tables 1, 2), Cux1-CreER is unique in targeting predominantly cortex- but not striatum-projecting IT subpopulations.

Fig. 3: Projection patterns of PyN subpopulations in SSp-bfd cortex.
figure 3

ac, Images at SSp-bfd injection site (inj site) (first row, arrowhead) and selected subcortical projection targets for eight driver lines: EGFP expression from Cre-activated viral vector (green) and background autofluorescence (red). Tamoxifen induction time points are indicated below each gene name. Arrows indicate axons. a, IT drivers project to the cortex and striatum. Note the near absence of projection to the striatum for the Cux1 driver. b, PT drivers project to many corticofugal targets, including the brainstem and spinal cord. c, CT drivers project predominantly to the thalamus. df, Schematics of main projection targets for each PyN subset generated in this study. d, IT drivers. e, PT drivers. f, CT drivers. Ipsilateral secondary motor area (iMOs); contralateral primary somatosensory area (cSSp); secondary somatosensory area (SSs); auditory areas (AUD); visual areas (VIS); retrosplenial area (RSP); temporal association areas (TEa); ectorhinal area (ECT); reticular nucleus of the thalamus (RT); ventral anterior–lateral complex of the thalamus (VAL); ventral posteromedial complex of the thalamus (VPM); ventral posterior–lateral complex of the thalamus (VPL); submedial nucleus of the thalamus (SMT); posterior complex of the thalamus (PO); substantia nigra, reticular part (SNr); superior colliculus (SC); caudoputamen (CP); mediodorsal nucleus of the thalamus (MD); paracentral nucleus (PCN); central medial nucleus of the thalamus (CM); parafascicular nucleus (PF); globus pallidus, external segment (GPe); lateral dorsal nucleus of the thalamus (LD); central lateral nucleus of the thalamus (CL); anterior pretectal nucleus (APN); ventral medial nucleus of the thalamus (VM); zona incerta (ZI); midbrain reticular nucleus (MRN); periaqueductal gray (PAG); pontine gray (PG); gigantocellular reticular nucleus (GRN); tegmental reticular nucleus (TRN); medullary reticular nucleus (MDRN); intermediate reticular nucleus (IRN); and parvicellular reticular nucleus (PARN).  Scale bars, 1 mm (first row, in c); 200 µm (second to eighth rows, in c (for each respective row); 100 µm (CST panel in c, which applies to the bottom row). Asterisks in b, c, e indicate passing fibres.

The supragranular layers comprise diverse IT types20, but only a few L2/3 drivers have been reported so far24 and none distinguish L2 versus L3 PyNs. We used a lineage and birth dating approach to dissect L2/3 PyNs. In our Tbr2-CreER driver targeting IPs, tamoxifen induction at E16.5 and E17.5 specifically labelled PyNs in L2/3 and L2, respectively (Fig. 2c, d). Combined with the CreER to Flp conversion strategy that converts lineage and birth timing signals to permanent Flp expression13, this approach enables adeno-associated virus (AAV) manipulation of L2 and L3 IT neurons.

The plexin D1–semaphorin 3E receptor–ligand system has been implicated in axon guidance and synapse specification25,26. In developing and mature cortex, Plxnd1 (encoding plexin D1) is expressed in large sets of IT PyNs10. Plxnd1-CreER and Plxnd1-Flp driver lines recapitulated endogenous expression and labelled projection neurons in the cerebral cortex, hippocampus, amygdala and striatum (Figs. 2b, 4e, Extended Data Figs. 4, 5a, Supplementary Tables 3, 4, Supplementary Videos 3, 4); in the neocortex, L5A and L2/3 IT PyNsPlxnd1 were labelled (Fig. 2b, Extended Data Fig. 6a–c). As Plxnd1 is also expressed in vascular cells, we bred Plxnd1-CreER mice with the neuron-specific reporter Snap25-LSL-EGFP to selectively label Plxnd1+PyNs (PyNsPlxnd1; Fig. 2b). Anterograde tracing from SSp-bfd revealed that PyNsPlxnd1 project to ipsi- and contralateral cortical and striatal regions (Fig. 3a, d, Extended Data Fig. 7a–c, f, Supplementary Tables 57, Supplementary Video 5). Thus, Plxnd1 drivers confer access to this major IT subpopulation and to Plxnd1+ subpopulations in the striatum and amygdala.

