Brain‐wide mapping of efferent projections of glutamatergic (Onecut3+) neurons in the lateral mouse hypothalamus

Abstract Aim This study mapped the spatiotemporal positions and connectivity of Onecut3 + neuronal populations in the developing and adult mouse brain. Methods We generated fluorescent reporter mice to chart Onecut3 + neurons for brain‐wide analysis. Moreover, we crossed Onecut3‐iCre and Mapt‐mGFP (Tau‐mGFP) mice to visualize axonal projections. A dual Cre/Flp‐dependent AAV construct in Onecut3‐iCre cross‐bred with Slc17a6‐FLPo mice was used in an intersectional strategy to map the connectivity of glutamatergic lateral hypothalamic neurons in the adult mouse. Results We first found that Onecut3 marks a hitherto undescribed Slc17a6 +/Vglut2 + neuronal cohort in the lateral hypothalamus, with the majority expressing thyrotropin‐releasing hormone. In the adult, Onecut3 +/Vglut2 + neurons of the lateral hypothalamus had both intra‐ and extrahypothalamic efferents, particularly to the septal complex and habenula, where they targeted other cohorts of Onecut3 + neurons and additionally to the neocortex and hippocampus. This arrangement suggests that intrinsic reinforcement loops could exist for Onecut3 + neurons to coordinate their activity along the brain's midline axis. Conclusion We present both a toolbox to manipulate novel subtypes of hypothalamic neurons and an anatomical arrangement by which extrahypothalamic targets can be simultaneously entrained.


| INTRODUCTION
The ever-increasing pace of single-cell RNA-seq to recognize cellular diversity in the brain promotes the association of transcription factors (TFs) as identity marks to neuronal subtypes. [1][2][3][4][5][6] While this approach has unparallele d intrinsic power, neuronal populations that are placed outside conventional locations (be these layers, nuclei, or brain areas) are often overlooked. Our recent efforts to chart hypothalamic neuronal subtypes 7,8 led us to recognize the TF family containing a single CUT domain and a distinct homeodomain (Onecut) 9 as a priority label for GABA/tyrosine hydroxylase + neurons at the hypothalamic midline. Onecut TFs cascade from Onecut1/2 expressed in progenitors to Onecut3, which labels postmitotic neurons in both the fetal and adult brains. 1,2 Nevertheless, the distribution of Onecut3 + neurons along the rostrocaudal axis of the brain remains unaccounted for. Likewise, it is unclear if non-GABA neurons could also express Onecut3, a concept that accords with the Ascl1 origin of Onecut3 + neurons 1,10 ; noting that hypothalamic Ascl1 + progenitors can generate both GABA and glutamate neurons during cascading proliferation events. 10 A major challenge in neurobiology is that the pace of the development of cellular and mouse models is inferior to that of advances in single-cell RNA-seq, as well as to spatial transcriptomics. This caveat is particularly relevant for the faithful manipulation of TFs. An efficient way to generate transgenic mice is the use of bacterial artificial chromosomes (BAC), which have large carrying capacity (hundreds of kilobases), for the efficient integration of promoter and regulatory elements of specified genes together with either transgenes (e.g., Cre recombinase, flippases [FLP]) or reporter molecules (e.g., DsRed, mCherry) into the host chromatin by homologous recombination. 11 Here, we generated Onecut3-mCherry and Onecut3-iCre mouse lines and used these to map the brain-wide distribution of Onecut3 + neurons, and their connectivity, in the developing and adult nervous systems. We then combined these transgenic tools with immunohistochemistry, in situ hybridization, and high-resolution tissue-wide imaging to reveal the temporal trajectory of how Onecut3 + neurons populate brain areas from mid-gestation on in mouse, including the septal complex, hypothalamus, habenula, midbrain, and hindbrain. While Onecut3 + nuclei form a quasi-continuum along the midline, we conspicuously found Onecut3 + neurons that populate the lateral hypothalamus. We then combined intersectional genetics (Onecut3-iCre; Slc17a6-FLPo) with an adeno-associated viral (AAV) approach 12 to map efferent projections of this Onecut3 + subgroup, many of which containing thyrotropin-releasing hormone (TRH). We found Onecut3 + neurons of the lateral hypothalamus to predominantly target Onecut3 + neurons in brain areas endowed with TRH receptors (Trhr), with their presynapses containing VGLUT2. Thus, we suggest that Onecut3 + excitatory neurons of the lateral hypothalamus are poised to coordinate the activity of both intra-and extrahypothalamic Onecut3 + neuronal pools.

