Layer-restricted Brn3b expression in the dorsal midbrain

Our initial screening for SC neuronal markers in the superficial SC showed that Brn3b is expressed at the bottom layer of the superficial SC [15]. The organization of superficial SC can be defined relative to the axonal projections of retinal ganglion cells (RGCs). To visualize RGC axons, cholera toxin B subunit (CTB)-conjugated fluorescent dyes were delivered into 1 eye (Fig 1A–1C). CTB labeling divides superficial SC into 2 layers: stratum griseum superficiale (SGS), a densely labeled top layer, and stratum opticum (SO), a weakly labeled bottom layer. CTB labeling verified that Brn3b+ neurons, identified by immunostaining are localized in SO (Fig 1D). Furthermore, Brn3b is expressed sparsely in the intermediate SC and abundantly in the deep SC/PAG.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Layer-restricted Brn3b expression in the dorsal midbrain.

(A, B) Schematic diagrams of the midbrain after CTB injection into the contralateral eye (A); detailed layer distribution in the coronal section of the dorsal midbrain (top) and a sagittal image showing the location (dotted line) of the imaged coronal section (bottom) (B). (C, D) Brn3b expression (green), visualized by anti-Brn3b antibody is detected in the sSC, iSC, dSC, and PAG at P11 (C). Magnified view of the boxed area (D) shows Brn3b expression confined to the bottom layer of the sSC (SO). CTB labeling (red) divides the superficial SC into the SGS and SO. The aqueduct (Aq) is also shown. DAPI (blue). (E) Schematic diagram of the conditional Brn3b allele. Conditional deletion (by En1-Cre) removes an open reading frame of Brn3b and places a human placental AP under the control of the Brn3b promoter. (F) Dorsal midbrain section from Brn3b^flox/+:: En1-Cre mouse histochemically processed for AP signals at P24. Brackets represent the layer segregation. (G-K) Axonal projections from the dorsal midbrain to adPBG and Pn (G), CnF (H), central lateral nucleus of the thalamus (CL) (I), lateral posterior nucleus (LP) and hypothalamus (HY) (J), vLGN and PRC (K). Each area was identified by its anatomical position (see Methods). (L) Schematic diagram of the midbrain Brn3b+ neuronal projections. Scale bars: 250 μm. adPBG, adjacent parabigeminal nucleus; AP, alkaline phosphatase; CnF, cuneiform nucleus; CTB, cholera toxin B; dSC, deep SC; iSC, intermediate SC; PAG, periaqueductal gray; Pn, pons; PRC, precommisural nucleus; SC, superior colliculus; SGS, stratum griseum superficiale; SO, stratum opticum; sSC, superficial SC; vLGN, ventral lateral geniculate nucleus.

https://doi.org/10.1371/journal.pbio.3002386.g001

Neurons in specific layers of the dorsal midbrain selectively innervate different subcortical areas [19,20]. Restricted expression of Brn3b, particularly in the bottom layer of superficial SC (SO), prompted us to examine axonal projections of Brn3b+ neurons to other brain areas. To trace Brn3b+ axonal projections, we used a conditional Brn3b mouse carrying the Brn3b gene flanked by loxP sites and a human placental alkaline phosphatase (AP) coding region inserted immediately downstream of 3′ loxP (Brn3b^flox/+, Fig 1E) [21]. Following Cre-mediated recombination, a copy of Brn3b is deleted and AP is expressed under the control of the Brn3b promoter. To visualize Brn3b+ neurons, we crossed Brn3b^flox/+ to the En1-Cre line, in which Cre expression is confined to the midbrain [22,23]. The layer-restricted distribution of AP staining in the SC matched the labeling from the Brn3b antibody (S1A–S1G Fig): both AP and immunostaining signals in the superficial SC were detected close to the pia at P1 when SGS and SO are barely distinguishable, and then became restricted to the SO at P10/P12 when superficial SC development was completed [24]. Both methods revealed abundant labeling of somas/nuclei in the deep SC/PAG at P1. These observations confirmed that AP signals report Brn3b expression.

We next examined axonal projections to other brain areas, based on AP signals, and found that Brn3b+ neurons clearly innervate several subcortical areas, including LP, ventral lateral geniculate nucleus (vLGN), precommisural nucleus, hypothalamus, central lateral nucleus of the thalamus, adjacent parabigeminal nucleus, pons, and cuneiform nucleus (Fig 1F–1L). To rule out the possibility that observed axonal projections originate from areas outside the midbrain, we first conducted immunostaining of the thalamus including LP, and found no Brn3b staining, excluding the contribution of local thalamic neurons to the detected AP signals (S1H and S1I Fig). Considering the broad expression of Brn3b in the retina, we also examined Brn3b^flox/+:: En1-Cre mice to ensure that labeling was not present in the retina’s output neurons, RGCs. Indeed, no AP signals were detected in the retina, confirming the midbrain-restricted Cre expression in the En1-Cre mouse (S1J Fig). Together, these results demonstrate that the observed axonal labeling represents Brn3b+ neuronal projections from the dorsal midbrain.

Next, we examined whether Brn3b+ neurons are glutamatergic or GABAergic, considering that SC projection neurons comprise both populations [9,25]. We initially attempted to mark glutamatergic and GABAergic neurons by either immunostaining or crossing Slc17a6-Cre (VGluT2-ires-Cre) or Gad2-Cre to a Cre-dependent td-Tomato reporter line (Ai14). However, very dense immunolabeling of synaptic terminals and overly abundant filling of neuronal processes by td-Tomato signals precluded visualizing individual somas. Instead, we performed in situ hybridization using Slc17a6 and Gad1 probes. The brain tissue for the hybridization was obtained from a newly generated Brn3b^GFP/+ mouse, in which an open reading frame of Brn3b was replaced with GFP by CRISPR-based genome editing (S2A–S2C Fig). First, we confirmed that GFP signals faithfully represent Brn3b expression by double immunostaining, which showed that approximately 95% of Brn3b+ cells express GFP and approximately 93% of GFP+ cells express Brn3b. Subsequent double labeling with anti-GFP combined with in situ probes to either Slc17a6 or Gad1 revealed that all Brn3b+ neurons are glutamatergic and not GABAergic (S2D–S2L Fig).