PT drivers

After early expression in a subset of dorsal pallial progenitors, Fezf2 becomes restricted to postmitotic L5/6 corticofugal PyNs, with higher levels in L5B PT neurons and lower levels in a subset of CT neurons10,27. At postnatal stages, Fezf2 drivers labelled projection neurons in the cerebral cortex, hippocampus, amygdala, and olfactory bulb (Extended Data Fig. 4, Supplementary Table 4, Supplementary Video 6). Within the neocortex, PyNsFezf2 reside predominantly in L5B and to a lesser extent in L6 (Fig. 2e, 1, Extended Data Figs. 4a, d, 6d–f); PyNsFezf2 are absent below the rhinal fissure (Supplementary Table 4, Supplementary Video 6). Anterograde tracing of PyNsFezf2 in SSp-bfd revealed projections to numerous somatomotor cortical (for example, ipsilateral vibrissal secondary motor area (MOs)) and subcortical regions including the striatum, thalamic ventral posteromedial nucleus of the thalamus (VPM) and posterior complex of the thalamus (PO), anterior pretectal nucleus, ipsilateral superior colliculus (iSC), pontine nucleus, corticospinal tract (CST) and contralateral spinal trigeminal nucleus (cSp5) (Fig. 3b, e, Extended Data Figs. 7a, b, d, f, 8f–r, Supplementary Tables 5, 6, Supplementary Video 7).

We further generated several lines targeting finer PT subpopulations. In Adcyap1- and Tcerg1l-CreER drivers, late embryonic induction labelled L5B subpopulations that project only to ipsilateral cortical targets and to a subset of targets innervated by PyNFezf2 (Figs. 2f, g, 3b, e, Extended Data Figs. 4, 6d–f, 7a, b, d, f, Supplementary Tables 46, Supplementary Videos 811). In Sema3e-CreER (ref. 28), postnatal induction labelled a subset of L5B PyNs that project to more-restricted subcortical areas, namely higher-order thalamic nucleus POm and pontine nucleus (Extended Data Figs. 4d, 8a–e, Supplementary Tables 4, 5, Supplementary Video 12). Together, the new set of PT drivers will enable a finer hierarchical dissection of molecularly and anatomically defined PT types.

CT drivers

Tbr1 is expressed in postmitotic L6 CT neurons and represses the expression of Fezf2 and Ctip2 (also known as Bcl11b) to suppress the PT fate3,29. In Tbr1-CreER;Ai14 mice, tamoxifen induction at P4 marked L6 CT neurons densely, with sparse labelling in L2/3, cerebral nuclei, hippocampus, piriform cortex and amygdala (Fig. 2h, Extended Data Fig. 4a–c, Supplementary Table 4, Supplementary Video 13). PyNsTbr1 from SSp-bfd projected to multiple thalamic targets, including primary and higher order nuclei, as well as reticular nucleus of the thalamus (Fig. 3c, f, Supplementary Tables 5, 6, Supplementary Video 14). Consistent with a study showing that some PyNsTbr1 project to the contralateral cortex30, we found labelling in the corpus callosum (Extended Data Fig. 6). It remains to be determined whether PyNsTbr1 with contralateral projections (Fig. 3f) represent a distinct type.

Tle4 is a transcription corepressor that is expressed in a subset of CT PyNs31,32. Our Tle4-CreER driver specifically labelled L6 CT PyNs across the cortex (Fig. 2i, Extended Data Figs. 4, 6d–f, Supplementary Table 4, Supplementary Video 15). Tle4 is also expressed in medium spiny neurons of the striatum, olfactory bulb, hypothalamus, iSC, cerebellum and septum (Extended Data Fig. 4c, Supplementary Table 4, Supplementary Video 15). PyNsTle4 in SSp-bfd specifically projected to first-order thalamic VPM and reticular nucleus of the thalamus (Fig. 3c, f, Extended Data Fig. 7a, b, e, f, Supplementary Table 4, Supplementary Video 16).

Foxp2 is expressed in many CT neurons from the postmitotic stage to the mature cortex33,34,35. In adult Foxp2-IRES-Cre mice36, systemic injection of Cre-dependent AAV9-DIO-GFP specifically labelled L6 PyNs; Foxp2+ cells were also found in the striatum, thalamus, hypothalamus, midbrain, cerebellum and inferior olive (Fig. 2j, Extended Data Fig. 4a–c, Supplementary Table 4, Supplementary Video 17). PyNFoxp2 in SSp-bfd projected to thalamus, tectum and some ipsilateral cortical areas (Fig. 3c, f, Supplementary Tables 4, 5, Supplementary Video 18). Compared to PyNsTle4, PyNsFoxp2 projected more broadly to the thalamus, largely overlapping with PyNTbr1 axons.

To further characterize several PyN driver lines, we performed a set of histochemical analyses (Extended Data Fig. 6). PyNs targeted in Fezf2, Tcerg1l and Adcyap1 drivers extensively co-labelled with PT markers. PyNs targeted in Tle4 and Tbr1 drivers co-labelled with CT markers. The laminar patterns and class-specific marker expression in these driver lines precisely recapitulated endogenous patterns (in situ hybridization data in the Allen Brain Map: Mouse Brain Atlas;, providing further evidence of the reliability and specificity of these driver lines.

