Paratenial thalamus engages in reciprocal and broadcast circuits with the prefrontal cortex

SUMMARY

The dorsal anterior midline thalamus (aMT) consists of two adjacent nuclei important for motivated and emotional behavior. Previous work focused on the paraventricular thalamus (PVT), with little known about the adjacent paratenial thalamus (PT). We examine neural circuits involving the PT using molecular profiling, anatomical tracing, electrophysiology, and optogenetics. We find that protein kinase C-delta (PKCd) selectively labels thalamocortical (TC) cells in the PT but is sparse in the PVT. We show that TC cells in the PT project to the infralimbic region (IL) of the medial prefrontal cortex (mPFC), where they engage L2/3 pyramidal cells. We find that mPFC primarily projects to the PT over PVT, making connections onto reciprocally connected TC cells. However, these cortical inputs are even stronger onto thalamostriatal (TS) and thalamoamygdala (TA) cells in the PT, allowing the mPFC to broadcast to the subcortex. These findings help parcellate aMT, highlight PT as a distinct nucleus, establish reciprocal connectivity with the mPFC, and show corticothalamic throughput to the subcortex.

Graphical abstract

In brief

Dao and Carter find protein kinase C delta as a selective marker of the paratenial thalamus in the anterior midline thalamus and show that the paratenial thalamus engages multiple circuits within the limbic system, including reciprocal loops with the infralimbic cortex and subcortical projections to the nucleus accumbens and basolateral amygdala.

INTRODUCTION

The dorsal anterior midline thalamus (aMT) governs diverse behaviors and physiological states, including feeding,¹ arousal,² salience encoding,³ reward learning,^4–6 aversive learning,^7–9 and social interactions.¹⁰ The aMT consists of two adjacent nuclei: the paraventricular thalamus (PVT) and the paratenial thalamus (PT).¹¹ Previous studies have focused on the PVT, which interacts with the medial prefrontal cortex (mPFC), nucleus accumbens (NAc), and amygdala.^12–14 In contrast, very little is known about the PT, including the different cell types and their input and output connections.

Anatomical studies indicate that the aMT receives prominent long-range afferents from the mPFC^15–17 and also contains thalamocortical (TC) cells that project back to the mPFC.^18,19 Corticothalamic (CT) cells in the mPFC make strong inputs to reciprocally connected TC cells in other thalamic nuclei, which enable persistent activity, learning, and memory.²⁰ For example, we previously found that CT cells in the prelimbic mPFC engage TC cells in several higher-order thalamic nuclei, including the mediodorsal (MD) and ventromedial (VM) thalamus.²¹ CT inputs from the mPFC are also thought to drive activity in the PVT, which is responsible for a variety of motivated and emotional behaviors.^6,22–24 However, the relative strength and influence of CT inputs to reciprocally connected TC cells in the PVT and PT remains unexplored.

In addition to this cortical output, the aMT projects to other parts of the limbic system, including the NAc and basolateral amygdala (BLA).^9,19,25 Through these connections, thalamostria tal (TS) and thalamoamygdala (TA) cells in the PVT are responsible for feeding,¹ arousal,² valence processing,²⁶ reward learning,^4,5 and aversive learning.^7,9 Dysfunction of these subcortical projections is also implicated in a variety of maladaptive behaviors and neuropsychiatric disorders, including stress,²⁷ pain,^28,29 and substance use disorders.^30–33 However, the cellular properties of TS and TA cells in the PT, their ability to be engaged by long-range inputs, and their long-range influence on the NAc and BLA have not been examined.