Brn3b is required for survival of neurons projecting to LP

To assess the role of Brn3b in circuit assembly of the midbrain, we crossed Brn3b^flox/flox to Brn3b^+/-:: En1-Cre mice to generate Brn3b^flox/-:: En1-Cre progeny (Brn3b cKO) and confirmed Brn3b loss in midbrain by immunostaining (S3A–S3E Fig). The conditional approach avoids confounding effects of retinal degeneration and, consequently, the compromised development of superficial SC in global Brn3b mutants [17,26]. Therefore, we proceeded to analyze and compare the phenotypes of control animals carrying conditional Brn3b and WT alleles (Brn3b^flox/+:: En1-Cre) and mutants carrying conditional and null Brn3b alleles (Brn3b^flox/-:: En1-Cre, cKO); in both cases, mice carry 1 copy of En1-Cre and 1 copy of Brn3b^flox, producing AP signals.

The AP signals faithfully represent Brn3b expression in both control and cKO mice. These signals disappeared in the superficial SC of the mutants, with no obvious changes in other layers. Thus, Brn3b deletion induces a selective loss of Brn3b+ neurons in the superficial SC (Fig 2A). We also examined projections of Brn3b+ neurons to other brain areas and found that axonal projections to LP and ventral LGN were decreased in cKO mice, most significantly in the rostral LP and ventral LGN (Fig 2B–2D). Quantification of AP signals in the LP revealed an approximately 48% decrease. Axonal projections to other subcortical areas exhibited no obvious differences (Fig 2B and 2C and Fig 2E–2G).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Brn3b deletion reduces the number of Brn3b+ neurons in the superficial SC and decreases projections to LP.

Brn3b+ neurons visualized by AP signals. (A) (Left) Midbrain sections from control and conditional Brn3b knockout (cKO) at P24. (Right) Magnified view of the boxed areas (i, ii) showing the loss of Brn3b+ neurons in the superficial SC (arrow) of the cKOs. Brackets represent the layer distribution. (B–D) Decreased projections to the LP and vLGN and no difference in projections to the hypothalamus (HY) and PRC in cKO brains. Magnified view of the boxed images (iii, vi) showing the LP areas used for quantification (D; n = 5 mice/group). The AP-labeled LP area, normalized to the average control value, was reduced in cKO brains. Unpaired two-tailed Student’s t test (mean ± SEM, p < 0.0001 [****]). (E–G) No obvious differences in axonal projections to the central lateral nucleus of the thalamus (CL), adPBG, Pn, and CnF between control and cKO at P24 (n = 5 mice/group). Scale bars: 250 μm. The data underlying this figure can be found in S1 Data. adPBG, adjacent parabigeminal nucleus; AP, alkaline phosphatase; CnF, cuneiform nucleus; LP, lateral posterior nucleus; Pn, pons; PRC, precommisural nucleus; SC, superior colliculus; vLGN, ventral lateral geniculate nucleus.

https://doi.org/10.1371/journal.pbio.3002386.g002

The cKO animals lacked Brn3b+ neurons at the bottom layer of the superficial SC (SO), where LP-projecting neurons localize, which apparently explains the altered projections to LP. However, because some axonal projections to LP were still detectable in these cKO mice, we considered the possibility that Brn3b+ neurons in other layers of SC may also innervate LP. To test this idea, we first evaluated whether superficial SC Brn3b+ neurons project to LP using a second Cre line (S4A–S4D Fig). We crossed Brn3b^flox/+ to Ntsr1-GN209-Cre mice, which express Cre in the LP-projecting superficial SC neurons [27,28], but we unexpectedly found AP signals also in the retina. However, projections of Brn3b+ neurons to LP persisted after enucleating both eyes, suggesting that labeled axons in LP originated from the SC and not from the retina. We also quantified the number of Brn3b+ neurons in the offspring of Ntsr1-GN209-Cre crossed to the Ai14 line expressing Cre-dependent td-Tomato. We found that approximately 23% of td-Tomato+ cells were Brn3b+. Interestingly, Brn3b mutants (Brn3b^flox/-:: Ntsr1-GN209-Cre) showed no changes in the survival of the superficial SC Brn3b+ neurons and projections to the LP, likely because Cre expression occurs too late during development in this Cre line (see Discussion). Therefore, for further analysis, we utilized En1-Cre-based mutants only.

To confirm that Brn3b+ neurons in superficial SC project axons to LP, we labeled these neurons using 2 constructs delivered by adeno-associated virus (AAV): GFP-dependent FLP recombinase (FLP-DOG) and FLP-dependent mCherry (fDIO-mCherry). The FLP-DOG (DOG: Dependent On GFP) is genetically engineered to be unstable and degrade, whereas binding to GFP prevents its degradation [29]. Inside GFP+ cells, FLP-DOG becomes stabilized and converts fDIO-mCherry to an active configuration, allowing mCherry expression. We injected AAV-FLP-DOG and AAV-fDIO-mCherry into the superficial SC of Brn3b^GFP/+ mouse to selectively label Brn3b+ neurons (S4E–S4I Fig). Double immunostaining revealed that approximately 98% of mCherry+ cells express GFP, suggesting high fidelity of the FLP-DOG method. Subsequent analysis revealed axons labeled with mCherry in LP, confirming that Brn3b+ neurons in superficial SC indeed innervate LP and that they are not local neurons. Lastly, we injected AAV-mCherry into the deep SC/PAG of wild-type mice and found axonal labeling in LP (S5A–S5D Fig). Together, our results indicate that LP is innervated by neurons of the superficial SC as well as deep SC/PAG and that Brn3b deletion by En1-Cre causes a loss of the superficial SC Brn3b+ neurons specifically, resulting in the decreased projection to the LP.

In addition to affecting the SC-LP connections, Brn3b deletion by En1-Cre completely eliminated innervation of the ventral LGN (Fig 2B and 2C). Given that the superficial SC Brn3b+ neurons project only to the LP and that delivering AAV-mCherry into the deep SC/PAG did not show projections to the ventral LGN, we speculated that the ventral LGN-projecting Brn3b+ neurons are located in the intermediate SC layer. To test this possibility, we delivered AAV-Cre into the intermediate SC of Brn3b^flox/+ mice, which successfully labeled Brn3b+ neurons in the intermediate SC and confirmed their axonal projections to the ventral LGN (S5E–S5G Fig).