Combinatorial targeting of projection types

To further dissect driver-line-defined subpopulations according to projection targets, we first used retrograde tracing. Within the PT population, retroAAV and fluorogold injections in the spinal cord of Fezf2-CreER mice specifically labelled L5B corticospinal PyNs in the sensorimotor cortex (Extend Data Fig. 10a–c, Supplementary Table 6). To explore PyNsFezf2 subpopulations jointly defined by projection targets and sublaminar position, we used the IS reporter13. Consistent with previous findings37, PyNsFezf2 that project to the thalamus and medulla resided in the upper and lower sublamina of L5B in the primary motor area (MOp), respectively (Extended Data Fig. 9d–f). In SSp-bfd, PyNsFezf2 with collaterals to the striatum resided in upper L5, those with collaterals to the superior colliculus or cSp5 resided in the middle and lower portion of L5B, and those projecting to thalamic POm resided both in middle to lower L5B and in L6 (Extended Data Fig. 9g–h, l–o). We then distinguished subsets of L5B PyNsFezf2 according to their expression of the calcium-binding protein parvalbumin using Fezf2-CreER;Pv-Flp;IS mice that differentially labelled PyNsFezf2+/PV and PyNsFezf2+/PV+ (in which PV represents parvalbumin; this gene is also known as Pvalb) (Extended Data Fig. 9i, j). Compared to PyNsFezf2+/PV, PyNsFezf2+/PV+ exhibited more depolarized resting membrane potentials. In addition, we designed a strategy (triple trigger) to target PyNsFezf2 jointly defined by a driver line, a projection target and a cortical location (Extended Data Fig. 10).

We also used retroAAV to dissect the CT and IT populations. In Tle4-CreER;IS mice, retrograde tracing from the thalamic VPM revealed two subpopulations of L6 PyNsTle4, one extending apical dendrites to the L4/5 border, the other to L1 (Extended Data Fig. 9q), suggesting differential inputs. In Plxnd1-CreER mice (Extended Data Fig. 9p, r–w), whereas L5A PyNsPlxnd1 projected to both the ipsi- and the contralateral striatum, L2/3 PyNsPlxnd1 projected mostly to the ipsilateral striatum.

In addition, consistent with the finding that some PyNsFezf2 extend contralateral cortical and striatal projections (Fig. 3e), retrograde cholera toxin subunit B (CTB) tracing from the striatum labelled a set of contralateral PyNsFezf2+ at the L5A–L5B border (Extended Data Fig. 11a–e), a characteristic IT feature. Indeed, a small set of PyNs at the L5A–L5B border co-expressed Fezf2 and Plxnd1 mRNAs; these PyNsFezf2/Plxnd1 occupied the very top sublayer of the PyNFezf2 population (Extended Data Fig. 11f–h), and thus probably contributed to their contralateral cortical and striatal projections (Fig. 3a, b, d, e). Single-cell reconstruction may reveal whether PyNsFezf2/Plxnd1 are typical IT cells or also project subcortically and represent an ‘intermediate PT-IT’ type.

Finally, we show highly specific targeting of PyN subtypes by combining their developmental, molecular and anatomical attributes. PyNsPlxnd1 localize to L5A, L3 and L2 and project to numerous ipsilateral and contralateral cortical and striatal targets (Figs. 2b, 3a, d, Extended Data Figs. 4, 7). We developed a method (Fig. 4a) to dissect PyNPlxnd1 subtypes on the basis of the developmental principle that PyN birth order correlates with laminar position and the observation that the majority of IT PyNs are generated from IPs17. In Plxnd1-Flp;Tbr2-CreER;Ai65 mice, the constitutive Plxnd1-Flp allele marks the whole population (Fig. 4e) and the inducible Tbr2-CreER allele enables birth dating (Fig. 4a). Notably, tamoxifen induction at E13.5, 15.5 and 17.5 selectively labelled L5Aand progressively more superficial PyNsPlxnd1 (Fig. 4b–d). We then bred Plxnd1-Flp;Tbr2-CreER;dual-tTA mice for anterograde tracing of projection patterns (Fig. 4f). AAV-TRE3g-mRuby injection into SSp-bfd in E13.5- and 17.5-induced mice labelled distinct subtypes of PyNsPlxnd1 with different projection patterns. E13.5-born PyNsPlxnd1(E13.5) resided in L5A and projected ipsilaterally to multiple cortical areas, contralaterally to homotypic SSp-bfd cortex and heterotypic cortical areas, and bilaterally to the striatum (Fig. 4g–m, u). By contrast, E17.5-born PyNsPlxnd1(E17.5) resided in L2; although they also extended strong projections to ipsilateral cortical and striatal targets and to the homotypic contralateral cortex, they had minimal projections to the heterotypic contralateral cortex and striatum (Fig. 4n–u). These birth-dated PyNPlxnd1 subsets further differed in their axon termination patterns within a cortical target area. Whereas PyNPlxnd1(E13.5) axons terminated throughout the thickness of L1 and L2/3, with few axon branches in L5A, PyNPlxnd1(E17.5) axons terminated strongly in L2/3 and L5A, with few branches in L1 (Fig. 4l–m, s–u). Thus, even within the same target areas, birth-dated PyNsPlxnd1 may preferentially select different postsynaptic cell types and/or subcellular compartments.