| Methodological considerations
Members of the Onecut family of TFs have so far been identified as critical for the development of hepatocytes, endocrine cells of the pancreas and the gastrointestinal tract, and the spinal cord. [13][14][15][16][17][18][19][20] Nevertheless, their localization in the brain remains incomplete particularly because of the lack of genetic tools to faithfully chart the distribution of neurons expressing Onecut paralogs, and to interrogate their connectivity to infer functions. We expect that generating tools to manipulate these TFs is of value particularly since the most subordinate member of the Onecut family, Onecut3, seems to be unconventional in the temporal and subregional regulation of its expression ( Figure 1A), as single-cell RNA-seq suggested its mRNA being retained even in adulthood. 2 Therefore, we used BAC-based genetic engineering ( Figure 1B, Tables 1 and 2) to generate mouse strains that carry either fluorescence reporters (mCherry) or iCre recombinase under regulatory elements of the Onecut3 promoter and are stably integrated in the mouse genome ( Figure 1B,C). (BAC)Onecut3-mCherry mouse lines exhibited cellular fluorescence at levels that a mandatory amplification step by indirect fluorescence was not required for successful localization/mapping by laser-scanning microscopy. Nevertheless, Onecut3-iCre mice, when crossed with, e.g., the Ai14 reporter line (Onecut3-iCre::Ai14) to express tdTomato ( Figure 1D), had fluorescence intensities in excess of (BAC)Onecut3-mCherry mice. Therefore, we only recognized a structure as a genuine Onecut3 + locus if (i) it was labeled in both (BAC)Onecut3-mCherry and Onecut3-iCre::Ai14 mice, and (ii) also by indirect immunofluorescence using anti-Onecut3 antibodies that had been quality-controlled in Onecut3 −/− mice. 21 Finally, Onecut3-iCre mice allowed us to develop intersectional strategies to identify both the neurotransmitter identity and brain-wide efferent maps of Onecut3 + neurons localized at an unexpected location in the lateral hypothalamus.