Brn3b deletion-induced cell death occurs around birth

We next analyzed development of Brn3b+ neurons to define specific time points of neuronal loss and altered projections to LP. During development, Brn3b expression in the dorsal midbrain becomes detectable as early as E13 (Fig 3A and 3B). Therefore, we examined the possibility that Brn3b could regulate early development including neurogenesis. However, we found that Brn3b expression is barely detectable in the layer defined by a marker of proliferating cells, Ki67, suggesting that Brn3b is expressed post-mitotically and regulates later aspects of neuronal development and differentiation. We subsequently examined control mouse brains at later developmental time points, E16, E18, P1, and P4 (Fig 3C–3E). Based on the AP signals, segregation of Brn3b+ neurons into the superficial SC was first detectable at E18 and became obvious at P1. In cKO mice, the loss of superficial SC Brn3b+ neurons was already evident at E18 and persisted until adult stages (Fig 2A). It is known that Brn3b deletion causes apoptotic death of RGCs during retinal development [17]. To examine whether Brn3b deletion likewise causes apoptotic death of Brn3b+ neurons in the superficial SC, we performed immunostaining with anti-cleaved caspase-3 (Fig 3F–3J). Indeed, Brn3b deletion increased the number of cleaved caspase-3+ cells at E18 (approximately 59% increase in cKO) and this labeling was mainly detected in superficial and intermediate SC, suggesting that Brn3b expression is necessary for neuronal survival in these layers.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Loss of the superficial SC neurons in Brn3b mutants occurs around birth.

(A) Schematic diagram illustrating a specific localization of the imaged sagittal section according to a lateral distance from the midsagittal line at E13. (B) (Left) Example of sagittal section. (Right) Magnified view of the boxed area. Brn3b+ neurons (red) were missing in the layer labeled by Ki67, a maker of proliferating cells (cyan) at E13 (n = 3 mice). (C) Schematic diagrams showing the brain area (boxed) of a coronal image used for analysis (top) and a sagittal image depicting the level (dashed line) where each coronal section was obtained (bottom) at E16-E18 (i) and at P1-P5 (ii). (D, E) Brn3b+ neurons, visualized by AP signals, were weakly detectable at E18 and become obvious at P1 and P4 (arrows) in the superficial SC of control (D). However, these neurons were missing at E18, P1 and P4 in the superficial SC of cKO (E) (n = 3 mice/group/development stage). (F) Schematic diagram depicting the brain areas (boxed) analyzed to examine the number of cleaved caspase-3+ cells in the sagittal section. The localization of imaged sections was shown as a lateral distance from the midsagittal line. (G, H) Cleaved caspase-3 immunoreactivity (red) revealed increased cell death at E18 in cKO (H), compared to control (G). DAPI (blue). (I, J) Quantification revealed increased cell death in cKO (35.7 ± 4.9/mm² for control, 56.6 ± 3.3/mm² for cKO; n = 5 mice/group), primarily occurring in the superficial and intermediate layers; unpaired two-tailed Student’s t test (mean ± SEM, p = 0.008 [**] in I and p = 0.03 [*] and 0.02 [*] in J). Brackets represent the layer segregation and dashed lines indicate the pial surface. Scale bars: 250 μm (B) and 150 μm (D–E and G–H). The data underlying this figure can be found in S1 Data. AP, alkaline phosphatase.

https://doi.org/10.1371/journal.pbio.3002386.g003

The lack of AP signals in the superficial SC of Brn3b mutants indicates the death of Brn3b+ neurons. However, extensive AP signals in other layers made it difficult to distinguish somas from processes, preventing visualization of individual Brn3b+ cells. To assess possible changes of Brn3b+ neurons in other layers, we utilized the molecular markers ETV1 and Brn3a. ETV1 is expressed in Brn3b+ neurons occupying the intermediate SC layer (S6A–S6G Fig). Double immunostaining showed that approximately 50% of Brn3b+ neurons were ETV1+ and approximately 49% of ETV1+ neurons were Brn3b+ at P2. Quantification of ETV1+ neurons at E16 and P2 revealed no difference between control and cKO mice, suggesting that Brn3b deletion has no effect on the development of Brn3b+/ETV1+ cells. We also examined Brn3b+ neurons expressing Brn3a (S6H–S6M Fig). Double immunostaining showed that approximately 60% of Brn3b+ neurons were Brn3a+ and approximately 45% of Brn3a+ neurons were Brn3b+ at P1 and that Brn3b+/Brn3a+ neurons are mainly located in deep SC/PAG. Quantification of Brn3a+ neurons at E15 and P1 also revealed no difference in their number and distribution in control and cKO mice, indicating that Brn3b deletion has no substantial effects on the development of Brn3b+/Brn3a+ neurons.

Considering that ETV1 and Brn3a are expressed in only about half of Brn3b+ neurons in their respective areas of co-expression (intermediate SC and deep SC/PAG), we also examined survival of Brn3b+ neurons using the Brn3b^GFP/+ mouse, in which GFP expression faithfully reports Brn3b expression (S2 Fig). We crossed Brn3b^flox/flox to Brn3b^GFP/+:: En1-Cre mice and confirmed Brn3b loss in Brn3b^{GFP/ flox}:: En1-Cre offspring (Brn3b-GFP cKO) by immunostaining (S7 Fig). Quantification of GFP+ signals at E16 revealed no difference between Brn3b^GFP/+ (control) and Brn3b-GFP cKO. Also, no changes in GFP+ signals were detected in deep SC/PAG at P1/P2. However, we found approximately 30% decrease of GFP+ cell number in superficial and intermediate SC, confirming that Brn3b deletion affects neuronal survival in these layers specifically. Collectively, these results suggest that the loss of Brn3b has no effect on neuronal survival in the layers of dorsal midbrain outside the superficial and intermediate SC, including the deep SC/PAG.

We also examined axonal development during several embryonic and postnatal stages. We observed no obvious differences between control and Brn3b mutants at E13, suggesting that Brn3b deletion had no effects on initial axonal growth (S8 Fig). Consistent with our observations at P24, no obvious changes of AP signals were detected in other brain areas at E13, E16, and P1/P2. Projections to LP became visible at P4/P5 in control and cKO mice (S9 Fig), indicating that Brn3b deletion causes no significant delay in axonal growth. However, despite seemingly unaffected developmental timing, projections to LP were clearly decreased in the mutants (approximately 80% decrease at P4/P5, approximately 60% at P8/P9, approximately 55% at P12/P13). Altogether, our findings suggest that Brn3b deletion in the midbrain causes decreased projections to the LP due to the loss of superficial SC Brn3b+ neurons occurring around birth.

Brn3b regulates expression of Tac2 in the deep SC/PAG

The deep SC/PAG highly expresses Brn3b but did not show obvious morphological alterations in Brn3b mutants. Considering the diverse roles of Brn3b in retinal development [21,30,31], we speculated that apart from specific structural changes, Brn3b loss could cause various transcriptional alterations, including dysregulation of the genes involved in visually triggered behavioral responses. To test this hypothesis, we conducted transcriptomic profiling of the dorsal midbrain and identified 1,992 genes that were significantly differentially expressed in Brn3b cKOs compared to controls (padj < 0.05; Fig 4A and 4B and S1–S3 Tables). We discovered that besides Brn3b itself, neuropeptide tachykinin 2 (Tac2) showed the highest level of down-regulation in the Brn3b mutants. Interestingly, Tac2 was previously implicated in fear learning [32,33]. Additionally, Tac2 expression was up-regulated by single housing-based social isolation, which caused enhancement of freezing responses to a visually threatening looming stimulus [34]. Therefore, we hypothesized that decreased Tac2 expression due to the loss of Brn3b could affect neural circuits mediating fear-related behavior.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Brn3b regulates expression of Tac2 in the deep SC/PAG.