Fig. 4: Combinatorial targeting of IT subtypes by lineage, birth time and anatomy.
figure 4

a, Strategy for combinatorial labelling of PyNPlxnd1 subtypes. In a Tbr2-CreER;Plxnd1-Flp;Ai65 mouse, tamoxifen inductions at successive embryonic times label deep or more superficial PyNPlxnd1 subsets born sequentially from intermediate progenitors. bd, Laminar subsets born at E13.5 (b), E15.5 (c) and E17.5 (d). e, The overall population is labelled in a Plxnd1-Flp;R26-FSF-tdTom mouse for comparison. The bottom panels in be show high-magnification images of the boxed regions in the top panels. f, In Tbr2-CreER;Plxnd1-Flp;dual-tTA mice, E13.5 or E17.5 tamoxifen inductions activate tTA expression in L5A or L2 PyNsPlxnd1, respectively, and AAV-TRE3g-memb-mRuby2 anterograde injection in SSp-bfd reveals the projection pattern of each laminar subset. g, n, Coronal sections display the injection site and several major projection targets. gm, Anterograde tracing from E13.5-born L5A PyNsPlxnd1 in SSp-bfd, with images (hm) of projection targets in several ipsi- and contralateral regions. nt, Anterograde tracing from E17.5-born L2 PyNsPlxnd1 in SSp-bfd, with images (os) of projection targets in several ipsi- and contralateral regions. The bottom panels in l, m, s and t show high-magnification images of the boxed regions in the top panels. The higher magnification of cSSp (l, s; bottom) and iMOs (m, t; bottom) display laminar axon termination differences between L5A and L2 PyNsPlxnd1. u, Schematics comparing E13.5-born L5A (left) and E17.5-born L2 (middle) PyNPlxnd1 projection patterns; note differences in the strength of several contralateral targets and in the laminar pattern of axon termination (right). Arrowheads indicate cell body positions; arrows indicate axons. Contralateral temporal association area (cTEa); ipsilateral striatum (iStr); contralateral striatum (cStr); white matter (Wm). Scale bars: 1 mm (be (top panels), g, n); 200 µm (be (bottom panels)); 50 µm (ht (including top panels in l, m, s, t)); 5 µm (l, m, s, t (bottom panels)).


Together with previous resources24,38,39, the PyN driver lines we present here provide much improved specificity, coverage and robustness for a systematic dissection of PyN organization from broad subclasses to finer types. By focusing on driver lines that recapitulate the expression of key transcription factors and effector genes that are implicated in specification and differentiation, these tools will enable the dissection and fate-mapping of biologically significant subpopulations of PyNs through their inherent developmental, anatomical and physiological properties; that is, ‘carving nature at its joints’. The precision and reliability of these drivers also allows the combinatorial targeting of finer projection types through the intersection of molecular, developmental and anatomical properties. The inducibility of driver lines enhances the specificity and flexibility of cell targeting, manipulation and fate-mapping. Inducibility also allows control over the density of labelling and manipulation, from dense coverage to single-cell analysis—the ultimate resolution for clarifying the stereotypical and variable features of neurons within marker-defined subpopulations40,41,42. Temporal control allows gene manipulations at different developmental stages to discover the cellular and molecular mechanisms of circuit development and function. Together, these tools and strategies establish a roadmap for dissecting the hierarchical organization of PyN types on the basis of their inherent biology. The incorporation of recently developed enhancer AAVs43 with these driver lines may further increase the specificity, ease and throughput of cell-type access.

Several transcription factors used in this study (for example, Cux1, Fezf2, Tbr1, Tbr2 and Foxp2) continue to evolve and diverge in primates44 and are implicated in developmental disorders such as autism23,35. Our transcription factor driver lines provide handles to track the developmental trajectories of PyN subpopulations in cortical circuit assembly, with implications in the cross-species evolution of PyNs and for deciphering the genetic architecture of neurodevelopmental disorders.


Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

Generation of knock-in mouse lines

Driver and reporter mouse lines listed in Supplementary Table 1 were generated using a PCR-based cloning, as described before and below13,45. All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Cold Spring Harbor Laboratory (CSHL) in accordance with NIH guidelines. Mouse knock-in driver lines are deposited at The Jackson Laboratory for wide distribution. Knock-in mouse lines were generated by inserting a 2A-CreER or 2A-Flp cassette in-frame before the STOP codon of the targeted gene. Targeting vectors were generated using a PCR-based cloning approach as described before. In brief, for each gene of interest, two partially overlapping BAC clones from the RPCI-23&24 library (made from C57BL/b mice) were chosen from the Mouse Genome Browser. The 5′ and 3′ homology arms were PCR-amplified (2–5 kb upstream and downstream, respectively) using the BAC DNA as template and cloned into a building vector to flank the 2A-CreERT2 or 2A-Flp expressing cassette as described47. These targeting vectors were purified and tested for integrity by enzyme restriction and PCR sequencing. Linearized targeting vectors were electroporated into a 129SVj/B6 hybrid embryonic stem (ES) cell line (v6.5). ES clones were first screened by PCR and then confirmed by Southern blotting using appropriate probes. DIG-labelled Southern probes were generated by PCR, subcloned and tested on wild-type genomic DNA to verify that they give clear and expected results. Positive v6.5 ES cell clones were used for tetraploid complementation to obtain male heterozygous mice following standard procedures. The F0 males were bred with reporter lines (Supplementary Tables 1, 3, 4) and induced with tamoxifen at the appropriate ages to characterize the resulting genetically targeted recombination patterns.

Tamoxifen induction

Tamoxifen (T5648, Sigma) was prepared by dissolving in corn oil (20 mg ml−1) and applying a sonication pulse for 60 s, followed by constant rotation overnight at 37 °C. Embryonic inductions for most knock-in lines were done in the Swiss-Webster background; inductions for Tis21-CreER, Fezf2-Flp intersection experiments were done in the C57BL6 background. E0.5 was established as noon on the day of vaginal plug and tamoxifen was administered to pregnant mothers by gavage at a dose varying from 2–100 mg kg−1 at the appropriate age. For embryonic collection (12–24 h pulse-chase experiments), a dose of 2mg kg−1 was administered to pregnant dams via oral gavage. For postnatal induction, a 100–200 mg kg−1 dose was administered by intraperitoneal injection at the appropriate age.


Postnatal and adult mice were anaesthetized (using Avertin) and intracardially perfused with saline followed by 4% paraformaldehyde (PFA) in 0.1 M PB. After overnight post-fixation at 4 °C, brains were rinsed three times and sectioned at a 50–75-µm thickness with a Leica 1000s vibratome. Embryonic brains were collected in PBS and fixed in 4% PFA for 4 h at room temperature, rinsed three times with PBS, dehydrated in 30% sucrose-PBS, frozen in OCT compound and cut by cryostat (Leica, CM3050S) in 20–50-µm coronal sections. Early postnatal pups were anaesthetized using cold shock on ice and intracardially perfused with 4% PFA in PBS. Post-fixation was performed similarly to older mice. Postnatal mice aged 1–2 months were anaesthetized using Avertin and intracardially perfused with saline followed by 4% PFA in PBS; brains were post-fixed in 4% PFA overnight at 4 °C and subsequently rinsed three times, embedded in 3% agarose-PBS and cut to a 50–100-μm thickness using a vibrating microtome (Leica, VT100S). Sections were placed in blocking solution containing 10% normal goat serum (NGS) and 0.1% Triton-X100 in PBS1X for 1 h, then incubated overnight at 4 °C with primary antibodies diluted in blocking solution. Sections were rinsed three times in PBS and incubated for 1 h at room temperature with corresponding secondary antibodies (1:500, Life Technologies). Sections were washed three times with PBS and incubated with DAPI for 5 min (1:5,000 in PBS, Life Technologies, 33342) to stain nuclei. Sections were dry-mounted on slides using Vectashield (Vector Labs, H1000) or Fluoromount (Sigma, F4680) mounting medium.

To perform molecular characterization of GeneX-CreER mouse lines, we stained 40-µm vibratome sections for CUX1 and CTIP2, that were imaged in a Nikon Eclipse 90i fluorescence microscope. Focusing on the somatosensory cortex, we counted tdTomato+ cells in a column of around 300-µm width and determined their relative position along the dorso-ventral axis that goes from the ventricular surface (0) to the pia (100%). As a reference, CTIP2+ and CUX1+ regions were plotted as green and blue bars, where the upper limits correspond to the mean relative position of the dorsal-most positive cells, and the lower limits correspond to the mean relative position of the ventral-most positive cells. Grey areas in histograms correspond to the s.d. of those limits. The frequency of tdTomato+ cells along the dorso-ventral axis was plotted in a histogram with a bin width of 5%. Number of cells: Fezf2-CreER, 2,781 cells; Tcerg1l-CreER, 185 cells; Adcyap1-CreER, 54 cells; Tle4-CreER, 2,737 cells; Lhx2-CreER, 1,380 cells; Plexind1-CreER, 809 cells; Cux1-CreER, 2,296 cells; Tbr1-CreER, 3,572 cells; Tbr2-CreER tamoxifen E16.5, 1,273 cells; Tbr2-CreER tamoxifen E17.5, 1,871 cells. For each line we quantified at least four sections from two embryos. Differences in cell numbers are due to differences in labelling density.