A major challenge for studying the aMT is the small dimensions and close proximity of the PT and PVT, which often makes it difficult to assign a precise location to cells and inputs.³⁴ For example, the PVT and PT are only a few hundred microns in diameter and have no discernable anatomical border, making it difficult to inject viral tools or assess if cells and inputs are restricted. In principle, identifying genetic markers could help delineate these nearby nuclei and better define their cells and connections.^35,36 Selective markers now exist for PVT, including neurotrophic receptor tyrosine kinase 1 (Ntrk1) and neurotensin (Nts).^26,36 Other markers used for PVT include the dopamine receptor 2 (Drd2)^22,33 and calretinin (Calb2).^37,38 However, finding selective markers for PT has proven difficult, limiting our ability to identify different cells and inputs and study neural circuits in this nucleus.

Here, we first examine the molecular profile of aMT, identifying protein kinase C delta (Prkcd or PKCd) as a specific marker for PT. Using virus-based circuit mapping, slice electrophysiology, and optogenetics, we then show that PKCd-expressing (PKCd+) neurons are TC cells, which are concentrated in the PT and project to the infralimbic (IL) mPFC. We find that TC cells in the PT receive a range of afferents, including from L6 CT cells in the IL mPFC. Moreover, we show that CT inputs are heavily biased onto PT over PVT, where they can drive action potential (AP) firing of TC cells. Last, we show that mPFC inputs are even stronger onto TS and TA cells, which they preferentially activate to engage the subcortex. Together, our data reveal cell types and synaptic connections in the PT, demonstrating the presence of both a cortico-thalamo-cortical loop and cortico-thalamo-limbic broadcast circuits.

RESULTS

PKCd is a molecular marker for paratenial thalamus

To distinguish adjacent nuclei in the aMT, we explored the distributions of genetic markers, including those known to label the PVT.^35,36 We first used fluorescent in situ hybridization (FISH) in thin coronal slices (10 μm) registered to the Allen Mouse Brain Common Coordinate Framework (CCF) using the DAPI signal³⁹ (Figure 1A). We found that both neurotrophic receptor tyrosine kinase 1 (Ntrk1)³⁶ and neurotensin (Nts)²⁶ labeled the PVT but not the PT (Ntrk1 count: PVT = 81.6 ± 5.8 cells, PT = 2.0 ± 0.6 cells, p < 0.0001; Nts count: PVT = 25.3 ± 7.5 cells, PT = 2.0 ± 0.5 cells, p = 0.01; N = 5 mice) (Figure 1B). In contrast, dopamine receptor D2 (Drd2) was minimally expressed in the aMT, despite prominent expression in the striatum and posterior PVT^22,33 (Drd2 count: PVT = 0.4 ± 0.4 cells, PT = 0.4 ± 0.2 cells, p > 0.99, N = 3 mice) (Figures 1B & S1A). Calretinin (Calb2) was prominent in the PVT but also extended into the PT^15,37,38,40 (Calb2 count: PVT = 126.7 ± 27.5 cells, PT = 118.4 ± 22.2 cells, p = 0.6, N = 5 mice) (Figure 1B). However, protein kinase C delta (Prkcd or PKCd) was almost exclusively expressed in PT, with minimal labeling of PVT (Prkcd count: PVT = 3.4 ± 0.3 cells, PT = 34.6 ± 4.5 cells, p < 0.001, N = 5 mice, two-way ANOVA: genes F(5,21) = 15.34, p < 0.0001, areas F(1,21) = 6.36, p = 0.019, genes × area F(5,21) = 57.82, p < 0.0001, Šídá k’s multiple comparison tests) (Figure 1B). The PT/PVT ratio indicated that Prkcd expression is 10-fold higher in PT than PVT, and also higher than all PVT-specific markers (Nts = 0.09, 95% confidence interval [CI] = 0.06–0.16; Ntkr1 = 0.02, 95% CI = 0.006–0.07; Calb2 = 0.9, 95% CI = 0.7–1.1; Drd2 = 0.8, 95% CI = 0.43–1.67; Prkcd = 9.8, 95% CI = 5.8–16.4; Prkcd vs. Nts: p < 0.0001, Prkcd vs. Ntrk1: p < 0.0001, Prkcd vs. Calb2: p = 0.22, Prkcd vs. Drd2: p = 0.3, Kruskal-Wallis χ²(4) = 21.11, p = 0.0003, Dunn’s multiple comparison tests) (Figure 1C). We confirmed this specificity with immunohistochemistry, labeling coronal slices for protein expression, again registering to the CCF (Figure 1D). We found that CALB2 protein was expressed in both PVT and PT (PVT = 2,319 ± 536 cells, PT = 2,145 ± 439 cells, p = 0.3, N = 4 mice), but PKCd protein was largely restricted to PT (PVT = 62.5 ± 12.1 cells, PT = 481.3 ± 20.5 cells, p < 0.001, N = 4 mice; two-way ANOVA: proteins F(1,6) = 16.35, p = 0.007, areas F(1,6) = 2.914, p = 0.138, proteins × areas F(1,6) = 17.24, p < 0.0001, Šídá k’s multiple comparison tests; PT/PVT ratio: CALB2 = 0.9, 95% CI = 0.7–1.1; PKCd = 8.0, 95% CI = 4.9–13.2; Mann-Whitney U = 0, p = 0.028) (Figures 1E and 1F). Together, these results indicate that PKCd is a selective marker for the PT in the aMT, distinguishing it from the nearby PVT and enabling the study of its cells and connections.