(A, B) Transcriptomic profiling (RNA-seq) using dorsal midbrain at P2-4 (n = 3 for control and 4 for cKO). (A) Volcano plot of the RNA-seq results. Brn3b and Tac2 were down-regulated in cKO (arrows). The vertical dashed lines indicate the lowest log2-fold change value of significantly down- or up-regulated genes. The horizontal dashed line indicates adjusted p value (p_adj) at 0.05. (B) Violin plots showing TPM values of Brn3b and Tac2 for each animal. The log2-fold change was 2.01 for Brn3b and 3.10 for Tac2. (C) Assessment of Tac1, Tac2, and Crh levels at P3-P5 showed decreased expression of Tac2 in cKO mice. The data for transcripts were normalized to the average control value. Quantification (Tac1: 1.01 ± 0.07 for control, 1.22 ± 0.12 for cKO; Tac2: 1.07 ± 0.16 for control, 0.54 ± 0.12 for cKO; Crh: 1.02 ± 0.11 for control, 1.37 ± 0.18 for cKO; Brn3b: 1.05 ± 0.17 for control, 0.01 ± 0.00 for cKO; n = 5 mice/group/gene). Unpaired two-tailed Student’s t test (mean ± SEM, p = 0.177 for Tac1, p = 0.030 [*] for Tac2, p = 0.136 for Crh and p = 0.0003 [***] for Brn3b). (D) Schematic diagram of the brain area analyzed for Tac2 expression in a coronal section (top) and a sagittal image depicting the level (dashed line) where the coronal section was acquired (bottom). (E, F) Brn3b deletion decreased Tac2 expression (red) in the deep SC/PAG, as detected by in situ hybridization. DAPI (blue). (G) Quantification of Tac2+ cell number (76.7 ± 6.5/mm² for control, 48.2 ± 3.3/mm² for cKO; n = 3 mice/group). Unpaired two-tailed Student’s t test (mean ± SEM; p = 0.017 [*]). Scale bars: 250 μm. The data underlying this figure can be found in S1 Data and S1–S3 Tables. PAG, periaqueductal gray; SC, superior colliculus; TPM, transcripts per million.

https://doi.org/10.1371/journal.pbio.3002386.g004

To further assess whether Tac2 may act downstream of Brn3b, we performed reverse transcription-quantitative PCR (RT-qPCR) on independent biological samples and confirmed that Brn3b deletion indeed significantly reduces Tac2 expression (approximately 50%, Fig 4C). Two related neuropeptides, tachykinin1 (Tac1) and corticotropin-releasing hormone (Crh), known to mediate effects of stress [35,36], showed no difference between controls and mutants. To determine whether Brn3b+ neurons express Tac2, we performed in situ hybridization with Tac2 probes on brain tissues of the Brn3b^GFP/+ mouse (S10 Fig). Tac2+ cells were mainly located in the deep SC/PAG and approximately 57% of Tac2+ cells were Brn3b+. No Tac2+ cells were found in the superficial and intermediate SC. Furthermore, Brn3b deletion reduced Tac2+ cell number (approximately 37% decrease, Fig 4D–4G), consistent with the finding that Brn3b regulates Tac2 expression. Unlike the superficial SC, altered Tac2 expression in the deep SC/PAG of the mutants very unlikely resulted from neuronal death, as neither an obvious loss of AP signals (Fig 2A) nor a change in the number of Brn3a+ cells and GFP+ cells was detected in that area (S6 and S7 Figs).

Brn3b-dependent circuits are required for behavioral responses to visual threat

Upon discovering that Brn3b loss disrupts SC-LP connections and down-regulates Tac2 expression, we proceeded to evaluate whether Brn3b deletion affects visually triggered fear responses using the looming stimulus assay (Fig 5A). This assay examines the mouse’s reaction to a dark, overhead expanding disk (a looming stimulus) that mimics an approaching predator (see Methods) [18]. The stimulus causes mice to either freeze or escape to a shelter. Here, we omitted the shelter to focus on freezing responses that are thought to depend on the SC-LP pathway [9,10,14]. The behavioral examination was performed using group-housed as well as single-housed mice to investigate potential enhancement of the freezing responses precipitated by social isolation and linked to Tac2 activity [34]. Basic visual abilities were examined using the light/dark exploration test [37,38].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Brn3b-dependent circuits are necessary for visually triggered freezing responses.

Fear-related defensive behavior was analyzed by examining freezing responses to visual threat in a looming stimulus assay. (A) Schematic diagram of the behavioral paradigms. The looming stimulus was an expanding dark disk on a gray background presented overhead (0.25 s expansion; 0.25 s at largest disk size; 0.5 s interstimulus interval; 10 repeats). Following the looming stimulus assay, the mice were tested on light/dark discrimination. For the experiments involving single housing, the mice were housed individually for 2 weeks prior to the behavioral tests. (B) Quantification of freezing responses for group-housed (32.4 ± 6.5 s for control, 22.6 ± 6.1 s for cKO; n = 9 for control and 8 for cKO) and single-housed (78.1 ± 12.3 s for control, 35.3 ± 7.4 s for cKO; n = 12 for control and cKO) mice. A significant difference in total freezing time was observed between control and cKO in single-housed but not in group-housed mice. Tukey post hoc analysis (mean ± SEM, p = 0.903 for group-housed and p = 0.006 [**] for single-housed animals). (C) Quantification of the time spent in the brightly lit area for group-housed (132.5 ± 14.0 s for control, 137.5 ± 21.9 s for cKO; n = 9 for control and 8 for cKO) and single-housed (108.7 ± 15.6 s for control, 136.1 ± 12.9 s for cKO; n = 9 for control and cKO) mice. No significant difference was detected in the light/dark exploration test. Tukey post hoc analysis (mean ± SEM, p = 0.996 for group-housed, p = 0.618 for single-housed animals). The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002386.g005

Our looming stimulus assay evoked prolonged freezing responses (Fig 5B). Indeed, nearly all mice froze within a few seconds of the onset of the 10 s looming stimulus and remained frozen for up to 3 min in some cases. A 2 × 2 ANOVA (genotype × housing condition) showed a main effect of genotype (F = 7.866, p = 0.008) and a main effect of housing condition (F = 9.727, p = 0.004) with no interaction (F = 3.099, p = 0.087). Specifically, the Brn3b cKO group froze less than the control group, and both groups froze longer following single housing. Follow-up post hoc analysis showed a significant difference between groups only in the single housing condition (t = 5.018, p = 0.006; group housing: t = 0.965, p = 0.903). All mouse groups behaved similarly in the light/dark exploration test (Fig 5C). A 2 × 2 ANOVA showed no statistical differences between genotypes (F = 1.011, p = 0.323) and housing conditions (F = 0.6073; p = 0.442). Overall, these results suggest that Brn3b cKO mice have overall normal vision; however, they exhibit impaired visually evoked freezing responses following single housing.