For colocalization determination, we obtained confocal z-stacks centred in layer 5 or 6 of the somatosensory cortex, of 320 × 320 × 40 µm3 volumes. For all tdTomato+ cells in the volume, we manually determined whether they were also positive for the desired markers by looking in individual z-planes. The percentage of positive cells was calculated for each area. Average number of tdTomato+ cells quantified per staining: Fezf2-CreER, 314 cells in layer 5 and 472 in layer 6; Tcerg1l-CreER, 162 cells; Adcyap-CreER, 20 cells; Tle4-CreER, 157 cells in layer 5 and 1,081 in layer 6; Lhx2-CreER, 294 cells; Plexind1-CreER, 468 cells in layers 4 and 5a; Cux1-CreER, 761 cells; Tbr1-CreER, 858 cells; Tbr2-CreER, 1,380 cells. For each line we quantified at least four sections from at least two embryos. Differences in cell numbers are due to differences in labelling density.


Anti-GFP (1:1,000, Aves, GFP-1020), anti-RFP (1:1,000, Rockland Pharmaceuticals, 600-401-379), anti-mCherry (1:500, OriGene AB0081-500), anti-mKATE2 for Brainbow 3.0 (a gift from D. Cai), anti-SATB2 (1:20, Abcam ab51502), anti-CTIP2 (1:100, Abcam 18465), anti-CUX1 (1:100, SantaCruz 13024), anti-LDB2 (1:200, Proteintech 118731-AP), anti-FOG2 (1:500, SantaCruz m-247), anti-LHX2 (1:250, Millipore-Sigma ABE1402) and anti-TLE4 (1:300, Santa Cruz sc-365406) were used.

Validation of PyN driver lines

ViewRNA tissue Assay (Thermo Fisher Scientific) fluorescent in situ hybridization (FISH) was carried out as per the manufacturer’s instructions on genetically identified PyNs expressing H2bGFP nuclear reporter (GeneX-CreER;LSL-H2bGFP) to validate the expression of PyN mRNA within Cre-recombinase dependent H2bGFP expressing cells in adult tissue (p24). Antibody validation with Cre-recombinase dependent reporter (GeneX-CreER;Ai14) was also used as it was available for use in adult tissue. For both FISH and antibody validation experiments, the percentage of total recombinase-dependent reporter-positive cells co-expressing PyN driver transcript or antibody was quantified.

Viral injection and analysis

Stereotaxic viral injection

Adult mice were anaesthetized by inhalation of 2% isofluorane delivered with a constant air flow (0.4 l min−1). Ketoprofen (5 mg kg−1) and dexamethasone (0.5 mg kg−1) were administered subcutaneously as preemptive analgesia and to prevent brain oedema, respectively, before surgery, and lidocaine (2–4 mg kg−1) was applied intra-incisionally. Mice were mounted in a stereotaxic headframe (Kopf Instruments, 940 series or Leica Biosystems, Angle Two). Stereotactic coordinates were identified (Supplementary Table 5). An incision was made over the scalp, a small burr hole drilled in the skull and brain surface exposed. Injections were performed according to the strategies delineated in Supplementary Table 5. A pulled glass pipette tip of 20–30 μm containing the viral suspension was lowered into the brain; a 300–400 nl volume was delivered at a rate of 30 nl min−1 using a Picospritzer (General Valve Corp); the pipette remained in place for 10 min preventing backflow, prior to retraction, after which the incision was closed with 5/0 nylon suture thread (Ethilon Nylon Suture, Ethicon) or Tissueglue (3M Vetbond), and mice were kept warm on a heating pad until complete recovery.

Systemic AAV injection

Foxp2-IRES-Cre mice were injected through the lateral tail vein at 4 weeks of age with a 100 µl total volume of AAV9-CAG-DIO-EGFP (UNC Viral Core) diluted in PBS (5 × 1011 vg per mouse). Three weeks after injection, mice were transcardially perfused with 0.9% saline, followed by ice-cold 4% PFA in PBS, and processed for serial two-photon (STP) tomography.