| Onecut3 expression during fetal development
As conventional for many TFs that define neuronal identity, the Onecut1 and Onecut2 paralogs are expressed in hypothalamic progenitors in the rodent forebrain. 1 In contrast, we found Onecut3 in post-mitotic cells that had exited the proliferative zones of the brain, alike in the spinal cord, (Figure 2A) by mid-gestation in mouse. This finding was validated by combining mouse genetics and indirect histochemistry (Figure 2A), which ensured the authentic presence of tdTomato + somata in the telencephalon, diencephalon, mesencephalon, rhombencephalon, and spinal cord. Besides localizing neuronal somata, we made an effort to reveal the extent, specific trajectories, and body-wide targets of Onecut3 + neurons of the nervous system. To this, we have crossed Onecut3-iCre and Mapt-(Tau)-mGFP mice, the latter serving as an axonal reporter through its membrane-bound GFP expression ( Figure 2A 1 ). Whole-mount imaging, incorporating growth-associated protein 43 (GAP43) to preferentially mark neurites and their growth cones during neuritogenesis, 22,23 demonstrated that GFP + processes, likely axons, radiated through the diencephalon (Figure 2A 1 ). In addition, we located Onecut3 protein in the spinal cord, [14][15][16] most likely motorneurons and spinal ventral interneurons, 15 with spinal nerves also positive for GFP ( Figure  2A 1 ). Finally, we localized GFP to the 3rd (oculomotor), 5th (trigeminal), 10th (vagus), 11th (accessory), and 12th (hypoglossal) nerves, indicating the presence of Onecut3 mainly in cranial nerve nuclei of neural crest origin. 24 Confirming a previous report, 25 Onecut3 mRNA and protein overlapped with mCherry in retinal ganglion cells of the eye ( Figure S1A). We then used tissue clearing and 3D imaging by lightsheet microscopy to produce a refined map of mCherry + neuronal cell bodies in the forebrain at E14.5 (both sagittal and coronal views are shown; Figure 2B,C). This identified midline structures, wherein neurons primarily migrate from the wall of the 3rd ventricle: the medial septal area, hypothalamus, and midbrain. Brain areas in which the mCherry reporter overlapped with Onecut3 mRNA, protein, or both, were only accepted as positively labeled ( Figure 2D-D 3 ; Figures S1B and S2A,B). The segregation of neuronal cohorts populating the prospective lateral vs. medial septal nuclei was noted as early as E14.5 ( Figure 2D), together with the impending accumulation of Onecut3 + neurons in the lateral habenula ( Figure  2D 1 ). The hypothalamus provided a curious case because the mCherry signal, including both perikaryal and processes, formed a continuous meshwork encompassing the periventricular nucleus (PeVN), mediolateral zone (MZ), and lateral hypothalamus (LH), and extending dorsally to the zona incerta (ZI), which partitions as a nucleus of the subthalamus. 1,4 A Onecut3 + cell group was also found positioned behind the isthmic organizer in the dorsal midbrain, yet its identity was not further pursued in this study ( Figure 2D 2 ). In the hindbrain, bundles of axons were seen that reached the spinal cord ( Figure 2D 3 ).