Tac2 overexpression in Brn3b mutants rescues freezing responses to visual threat

Our findings indicated that Brn3b deletion reduces Tac2 expression in the deep SC/PAG and that Brn3b mutants exhibit decreased freezing reactions to visual threat precipitated by single housing. Given that single housing was reported to enhance freezing behavior via Tac2 up-regulation [34], we speculated that decreased Tac2 expression in Brn3b mutants prevents visual fear augmentation induced by social isolation. However, due to a lack of direct evidence that increased expression of Tac2 in dorsal midbrain can affect visually triggered fear behavior, we first examined whether Tac2 up-regulation in the SC/PAG of WT mice is sufficient to enhance freezing responses to the looming stimulus (S11 Fig). To overexpress Tac2, we generated and delivered AAV-Tac2 as well as AAV-mCherry necessary for visualization. The fidelity of co-expressing 2 AAV constructs has been assessed previously [22] and Tac2 overexpression was confirmed by RT-qPCR. AAV-mCherry alone was delivered as a control. Four to 6 weeks after AAV injection, behavioral experiments were conducted. Analysis revealed that Tac2 overexpression in the deep SC/PAG of WT mice is sufficient to increase freezing in the response to a visual threat but has no effect on general visual ability. Subsequent histological analysis verified expression of mCherry in the deep SC/PAG.

Next, we examined whether restoring Tac2 levels in Brn3b mutants would ameliorate impairment in visually triggered defensive freezing (Fig 6A and 6B). We delivered AAV-Tac2 with AAV-mCherry into the deep SC/PAG of cKO mice. AAV-mCherry alone was injected into both control and cKO mice. Four to 6 weeks after AAV injection, animals were individually housed for 2 weeks prior to behavioral experiments.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Overexpression of Tac2 ameliorates behavioral deficit in Brn3b mutants.

(A) Schematic diagram of Tac2 + mCherry delivery into the deep SC/PAG layer. AAV was used as a delivery vehicle. (B) Schematic diagram of the behavioral assays. (C–E) Representative images of mCherry expression in control brains (red, C), mCherry expression in cKO brains (red, D), and Tac2 + mCherry co-expression in cKO brains (red, E). The sections were collected following behavioral assays. DAPI (blue). (F, G) Quantification of freezing responses in the looming stimulus assay (63.4 ± 14.4 s for mCherry in control, 15.9 ± 3.1 s for mCherry in cKO, 64.0 ± 17.2 s for Tac2 + mCherry in cKO; n = 9 for mCherry in control, 10 for mCherry and Tac2 + mCherry in cKO) and time spent in the brightly lit area in the light/dark exploration test (100.1 ± 10.1 s for mCherry in control, 128.2 ± 13.7 s for mCherry in cKO, 114.5 ± 18.1 s for Tac2; n = 9 for mCherry in control, 10 for mCherry and Tac2 + mCherry in cKO). Tac2 overexpression in cKO mice specifically increased their total freezing time, causing cKO animals to freeze similarly to control mice. No significant difference was detected during the light/dark exploration test. Unpaired two-tailed Student’s t test (mean ± SEM, p = 0.004 [**] or p = 0.01 [*] for freezing time; p = 0.123 or 0.553 for time spent in the light). Scale bars: 250 μm. The data underlying this figure can be found in S1 Data. AAV, adeno-associated virus; PAG, periaqueductal gray; SC, superior colliculus.

https://doi.org/10.1371/journal.pbio.3002386.g006

The looming stimulus assay following the single housing showed that Tac2 overexpression rescued the behavioral deficit in the Brn3b mutants, increasing freezing responses to the level exhibited by control mice, thereby ameliorating behavioral deficit displayed in the mutants injected with AAV-mCherry alone (Fig 6F and 6G), whereas there was no effect in the light/dark exploration test. We also confirmed mCherry expression in the deep SC/PAG by the post hoc anatomical analysis (Fig 6C–6E). Collectively, our findings indicate that Tac2 acts downstream of Brn3b in regulating visually triggered fear-related defensive behavior.

Ethics statement

All animal procedures were approved by the Institutional Animal Care and Use Committee at Yale University (protocol number 2022–11368) and were in compliance with federal guidelines (National Research Council (US) Committee).

Animals

Both male and female animals were used in all studies. The age of the animals was specified in each experiment. In the experiments requiring timed pregnancy, the date of a copulation plug detection was considered to be E0.5.

The following mouse lines were used: (1) global Brn3b knockout [47]; (2) conditional Brn3b knockout [21]; (3) Brn3b^GFP/+(this study); (4) engrailed1-Cre (En1-Cre, in which restricted Cre expression in the midbrain and hindbrain has been confirmed [22,23]); (5) Ntsr1-GN209-Cre, in which selective Cre expression was found in LP-projecting SC neurons [27,28]; (6) Ai14 (Cre-dependent td-Tomato expression; JAX # 007914); and (7) C57BL/6J (JAX # 000664).

Mouse generation

The Brn3b^GFP/+ mouse line was generated by the Yale Genome Editing Center using the CRISPR/Cas9 strategy. Briefly, 2 sgRNAs for Brn3b (for 5′ and 3′ ends) were designed using an online tool (http://crispor.tefor.net). A targeting vector, containing GFP with SV40 small t intron sequences flanked by homology arms to 162 bp upstream and to 100 bp downstream of Brn3b coding sequence, was synthesized by Genewiz. Two sgRNAs, the targeting vector and Cas9 were injected into fertilized zygotes that were then implanted in pseudo-pregnant females. PCR, followed by DNA sequencing analysis, identified 4 founders in which the GFP sequence replaced the open reading frame of Brn3b. All 4 founders were crossed to wild-type mice for 2 generations before conducting histological analysis. Double labeling showed approximately 95% overlap between GFP and Brn3b signals in the progeny of all 4 founders (1 animal/founder).

The sequences of the sgRNA were:

for 5′ end, GCTCCGGCCGGGTACTTCTC; for 3′ end, GGAGAAGGGTCCCTAAATGC.