AAVs serotype 8, 9, DJ PHP.eB or rAAV2-retro (retroAAV) packaged by commercial vector core facilities (UNC Vector Core, ETH Zurich, Biohippo, Penn, Addgene) were used as listed in Supplementary Table 5. In brief, for cell-type-specific anterograde tracing, we used either Cre- or Flp-dependent or tTA-activated AAVs combined with the appropriate reporter mouse lines28 (Supplementary Table 7), or dual-tTA (Fig. 4 and Extended Data Fig. 10) to express EGFP, EYFP or mRuby2 in labelled axons. retroAAV-Flp was used to infect axons at their terminals46 in target brain structures to label PyNs retrogradely according to the experiments detailed in Supplementary Table 5.

Microscopy and image analysis

Imaging was performed using Zeiss LSM 780 or 710 confocal microscopes, Nikon Eclipse 90i or Zeiss Axioimager M2 fluorescence microscopes, or whole-brain STP tomography (detailed below). Imaging from serially mounted sections was performed on a Zeiss LSM 780 or 710 confocal microscope (CSHL St. Giles Advanced Microscopy Center) and Nikon Eclipse 90i fluorescence microscope, using objectives ×63 and ×5 for embryonic tissue, and ×20 for adult tissue, as well as ×5 on a Zeiss Axioimager M2 System equipped with MBF Neurolucida Software (MBF). Quantification and image analysis was performed using Image J/FIJI software. Statistics and plotting of graphs were done using GraphPad Prism 7 and Microsoft Excel 2010.

Twenty-four-hour pulse-chase embryonic experiments

For 24-hour pulse-chase embryonic experiments (Fig. 1, Extended Data Fig. 1), high-magnification insets are not maximum intensity projections. To observe the morphology of RGs, only a few sections from the z-plane in low-magnification images have been projected in the high-magnification images.

Quantification and statistics related to progenitor fate-mapping

Quantification for top panels in Fig. 1h, Extended Data Fig. 2m–o (n = 5–6 from two litters): mean values, number of progenitors ± s.e.m. *P < 0.05 (compared with bin M, RGsLhx2+), #P < 0.05 (compared with bin M, RGsFezf2+), one-way ANOVA, Tukey’s post-hoc test. P < 0.05 (compared with RGsLhx2+ for corresponding bins), unpaired Student’s t-test. Quantification for bottom panels in Fig. 1h, Extended Data Fig. 2m–o (n = 3 from two litters): mean values for percentage total PyNs (S1)  ± s.e.m. *P < 0.05 (compared with PyNsLhx2+). Quantification for Fig. 1k, Extended Data Fig 2t: top panel: (n = 3 from two litters): mean values, number of progenitors ± s.e.m. *P < 0.0001 (compared with RGsLhx2+Fezf2), unpaired Student’s t-test. Bottom panel (n = 3 from two litters): mean values, number of progenitors ± s.e.m. *P < 0.05 (compared with rostral RGsLhx2+Fezf2), #P < 0.05 (compared with rostral RGsLhx2+Fezf2+), one-way ANOVA, Tukey’s post-hoc test. P < 0.05 (compared with RGsLhx2+/Fezf2 for corresponding regions), unpaired Student’s t-test.

Target-specific cell depth measurement

Cell depth analysis for retrogradely labelled projection-specific genetically identified PyNs (GeneX-CreER) were obtained using 5× MBF fluorescent widefield images of 65-µm thick coronal sections in MO and SSp-bfd. MO cell depths are presented in micrometres owing to the absence of a defined white matter border in frontal cortical areas and SSp-bfd depth ratio measurements were normalized to the distance from pia to white matter. For each condition we quantified at least four sections taken from two mice.

Whole-brain STP tomography and image analysis

Perfused and post-fixed brains from adult mice were embedded in oxidized agarose and imaged with TissueCyte 1000 (Tissuevision) as described48,49. We used the whole-brain STP tomography pipeline previously described48,49. Perfused and post-fixed brains from adult mice, prepared as described above, were embedded in 4% oxidized-agarose in 0.05 M PB, cross-linked in 0.2% sodium borohydrate solution (in 0.05 M sodium borate buffer, pH 9.0–9.5).The entire brain was imaged in coronal sections with a 20× Olympus XLUMPLFLN20XW lens (NA 1.0) on a TissueCyte 1000 (Tissuevision) with a Chameleon Ultrafast-2 Ti:Sapphire laser (Coherent). EGFP/EYFP or tdTomato signals were excited at 910 nm or 920 nm, respectively. Whole-brain image sets were acquired as series of 12 (x) × 16 (y) tiles with 1 μm × 1 μm sampling for 230–270 z sections with a 50-μm z-step size. Images were collected by two PMTs (PMT, Hamamatsu, R3896), for signal and autofluorescent background, using a 560-nm dichroic mirror (Chroma, T560LPXR) and band-pass filters (Semrock FF01-680/SP-25). The image tiles were corrected to remove illumination artifacts along the edges and stitched as a grid sequence47,49. Image processing was completed using ImageJ/FIJI and Adobe/Photoshop software with linear level and nonlinear curve adjustments applied only to entire images.