| Onecut3 expression in adult brain
Within the early postnatal brain, the pattern of mCherry + neuronal structures recapitulated those identified at E14.5 ( Figure 3A-B 1 ; Figures S2A,B and S3A). When using the anatomical mapping tools as for embryonic brains, we found the co-existence of mCherry and Onecut3 protein in the adult lateral and medial septal nuclei ( Figure  3B 1 ,C). Within the hypothalamus, tdTomato + neurons overlapped with Onecut3 + protein and populated the anterior hypothalamic area, with their positions spread toward the wall of the 3rd ventricle, corresponding to the periventricular nucleus (PeVN; Figure 3D). Moreover, the MZ also harbored Onecut3 + neurons ( Figure 3D). In the lateral habenula (LHb), Onecut3 + neurons concentrated in its magnocellular division (Figure 3B 1 ,D 1 ). In the medial habenula, a tdTomato + cluster of neurons resided in the superior region; though we failed to detect Onecut3 protein in this area. mCherry + fiber labeling was found in the cingulum bundle, radiating toward layer 1 in the somatomotor area ( Figure 3E). In addition, we found fine-caliber fibers coursing mainly through layer 5/6 (its lamina distinguished through the density of both parvalbumin + interneurons and SMI-32 + pyramidal-like cells, Figure S3B) 26 throughout the sensory motor and visual cortices into the hippocampal CA1-CA3 subfields ( Figure S3B 1 ,B 2 ), while being excluded from the dentate gyrus ( Figure S3B 3 ). From a methodological standpoint, we find it important to emphasize that Onecut3-Cre::Ai14 mice confirmed the cellular distribution maps we have charted by using (BAC)Onecut3-mCherry mice, as shown for both the PeVN/MZ of the hypothalamus ( Figure 3D, Figure S3B 5 ), and LHb ( Figure 3D 1 ). Nevertheless, we caution that ectopic mCherry labeling (that is, no somatic co-localization with Onecut3 protein) occurred in the rostral migratory stream, olfactory bulb, and piriform cortex F I G U R E 2 Onecut3 + neurons and their projections in mouse embryos. (A) A chain of Onecut3 + territories in the brain and along the spinal cord, including mCherry + somata (arrowheads, A) and GFP + axons (arrowheads, A 1 ), were identified using Onecut3-iCre mice. Insets in (A 1 ) show the oculomotor nerve (1), corticospinal tract (2), and spinal segments (3). Sagittal (B) and coronal (C) brain maps on E14.5 used to schematically illustrate the distribution of mCherry + cell bodies (red circles) and processes (dashed red lines). Histochemistry for Onecut3 (green) overlapped with mCherry (red, arrowheads), thus confirming faithful localization of the genetic signal. (D-D 3 ). Onecut3 + /mCherry + perikarya were localized to the medial septum (D), hypothalamus (hyp) and habenula (Lhb; D 1 ) midbrain (MB; D 2 ) and hindbrain (D 3 ). III,V, X, XI, XII, cranial nerves; 3V, 3rd ventricle; CX, cortex; den, diencephalon; HB, hindbrain; Hi, hippocampus; LV, lateral ventricle; MB, midbrain; mes, mesencephalon; MZ, mediolateral zone; OC3, Onecut3; OB, olfactory bulb; rhom, rhombencephalon; SEP, septum; sp, spinal cord; VZ, ventricular zone. Scale bars = 500 μm (A, B), 200 μm (D 1 ), 100 μm (insets in A, D, D 2 , D 3 ), 40 μm (insets in D, D 1 ). in (BAC)Onecut3-mCherry mice, which are known for unwanted recombination also in some other (BAC) models (data not shown). 27 For the Onecut3-iCre::Ai14 line in particular, we found non-complementary expression between tdTomato and Onecut3 protein at some sites. These data could suggest developmentally regulated transient Onecut3 expression, with Cre-mediated recombination conferring "life-time" tracing of these cell groups or ectopic recombination events. Therefore, (BAC)Onecut3-mCherry mice are best suited for anatomical mapping studies of connectivity, while Onecut3-iCre are useful for circuit reconstruction of novel cell groups through viral transduction so long as cellular targets are confirmed by a combination of transgenic and antibody-based histochemical methods.