The primer sequences for screening were:

for 5′ end, forward-GGGGACTATAACTCCACCGC, reverse-TCGATGCCCTTCAGCTCGAT;

for 3′ end, forward-AGGACGACGGCAACTACAGG, reverse-GCGAAACCGGTTCACAATCT.

The sequences of homology arms and small t intron were:

left arm: CGGAGCTAGCGGCCACACTGGGAGAACCGGGCCTGGAAGCAGTGGCGGCTGAGCACAACTTTGCAGTGTTGTTCCCTCTGCTGCTCCGGCCGGGTACTTCCTCAGAGGGTCGGGTAGCTGGGACCGGAGTGCGCCAGCGACGAGCGCGCCGCGCAAGGAAAG;

small t intron: GTAAATATAAAATTTTTAAGTGTATAATGTGTTAAACTACTGATTCTAATTGTTTGTGTATTTTAG;

right arm: GGACCCTTCTCCAGGGATGGCCCTTTCCCTTCGCCCTCTTTTTTTCTAACCCCCTTCTTGTCTCTTCTGCCTCTTTCCTTTCTACTTTGGCTATCAGAAA.

Immunohistochemistry

Mice were anesthetized by intraperitoneal injection of 100 mg ketamine plus 10 mg xylazine/kg of bodyweight and perfused transcardially with 4% paraformaldehyde (PFA)/PBS. Following perfusion, the brains were dissected, postfixed overnight at 4°C, incubated successively with 15% sucrose/PBS and 30% sucrose/PBS overnight at 4°C, and sectioned with a cryostat (12 to 20 μm). For vibratome sectioning, the tissues were postfixed overnight at 4°C, washed with PBS, and sectioned (50 to 100 μm). For immunostaining, sections were washed twice with PBS, blocked with 3% donkey serum/0.1% Triton X-100/PBS for 30 min at room temperature, incubated with primary antibodies for 1 to 3 days at 4°C and then with secondary antibodies for 2 h at room temperature.

Primary antibodies used were: rabbit anti-GFP (1:1,000, Millipore, AB3080P), chicken anti-GFP (1:1,000, Aves Laboratories, GFP-1020), rabbit anti-cleaved caspase-3 (1:1,000, Cell Signaling Technology, 9661), rabbit anti-calbindin D-28K (1:4,000, Swant, CB38), rabbit anti-Ki67 (1:500, Thermo Scientific, RM9106), rabbit anti-ETV1 (1:1,000, Abcam, AB36788), rabbit anti-DsRed (1:1,000, Clontech, 632496), mouse anti-Brn3a (1:500, Millipore, MAB1585), goat anti-Brn3a (1:3,000, Santa Cruz, SC31984), mouse anti-calretinin (1:2,000, Millipore, MAB1568), goat anti-Brn3b (1:500, Abcam, AB235268), goat anti-Brn3b (1:250, Santa Cruz, SC31989). Secondary antibodies conjugated to Alexa Fluor-488, Cy3 or Alexa Fluor 647 were diluted at 1:500 (Jackson ImmunoResearch Laboratories, 703-545-155, 711-545-152, 711-165-152, 711-605-152, 715-165-151, 705-165-147, or Thermo Fisher, A-2202, A-11055).

For AP staining, brains were sectioned on a vibratome (50 μm for postnatal mice and 100 to 200 μm for embryos). The tissues were washed with 2 mM MgCl₂/PBS twice, incubated for 2 h at 65°C to inactivate the endogenous AP activity and then washed 3 times with AP buffer (0.1 M Tris, 0.1 M NaCl, 0.05 M MgCl₂ (pH9.5)). AP staining was conducted in the AP buffer with 2 mM levamisole and NBT/BCIP (50X, Roche, 11681451001) for 4 to 16 h. After staining was complete, the tissues were washed with PBS, postfixed with 4% PFA/PBS overnight and dehydrated through EtOH incubation (50%-75%-85%-95%-100%). Just before imaging, the tissues were cleared with benzyl benzoate and benzyl alcohol (2:1).

Construction and generation of AAV

To generate an AAV vector carrying Tac2, a full-length Tac2 was amplified from P1 mouse brain cDNA by generating BamHI and EcoRI restriction sites at each end and subcloned into the AAV vector (Addgene, #74291) [48]. Generation of AAV-YFP and AAV-mCherry constructs was previously described [22]. AAV-Cre construct was obtained from (Addgene, #55636) [49]. Detailed strategies of AAV-FLP-DOG and AAV-fDIO-mCherry were previously described [22,29].

AAV production was based on a triple-transfection, helper-free method, and AAVs were purified as described [22]. A plasmid carrying either AAV capsid 2/1 or 2/9 genes (UPenn Vector Core) was used. The titer of the purified AAVs was determined by quantitative PCR using primers that recognize WPRE; the concentrated titers were >10¹³ viral genome particles/ml in all preparations. Viral stocks were stored at −80°C.

Transcriptome profiling (RNA-seq)

Dorsal midbrains were dissected at P2-P4, snap-frozen on dry ice and stored at −80°C. Frozen samples were sent to the Yale Center for Genome Analysis for RNA extraction, library preparation, and RNA sequencing. Briefly, total RNA was prepared using the RNeasy PowerLyzer Tissue & Cell kit (Qiagen). RNA quality was assessed by estimating the A260/A280 and A260/A230 ratios by nanodrop and Agilent Tapestation 4200 RNA Screen Tape Assay. Samples with RIN values of 7 or greater were used for library prep.

To generate cDNA libraries, 1 μg of total RNA per sample was enriched for polyA RNA and converted into libraries using the Kapa mRNA Hyper Prep Kit for Illumina Platforms (KAPA Biosystems). Briefly, mRNA was purified with oligo-dT beads and sheared by incubation at 94°C in the presence of Mg. Following first-strand synthesis with random primers, second strand synthesis and A-tailing were performed with dUTP for generating strand-specific sequencing libraries. Adapter with 3′ dTMP overhangs were ligated to library insert fragments. Library amplification amplified fragments carrying the appropriate adapter sequences at both ends. Strands marked with dUTP were not amplified. Indexed libraries that meet appropriate cut-offs for both were quantified by qRT-PCR using a commercially available kit and insert size distribution was determined by the LabChip GX or Agilent Bioanalyzer. Samples with a yield of ≥0.5 ng/μl were used for sequencing. The samples were sequenced on an Illumina NovaSeq S2. Briefly, the sample concentrations were normalized to 1.2 nM and loaded onto an Illumina NovaSeq flow cell at a concentration that yields 25 million passing filter clusters per sample. Samples were sequenced using 100 bp paired-end sequencing on an Illumina NovaSeq according to Illumina protocols.