Figure 1. PKCd expression is prominent in the paratenial thalamus.

(A) Confocal images of CCF-aligned coronal slices containing the anterior midline thalamus (aMT) (upper left), with individual panels showing magnification of the yellow dashed box, including neurotrophic receptor tyrosine kinase 1 (Ntrk1, orange), neurotensin (Nts, green), dopamine receptor 2 (Drd2, gray), calretinin (Calb2, red), and protein kinase C delta (Prkcd, magenta) mRNAs in the paratenial thalamus (PT) and the paraventricular thalamus (PVT). Note that Ntrk1 and Drd2 images come from a different slice. Scale bars: 1 mm and 200 μm.

(B) Summary of Calb2-, Ntrk1-, Nts-, Drd2-, and Prkcd-expressing cell counts in the PT and PVT (N = 3–5 mice each).

(C) Summary of Calb2, Ntrk1, Nts, Drd2, and Prkcd PT/PVT ratio.

(D) Confocal images of immunohistochemically labeled CALB2 (red) and PKCd (magenta) proteins in the aMT. Scale bars, 200 μm.

(E) Summary of CALB2+ and PKCd+ cell counts in PT and PVT (N = 4 mice).

(F) Summary of CALB2+ and PKCd+ PT/PVT ratio.

Values are mean ± SEM (B and E) or geometric mean ±95% CI (C and F). *p < 0.05 by two-way ANOVA followed by Šídá k’s multiple comparison tests (B and E), by Kruskall-Wallis followed by Dunn’s multiple comparison tests (C), or by Mann-Whitney’s test (F) (see also Figures S1).