| Brain-wide circuit mapping using intersectional mouse genetics
Next, we used our Onecut3-iCre model to define the extent of both local and long-range projections of these poorly understood lateral hypothalamic neurons. We first used a pAAV-hSYN-DIO-mCherry-driven adeno-associated virus (AAV) approach, harboring mCherry under the neuron-specific human synapsin promoter, in the lateral hypothalamus of adult Onecut3-iCre mice (AP = −0.94, L = −0.88, DV = −5.52 mm, relative to bregma and dura as appropriate; n = 6, males and females; Figure 4A). After 3 weeks, mCherry signal was restricted to neurons in the lateral hypothalamus, indicating successful targeting ( Figure 4B,H). Histochemistry for Onecut3 showed that these cells were correctly visualized by mCherry ( Figure 4H), and had medium-to-thick processes, likely dendrites. Reminiscent to data from (BAC)Onecut3-mCherry mice, sparse fiber labeling was detected in the olfactory bulb ( Figure 4C), layers 1 and 2/3 of the somatosensory cortex ( Figure 4C 1 ), as well as the perirhinal and entorhinal areas (Figure 4C 2 ). We also found fine-caliber mCherry + processes at the outer border of the granule cell layer of the hippocampal dentate gyrus ( Figure 4D), which likely invaded through the fimbria hippocampi ( Figure  4D 1 ). Finally, we mapped long-range mCherry + fibers in the LHb (Figure 4E), the lateral and medial septal nuclei ( Figure 4F,F 1 ), as well as the tuberal region and lateral and medial preoptic area (Figure 4G), resolving a circumscribed neurocircuit with primary terminal fields dominating in brain regions that themselves harbor Onecut3 + neurons (such as the LHb and septum; Figure 3). Given the anatomical position of the ZI, we in some cases virally labeled its Onecut3 + neurons, with their efferent projections mapped separately ( Figure S4). Considering that the presence of GABA/Onecut3 + neurons 1 in the lateral hypothalamaus cannot be excluded per se, even if these cells predominate in the PeVN, we opted for an intersectional approach to strengthen our conclusions on brainwide efferents of Slc17a6/Onecut3 + neurons.
To this end, we crossed Onecut3-iCre and Slc17a6-FLPo mice (the latter expressing flippase under the Vglut2 promoter; n = 4; adult), and microinjected a dual Cre/FLPdependent pAAV-Ef1a-Con/Fon-mCherry virus, 12 expressing mCherry under the Ef1a promoter ( Figure 5A). Thus, we restricted viral labeling to neurons co-expressing Slc17a6 and Onecut3. Alike above, we found mCherry + large-caliber varicose dendrites coursing in the lateral hypothalamus and the tuberal area, often even reaching the pial surface ( Figure 5B-C 1 ). Small caliber varicose fibers, likely axons, accumulated around the 3rd ventricle, including its contralateral surface by traversing through the basal retrochiasmatic area directly underneath the 3rd ventricle ( Figure 5D), but avoiding the median eminence. Most unexpectedly, we also found mCherry + efferents in the contralateral lateral hypothalamus and tuberal area ( Figure 5E). Besides, our intersectional approach recapitulated data on terminal fields in the LHb (Figure 5F), the lateral and medial septal nuclei ( Figure 5G-G 2 ), and posterior cortical nucleus of the amygdala ( Figure S5A,A 1 ). However, these efferents were restricted unilaterally within the injected hemisphere. By using an intersectional approach, we did not recapitulate the fiber labeling in the hippocampus and cortex, indicating that these fibers are unlikely to be of Onecut3 + /Vglut2 + origin ( Figure S5B,C). Similar to the tracing above (Figure 4), we found projections to regions containing Onecut3 + neurons ( Figure 6A). Axonal projections were intermingled between Onecut3 + territories in the lateral hypothalamus/tuberal region, periventricular nucleus, and preoptic area on both sides, as well as the ipsilateral medial and lateral septum, and LHb ( Figure 6B). Immunohistochemistry confirmed the presence of VGLUT2 in varicosities throughout the projections, as well as in the vicinity of Onecut3 + neurons, suggesting direct glutamatergic innervation ( Figure 6B-E).