Low-quality base calls, sequences with low-complexity tails and adaptor sequences were removed using fqTrim2.py (Yale Center for Genome Analysis). Trimmed paired-end (PE) reads were aligned to the mouse genome (GRCm38.p5) using Tophat2/Bowtie 2 [50]. Raw read counts per gene were calculated using HTSeq (version v.0.6.1p1) and expression levels in transcripts per million (TPM) were estimated by StringTie (version 1.3.3b) on GENCODE vM15 annotation [51,52]. Three independent replicates for control and 4 for Brn3b mutants were used for further analysis. Differential expression analysis was performed using DESeq2, which produced fold changes and statistical significance of changes between samples [53].

Reverse transcription-quantitative PCR (RT-qPCR)

Midbrains including SC and PAG (around central aqueduct) were isolated. Total RNA was prepared using the RNeasy Mini Kit (Qiagen) and cDNA was synthesized using the Superscript III First-Stand Synthesis SuperMix (Thermo Fisher, #18080–400). RT-qPCR was performed in duplicates using iQ SYBR Green Supermix (Bio-Rad) on the CFX96 real-time system (Bio-Rad). The Ct values of the samples were normalized to that of GAPDH and expression level of each gene relative to that of controls was calculated by ΔΔCt method. Primers used for qPCR were as follows: GAPDH, 5′-GTGGAGTCATACTGGAACATGTAG-3′ and 5′-AATGGTGAA GGTCGGTGTG-3′; Tac1, 5′-TTTCGTAGTTCTGCATCGCG-3′ and 5′-TGGCCAGATCTCTCAC AAAA-3′; Tac2, 5′- GATGTCTCCTTTGGTCCCAC-3′ and 5′-AGGGAGGGAGGCTCAGTAAG-3′; Crh, 5′-ATCTCTCTGGATCTCACCTTCC-3′ and 5′-CCCGATAATCTCCATCAGTTTCC-3′; Brn3b, 5′-ATTGAAGAGCTCCGGCTTAG-3′ and 5′-CTGTCACACAACAACATGATCG-3′.

In situ hybridization

Tac2, Slc17a6, and Gad1 expression in brain tissues was detected by in situ hybridization following the method previously described [15]. Briefly, we amplified an entire coding region of Tac2 (NCBI reference sequence, NM_009312.2) and Slc17a6 (NM_080853) or the partial coding sequence of Gad1 (NM_008077) from the brain cDNA and subcloned into pCR8/GW/TOPO TA vector (Invitrogen) or pGEM Teasy (Promega). Antisense riboprobes were synthesized using digoxigenin-labeled UTP. Signals were detected by peroxidase-conjugated anti-digoxigenin antibodies and by the tyramide signal amplification system (TSA-Plus, PerkinElmer Life Sciences). To examine overlap between Tac2+ and Brn3b+ cells in Brn3b^GFP/+ mouse brain, the signals were first amplified by anti-digoxigenin antibody to Tac2 and then by antibodies to GFP (i.e., Brn3b).

Intraocular injection and enucleation

Animals received analgesic buprenorphine (0.05 to 0.1 mg/kg of body weight) before surgery and were then anesthetized with a mixture of 100 mg ketamine plus 10 mg xylazine/kg of bodyweight. All drugs were administrated intraperitoneally. To visualize axonal distribution of RGCs in the SC, a small hole was made in an eye with an insect pin (size 00) to release intraocular pressure. Cholera toxin B subunit conjugated to Alexa Fluor-555 (CTB-555; 1 μl of 1 mg/ml, Invitrogen) was injected through the same hole using a Hamilton syringe. Intraperitoneal injections of buprenorphine (0.05 to 0.1 mg/kg of body weight) were given for 48 h after the surgery.

For enucleation, the mice were treated with buprenorphine and meloxicam and anesthetized with a mixture of ketamine and xylazine described above. Before removal of eyes, lidocaine (<7 mg/kg of body weight) was administrated around the surgical area by subcutaneous injection. The eyes were removed using sterile surgical scissors. Eyelids were sutured to prevent unwanted infection using surgical specialties. Animals then received intraperitoneal injection of buprenorphine (0.05 to 0.1 mg/kg of body weight) and meloxicam (1 to 5 mg/kg of body weight) for 48 h after surgery.

Brain injections

To trace axonal projections from distinct layers of the midbrain, we injected AAV-mCherry (or AAV-YFP) into wild-type mice or AAV-Cre into Brn3b^flox/+ mice. The mice received intraperitoneal injection of buprenorphine (0.05 to 0.1 mg/kg of body weight) before surgery and were then anesthetized by intraperitoneal injection of a mixture of 100 mg ketamine plus 10 mg xylazine/kg of bodyweight before being mounted on a stereotaxic apparatus. Lidocaine (<7 mg/kg of body weight) was administrated around the surgical area by subcutaneous injection before incision. A small craniotomy was made over the SC using a dental drill. Coordinates used for injection were bregma −4.5 mm, lateral ± 0.6 mm, dura −1.2 mm (for the superficial SC), and −2.4 mm (for the deep SC/PAG). AAVs (approximately 50 nl) were unilaterally injected with a glass pipette at a rate of approximately 10 nl/min, and the pipette was left in place for 5 min after injection. Enucleation was conducted after the brain injection on the same day when AAVs were delivered into the intermediate SC. Intraperitoneal administration of buprenorphine (0.05 to 0.1 mg/kg of body weight) and meloxicam (1 to 5 mg/kg of body weight) was given for 48 h after the surgery. Three to 4 weeks later, the brain tissues were dissected and processed to visualize axonal projections.

To overexpress Tac2, we injected a mixture of AAV-Tac2 and AAV-mCherry, or AAV-mCherry alone for control into the deep SC and PAG. General surgical procedures were performed as described above. Coordinates used for injection were bregma −4.5 mm, lateral ± 0.6 mm, dura −2.4 mm. Four to 6 weeks later, the animals were single-housed for 2 weeks before conducting behavioral tests. After completion of the tests, the brain tissues from all tested animals were dissected, sectioned, and imaged to confirm mCherry expression in the deep SC and PAG.

To label superficial Brn3b+ neurons by FLP-DOG strategy, we injected AAV-FLP-DOG and AAV-fDIO-mCherry into the SC of Brn3b^GFP/+ mouse at P1. Mice received oral injection of meloxicam (0.3 mg/kg of bodyweight) before surgery and were anesthetized by chilling on ice. Pups had no direct contact with ice, as they were placed on a barrier covering the crushed ice. A small incision (maximum length 3 mm) was made on the skin over the SC. The skull and brain tissue were penetrated at the same time by disposable glass pipettes (<0.1 mm outer diameter) containing AAVs. A pressure injector was used to deliver approximately 50 nl of AAVs unilaterally. The pipette was gently withdrawn after injection and the skin was sealed using cyanoacrylate glue. Oral injection of meloxicam was given for 48 h after surgery. Three to 4 weeks later, the brain tissues were dissected and further processed for histological analysis.