Thalamocortical cells express PKCd in paratenial thalamus

Neurons in the dorsal aMT send outputs to several targets, including mPFC, NAc, and BLA. We next used retrograde labeling to assess the distributions of these different projection neurons in the PT and PVT. We found that injection of rabies virus expressing GFP (RV-GFP) into the mPFC labeled presynaptic TC cells primarily in the PT, with much less labeling in the nearby PVT (PVT = 195.6 ± 32.8 cells, PT = 549.0 ± 73.5, p < 0.0001, N = 6 mice) (Figures 2A and 2B). To rule out any tropism for PT or PVT projection neurons, we compared retrograde tracing using RV, AAV-retro (AAVrg), and cholera toxin subunit B (CTB). Similar preferential labeling of the PT was observed with each method, although AAVrg labeled fewer cells in the thalamus (Figures S1B and S1C). Staining for PKCd confirmed the location of the PT and indicated that most retrogradely labeled TC cells are PKCd positive (PKCd+) (Figures 2C and S1D). In contrast, injections into the NAc or BLA labeled TS and TA cells in both PT and PVT, with more labeling in the latter (TS: PVT = 491.8 ± 38.6 cells, PT = 238.2 ± 35.7 cells, p = 0.007, N = 5 mice; TA: PVT = 340.4 ± 43.3 cells, PT = 115.8 ± 28.8 cells, p < 0.001, N = 5 mice, two-way ANOVA: cell types F(2,13) = 6.42, p = 0.01, areas F(1,13) = 3.09 p = 0.101, cell types × areas F(2,13) = 73.0, p < 0.0001, Šídá k’s multiple comparison tests; PT/PVT ratio: TC = 2.8, 95% CI = 2.2–3.5; TS = 0.4, 95% CI = 0.3–0.7; TA = 0.3, 95% CI = 0.2–0.4, p < 0.0001, Kruskal-Wallis χ²²(2) = 12.53, p < 0.0001, with Dunn’s multiple comparison tests) (Figures 2A and 2B). Neither TS nor TA cells showed any overlap with PKCd labeling, indicating that they represent a separate population of projection neurons in the PT (PKCd+ co-labeled count: TC = 274.5 ± 32.6 cells, TS = 7.8 ± 2.3 cells, TA = 6.6 ± 1.1 cells, TC vs. TS: p = 0.01, TC vs. TA: p = 0.008, Kruskal-Wallis χ²(2) = 10.66, p < 0.0001, with Dunn’s multiple comparison tests) (Figures 2C and S1D). Dual retrograde injections into the NAc and BLA of the same animals showed some co-labeling of cells in both PT and PVT, suggesting that at least a fraction of these neurons can bifurcate to both subcortical targets (Figure 2D). All three types of projection neurons were co-labeled with CALB2 (Figures S1E and S1F), in agreement with previous reports suggesting that CALB2+ cells in the aMT project broadly to multiple structures.³⁷

Figure 2. Characterizing projection neurons in the paratenial thalamus.

(A) Left: images of RV-H2B-GFP injections into mPFC (Ai, green), NAc (Aii, cyan), and BLA (Aiii, yellow), with injection sites outlined in white. Middle: magnified confocal images showing retrogradely labeled TC, TS, and TA cells in the PT and PVT, overlaid with PKCd IHC (purple). Right: further magnified confocal images showing PKCd IHC (top), RV-labeled cells (middle), and merged images (bottom). Scale bars: 1 mm, 200 μm, and 50 μm.

(B) Left: summary of total retrogradely labeled TC, TS, and TA cell counts in the PT and PVT. Right: summary of PT/PVT ratio (right) (N = 5–6 mice each).

(C) Summary of PKCd+ retrograde-labeled TC, TS, and TA cell counts in the PT (N = 5–6 mice each).

(D) Left: confocal image of TS (blue) and TA (yellow) cell labeling in the aMT. Scale bar, 200 μm. Right: summary of TS-only, TA-only, and dual-labeled cells in PT (N = 3 mice).

(E) Left: representative image of AAVrg-labeled TC cells (green) in the PT identified by PKCd IHC (purple). Middle: dendritic reconstruction. Right: membrane potential responses to current injections. Scale bars: 200 μm, 50 μm, and 20 mV × 100 ms.

(F) Summary of frequency-current (F-I) curves (n = 8–9 cells each, N = 4–5 mice each).

(G) Summary of total dendritic lengths (n = 5–9 cells, N = 4–6 mice).

Values are mean ± SEM (B, C, D, F, and G) or geometric mean ±95% CI (B). *p < 0.05 by two-way ANOVA followed by Šídá k’s multiple comparison tests (B), or by Kruskall-Wallis followed by Dunn’s multiple comparison tests (C) (see also Figures S1 and S2).