| DISCUSSION
The understanding of neuronal complexity and connectivity in discrete brain regions had recently received major support from the introduction of single-cell RNA-seq technologies, allowing the revised classification of neuronal subtypes based on their RNA landscapes. 1-6 Molecular profiling, particularly over successive developmental stages, predominantly relies on the spatiotemporal expression of TFs and their gene regulatory networks instructing the acquiring of distinct neurochemical make-ups. 1 Derived from recent studies of ourselves, 1,2 as well as open-label datasets, 4 the peculiar case of the Onecut3 TF was selected, because it is retained during postnatal development and even in adulthood. This feature could suggest novel functions likely related to the maintenance of synaptic neurotransmission and/or connectivity and cellular plasticity, as seen for Sox and POU TFs in the cerebral cortex. [31][32][33] Nevertheless, a caveat in existing knowledge on the precise neurocircuit established by Onecut3 + neurons precludes correct inferences on postnatal functions.
To overcome these limitations, we have produced transgenic mice that allow for either fluorescent reporters or iCre recombinase be expressed under the control of regulatory elements of the Onecut3 promoter. By using our transgenic lines, we present the detailed mapping of the Onecut3 connectome in both the fetal and adult brains. Besides our immediate aims, the combination of Onecut3-iCre mice with the viral delivery of either optogenetic 34 or DREADD constructs 35 for the temporally restricted control of neuronal activity presents powerful means to study the functionality of Onecut3 + neurons in intact systems in vivo in the future. Likewise, the manipulation of select genes (or gene sets) in Onecut3 + neurons by, e.g., Cre-dependent shRNA expression 36 can provide information pertinent to the regulation of neuronal output. Here, we have not only qualitycontrolled these new mouse models but also provided substantial novel insights in the neurochemical heterogeneity, morphology, and connectivity of a novel subset of Onecut3 + /Slc17a6 + neurons in the lateral hypothalamus, of which many are TRH + .
The tri-peptide TRH was historically isolated from the hypothalamic median eminence, where it is released into the bloodstream from release terminals of neuroendocrine cells of the paraventricular nucleus 37,38 to instruct the pituitary gland. However, TRH signaling seems equally important in many other extrahypothalamic regions, 39,40 rich in TRH receptor mRNA (Trhr) and protein. 41 For instance, TRH modifies action potential waveforms in cortical pyramidal cells and also modulates their acetylcholine-induced excitation. 42 However, previous anatomical studies were unable to elucidate the efferent projections that could deliver TRH to these forebrain regions, contrasting, for example, the well-described glutamatergic hypocretin/ orexin neurons in the same area 43,44 that gives rise to ascending projections to wide cortical and subcortical regions. 45 Considering the distribution of virally-labeled efferents, we suggest that the lateral hypothalamus can directly contribute to TRH signaling, likely modulating cortical network activity.
For Onecut3 + neurons, their GABA contingent in the PeVN has recently received significant attention. 1,23 However, the excitatory subtypes in the lateral hypothalamus has not been studied in detail. Here, we describe an alternatively produced glutamatergic Onecut3 + neuronal subtype. Major findings of our work include that (i) Slc17a6 + /Onecut3 + neurons acquire their final positions by mid-gestation in the mouse. This suggest that the wiring of their neurocircuits is an early event and might have a significant role in the activity-dependent development of their postsynaptic targets. 28 (ii) Slc17a6 + /Onecut3 + neurons of the lateral hypothalamus innervate midline structures that are endowed with other Onecut3 + neurons. (iii) The assembly of spatially segregated extrahypothalamic Onecut3 + neuronal populations, if synaptically connected, could suggest a reinforcement loop to coordinate the output of phenotypically similar neurons across fore-, mid-, and hindbrain areas. While these Onecut3 + regions do not share the same function per se, future studies on functional connectivity might be well-placed to elucidate how these brain areas are interconnected to synchronize outputs for complex behaviors. Even more so, mutual innervation of the lateral hypothalamus in the contralateral hemisphere (while other targets are strictly stereotypically mapped) suggests precise brain-wide coordination of network output. (iv) The selectivity of circuit components suggests that behavioral outputs, if any, could include reward, reactivity to adverse stimuli through fear and anxiety, [46][47][48] and circadian synchronization of TRH with synaptic output (e.g., for motivational aspects including sexual behavior) [49][50][51] through the recruitment of the LHb and the septal complex, as well as other Onecut3 − structures including the cortical amygdala complex. 52 Thus, our study identifies the efferent connectivity of Onecut3 + neurons in the lateral hypothalamus as a prototypic blueprint for the wiring of other hypothalamic neurons, with their integration into extrahypothalamic circuitries to link neuropeptide action on neurocircuits driving motivational aspects of cognition.

| MATERIALS AND METHODS
Mice were housed in a temperature-controlled environment with a 12-h/12-h dark--light cycle, and free access to food and water. All experimental procedures were planned to reduce suffering, as well as animal numbers. Transgenic lines used in this study were listed in Table 1.