Behavioral tests

Both male and female mice (16 to 20 weeks old) were utilized in the behavioral experiments; the subject numbers were sex-adjusted. Different behavioral assays were conducted 3 to 4 days apart for group-housed mice and 5 to 7 days apart for single-housed mice to allow for additional post-experimental recovery of the animals that experienced social isolation. All behavioral tests were performed and analyzed in a double-blind fashion.

Visually triggered defensive behavior was tested using a looming stimulus assay as described in Yilmaz and Meister. Briefly, the behavioral apparatus consisted of an open field-type arena with a computer monitor mounted on the top of the plexiglass chamber. All sidewalls of the chamber (40 × 40 × 30 cm) were coated with a matte finish (Krylon) to create a unidirectional visual perspective. The mouse was initially habituated to the test room for approximately 45 min and then placed into the arena to acclimate for approximately 10 min. During this time, the computer monitor illuminated the arena with a gray screen (mean luminance). Mouse behavior in the arena was recorded by a camera mounted on the chamber and the video was stored on a computer using video capture software. Following acclimation, the computer screen, regulating illumination of the test area, began presenting an expanding dark spot—a negatively contrasted circle expanding from 2 to 20 degrees of visual angle over 250 ms and then remaining at its largest size for an additional 250 ms. After a 500-ms interstimulus interval, during which the gray screen was presented, the looming stimulus repeated for a total of 10 times (1 stimulus s^-1). The stimulus was generated in MATLAB. Mouse freezing responses were tested in the absence of a shelter and analyzed using Boris 7.13, an open-source event tagging software [54]. Freezing time was calculated as a time period when the mouse remained completely immobile during 3 min of poststimulus video recording.

Visual discrimination ability was tested using a light/dark exploration test as previously described [38,55]. Briefly, the light/dark chamber (40 × 40 × 30 cm) was made of transparent plexiglass and contained a black plexiglass exploration box (20 × 40 × 30 cm) occupying about half of the total chamber area. The dark box had a removable black plexiglass cover and small exit/entrance space at the center. The light compartment was diffusely illuminated by a bright 120 W lamp centered over the lit area. A mouse was habituated to the test room for approximately 45 min before the experiment. A trial began by placing the mouse into the dark compartment and allowing it to freely explore the chamber. Mouse behavior was video recorded for 10 min. The video-recorded behavior, including the total time spent by the animal in the brightly lit area, was analyzed manually using Boris version 7.13.

Image acquisition and analysis

Fluorescent images were collected using a Zeiss Imager M2 fluorescence microscope and a Zeiss LSM 800 confocal microscope. The z-stacks were obtained with 1 μm steps using a 20× objective (NA 0.8) or 4 μm steps using a 10× objective (NA 0.45). AP images were acquired using a Zeiss Axioskop equipped with SPOT software (Diagnostic Instruments, Sterling Heights, Michigan, United States of America). Data analysis was conducted by ImageJ software (National Institutes of Health, Bethesda, Maryland, USA).

To indicate anatomical location of brain sections/images used for analysis, the Bregma levels were indicated on the figures depicting sections obtained from approximately 24-day-old mice. For younger animals when the brain is still developing and/or a clear Bregma is lacking, the following localization and visualization approaches were employed. For coronal images used for analysis, diagrams of the coronal sections showed their specific brain positions and the corresponding sagittal images depicted the levels where each section was obtained. Anatomical structures of the coronal sections were defined using information from the Atlas of the Developing Mouse Brain [56]. For sagittal images used for analysis, the level of each section was shown as a lateral distance from the midsagittal line. For S8 Fig, the anatomical level for images was described in the figure legend.

To quantify axonal projections to the LP, AP images were taken using 50-μm thick coronal sections, approximately 300 μm apart, extending from the suprachiasmatic nucleus to the rostral cerebellum. Sections containing the LP adjacent to the dorsal lateral geniculate nucleus (dLGN) were chosen for further analysis to ensure the correct LP location. The poor quality of antibodies to AP made it difficult to conduct double immunostaining with an antibody to calretinin, which demarcate LP [22]. Therefore, LP localization was determined based on anatomical criteria. The areas covered by axon terminals were measured using ImageJ. Total LP areas from 2 or 3 sections (depending on the animal’s age) were combined, quantitatively compared to controls, and presented as percentages indicating relative differences. This approach allowed us to rule out the possibility that different sectioning levels along the anterior-posterior axis were compared between animals. To visualize dLGN and LP by double immunostaining at P10, antibodies to calbindin-D28K were used [57].

To assess and compare the total number and distribution of cells stained with specific antibodies, sagittal sections anatomically matched in controls and mutants, were used unless noted otherwise. Images were taken from the sections located at approximately 100 μm from the medial edge and continued laterally, approximately 50 μm apart by E16, and approximately 75 μm apart by P1/P2. For cell counting, the images were cropped to 350× up to 450 μm at E15; to 350× up to 500 μm at E16; to 350× up to 800 μm at E18; and to 350 × 1,000 μm at P1/P2. Such cropping covers an area from the pia to the dorsal PAG at each developmental stage. For ETV1+ neurons, the entire image was used for the analysis. To assess the distribution of Brn3a+ neurons and GFP+ neurons, the counted cells were grouped within 50 μm intervals for E15/E16 analysis and within 100 μm intervals for P1 analysis. To analyze the number of Tac2+ neurons, the images were taken from coronal sections located just rostrally to the inferior colliculus and continued rostrally approximately 240 μm apart at P13-P18; the images were cropped to 700 × 700 μm for counting.

Experimental design and statistical analysis

Analysis of axonal projections was conducted in 2 to 3 different sections per animal and averaged for 4 to 5 animals per condition. The number and distribution of cells were analyzed using 2 to 3 different sections per animal and averaged for 3 to 5 animals per condition. Animals of either sex were utilized for the analysis.

All data were reported as mean ± SEM and analyzed using GraphPad Prism software (GraphPad Software; Dotmatics, San Diego, California, USA). Statistical analyses were conducted using an unpaired two-tailed Student’s t test when comparing means between 2 groups or two-way ANOVA followed by Tukey post hoc test when comparing means between multiple groups. The number of animals used per condition, the mean values ± SEM for each data set, and statistical significance were provided in each figure/figure legend. Statistical significance was determined when p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).