Cell body detection from whole-brain STP data

PyN somata were automatically detected from cell-type specific reporter lines (R26-LSL-GFP or Ai14) by a convolutional network trained as described previously48. Detected PyN soma coordinates were overlaid on a mask for cortical depth, as described48.

Axon detection from whole-brain STP data

For axon projection mapping, PyN axon signal based on cell-type-specific viral expression of EGFP or EYFP was filtered by applying a square root transformation, histogram matching to the original image, and median and Gaussian filtering using Fiji/ImageJ software50 so as to maximize signal detection while minimizing background auto-fluorescence, as described before51. A normalized subtraction of the autofluorescent background channel was applied and the resulting thresholded images were converted to binary maps. Three-dimensional rendering was performed on the basis of binarized axon projections and surfaces were determined based on the binary images using Imaris software (Bitplane). Projections were quantified as the fraction of pixels in each brain structure relative to each whole projection.

Axon projection cartoon diagrams from whole-brain STP data

To generate cartoons of axon projections for a given driver line, axon detection outputs from all individual experiments were compared (sorting the values from high to low), and analysed side-by-side with low-resolution image stacks (and the CCFv3 registered to the low-resolution dataset for brain area definition) to get a general picture of the injection, as well as high-resolution images for specific brain areas.

Registration of whole-brain STP image datasets

Registration of brain-wide datasets to the Allen reference Common Coordinate Framework (CCF) version 3 was performed by 3D affine registration followed by a 3D B-spline registration using Elastix software52, according to established parameters52. For cortical depth and axon projection analysis, we registered the CCFv3 to each dataset so as to report cells detected and pixels from axon segmentation in each brain structure without warping the imaging channel.

In vitro electrophysiology

Brain slice preparation

Mice (>P30) were anaesthetized with isoflurane, decapitated, brains dissected out and rapidly immersed in ice-cold, oxygenated, artificial cerebrospinal fluid (section ACSF: 110 mM choline-Cl, 2.5 mM KCl, 4 mM MgSO4, 1 mM CaCl2, 1.25 mM NaH2PO4, 26 mM NaHCO3, 11 mM d-glucose, 10 mM Na ascorbate, 3.1 Na pyruvate, pH 7.35, 300 mOsm) for 1 min. Coronal cortical slices containing somatomotor cortex were sectioned at a 300-µm thickness using a vibratome (HM 650 V; Microm) at 1–2 °C and incubated with oxygenated ACSF (working ACSF; 124 mM NaCl, 2.5 mM KCl, 2 mM MgSO4, 2 mM CaCl2, 1.25 mM NaH2PO4, 26 mM NaHCO3, 11 mM d-glucose, pH 7.35, 300 mOsm) at 34 °C for 30 min, and subsequently transferred to ACSF at room temperature (25 °C) for more than 30 min before use. Whole-cell patch recordings were directed to the somatosensory and motor cortex, and the subcortical whiter matter and corpus callosum served as primary landmarks according to the atlas (Paxinos and Watson Mouse Brain in Stereotaxic Coordinates, 3rd edition).

Patch-clamp recording in brain slices

Patch pipettes were pulled from borosilicate glass capillaries with filament (1.2 mm outer diameter and 0.69 mm inner diameter; Warner Instruments) with a resistance of 3–6 MΩ. The pipette recording solution consisted of 130 mM potassium gluconate, 15 mM KCl, 10 mM sodium phosphocreatine, 10 mM HEPES, 4 mM ATP·Mg, 0.3 mM GTP and 0.3 mM EGTA (pH 7.3 adjusted with KOH, 300 mOsm). Dual or triple whole-cell recordings from tdTomato+ and EGFP+ PyNs were made with Axopatch 700B amplifiers (Molecular Devices) using an upright microscope (Olympus, BX51) equipped with infrared-differential interference contrast optics (IR-DIC) and a fluorescence excitation source. Both IR-DIC and fluorescence images were captured with a digital camera (Microfire, Optronics). All recordings were performed at 33–34 °C with the chamber perfused with oxygenated working ACSF.

Recordings were made with two MultiClamp 700B amplifiers (Molecular Devices). The membrane potential was maintained at −75 mV in the voltage clamping mode and zero holding current in the current clamping mode, without the correction of junction potential. Signals were recorded and filtered at 2 kHz, digitalized at 20 kHz (DIGIDATA 1322A, Molecular Devices) and further analysed using the pClamp 10.3 software (Molecular Devices) for intrinsic properties.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.