| Generation of transgenic mice
Mice expressing either mCherry or iCre recombinase under the control of regulatory elements of the Onecut3 promoter were generated using a BAC-based approach. (BAC)Onecut3-mCherry and (BAC)Onecut3-iCre mice were custom designed, ES cells injected on an FVB background, and backcrossed onto the C57Bl/6J background for >5 generations before use (Table 1). BAC clones RP23-161H22 (Source BioScience) and B6Ng01-345A14 (RIKEN BioResource Research Center) were used for (BAC)Onecut3 lines. Modified BAC cassettes contained cDNA of either mCherry or iCre recombinase, Woodchuck hepatitis virus posttranscriptional regulatory elements (WPRE), human growth hormone polyadenylation hormone (hGH-PA), and a neomycin selection cassette flanked by flippase recognition sites. The transgene was fused into the translation initiation codon (ATG site) of the Onecut3 gene. The Onecut3-mCherry construct additionally included Sleeping Beauty inverted repeats that were designed to increase integration efficiency. The neomycin cassette was excised through FLP-mediated recombination, and the isolated transgenic fragment was microinjected into the pronucleus of fertilized mouse eggs (FVB.129P2-Pde6b + Tyrc-ch/AntJ, Jackson Laboratory, stock number #004828). Mice were genotyped by standard genotyping ( Table 2) to confirm transgene expression.

| Genotyping
DNA was extracted from tail (embryos) or toe clips (postnatal pups) using a standard lysis protocols with 50 mM NaOH solution (Sigma). Genotypes were subsequently analyzed with AccuStart II PCR SuperMix (Avantor/ VWR) and appropriate primer pairs (Table 2) by using a BioRad thermocycler T100.

| Tissue collection and fixation
Timed pregnancies were produced by housing a male with one or two female mice and detecting a vaginal plug the morning after intercourse (designated as embryonic [E] day 0.5). Whole embryos (up to stage E12.5), whole heads of embryos (E14.5) or dissected brains (P3) were collected, and immersion fixed with 4% paraformaldehyde (PFA) in 0.1 M PB (pH 7.4) at 4°C for 2-24 h under continuous agitation. For older postnatal and adult brains, mice were transcardially perfused with a 4% PFA in 0.1 M PB (pH 7.4) and post-fixed in the same solution overnight (at 4°C with agitation). Samples were cryoprotected in 30% sucrose in 0.1 M PB for >2 days prior to cryosectioning.

| Quantitative polymerase chain reactio
Total RNA was extracted from brain tissue samples using the Aurum Total RNA kit (BioRad), and reverse transcribed into cDNA with a High-Capacity RNA-to-cDNA Kit (Applied Biosystems). A total of 5-20 ng of cDNA was used for quantitative real-time PCR (CFX Connect, BioRad) when mixed with a SYBR Green Master Mix Kit (Life Technologies). Primers were custom designed (Primer Blast; National Center for Biotechnology) and listed in Table 3. Expression levels were normalized to TATA box-binding protein (Tbp), a housekeeping gene.

| AAV-based tracing in vivo
Adult mice were anesthetized with an i.p. injection of ketamine (90 mg/kg) and xylazine (10 mg/kg) and their heads placed in a stereotaxic frame (Kopf instruments). Mice were kept anesthetized by inhalation of N 2 O/isoflurane (0.5%-1%, 1 L/min flow rate). Each mouse received an injection of 30 μL meloxicam (2 mg/mL; given s.c.) and 50 μL enrofloxacin (25 mg/mL; given s.c.) prior to the surgical procedure. The skull was exposed, and a burr-hole craniotomy performed to make the brain at the AP = −0.94 mm, L = −0.88 mm, and DV = 5.52 mm coordinates (relative to bregma). A stereotaxic injector connected to a capillary made of borosilicate glass was used to infuse 200 nL of viral particles (see Table 5 for a complete list of virus constructs) in 1 min. After 5 min, the glass capillary was slowly withdrawn. The incision site was stitched, and the mice kept under an infrared lamp to aid their recovery, and closely monitored until they have regained consciousness. Mice were sacrificed 3 weeks later and processed for immunohistochemistry as earlier.