In separate experiments, we also explored the morphological and physiological properties of TC, TS, and TA cells in PT, as well as TC cells in PVT. After injecting retrograde tracers, we prepared acute brain slices and made whole-cell current-clamp recordings from labeled cells (Figure 2E). We injected current steps to assess AP firing and passive properties and filled cells with Alexa 594 to image morphology. We found that all three cell types fired APs in response to positive current injections and displayed rebound firing to negative current injections (Figure 2F), similar to TC cells in other parts of the thalamus.^21,41 We also found that all three cell types had characteristic radial dendrites (Figure 2G), which were broadly similar to TC cells in other parts of the thalamus.^21,41 We found minimal differences in the resting or passive properties of these cells, but did observe subtle differ ences in rebound AP firing (Figures S2A and S2B). We further quantified dendritic morphology with Sholl analysis and found no differences between these cell types (Figure S2C). These findings indicate that cells in the PT have morphological and physiological properties similar to other parts of the thalamus.

PKCd+ cells in the paratenial thalamus engage the infralimbic mPFC

To explore the connectivity of PKCd+ cells in the PT, we next injected AAV-DIO-YFP into the aMT of PKCd-Cre mice (N = 5 mice) that express Cre under the PKCd promoter.⁴² YFP-expressing cells overlapped with PKCd labeling in PT but not PVT (Figure 3A), consistent with our FISH and immunohistochemistry (IHC) labeling experiments and confirming the specificity of this mouse line. We looked for green axons in serial coronal sections of the brain and observed prominent labeling in several frontal cortical areas, with the most labeling in the IL cortex (Figures 3B and 3C). In contrast, we observed minimal axonal labeling in subcortical areas that are potential targets of aMT, including NAc and BLA (Figures 3B and 3C), consistent with the lack of PKCd overlap in TS and TA cells in our retrograde labeling experiments. Quantifying fluorescence across layers of the IL indicated that axons were concentrated in both cortical layers 1 (L1) and 2/3 (L2/3) (Figure 3D), similar to equivalent inputs from the MD thalamus to mPFC,^21,43 but in contrast to inputs from the VM thalamus.^44,45 Interestingly, PKCd+ axons closely overlapped with PKCd IHC in the mPFC (Figure 3D) and PKCd IHC densely labels thalamo-recipient layers in many other parts of the cerebral cortex (Figure S3A), suggesting that this may be a broad marker of TC connections across the brain. The intrinsic properties and dendritic morphology of these PKCd+ cells resemble those of retrogradely labeled TC cells (n = 8 cells, N = 3 mice) (Figures S3B–S3E). These findings confirm that PKCd+ cells in the aMT are primarily restricted to the PT and indicate that they are TC cells.

Figure 3. PKCd+ cells in the paratenial thalamus drive the infralimbic mPFC.

(A) Left: schematic of AAV-DIO-YFP injection into the anterior midline thalamus (aMT) of PKCd-Cre mice. Middle: CCF-aligned coronal slice showing YFP labeling (green) and PKCd labeling (purple). Right: magnification of the PT and PVT, showing labeled cells only in the PT. Scale bars: 1 mm and 200 μm.

(B) Coronal slices showing PKCd+ axon labeling in the IL (left) and absence of labeling in the NAc (middle) and BLA (right). Areas of interest are outlined in white. Scale bars, 1 mm.

(C) Summary of PKCd+ axon distributions across cortical and subcortical areas (N = 5 mice).

(D) Left: confocal image of PKCd+ axons (green) and PKCd IHC labeling (purple) in the IL. Right: summary of PKCd+ axons and PKCd protein expression in the IL (N = 5 mice).

(E) Left: average PKCd+ cell-evoked EPSCs at L2/3 pyramidal cells in the IL (top), MSNs in the NAc medial shell (middle), and pyramidal cells in the BLA (bottom). Scale bars, 50 pA × 50 ms. Right: summary of EPSC₁ amplitudes in the IL, NAc, and BLA (n = 7–9 cells, N = 5 mice each). Note that the NAc and BLA values are approximately zero and overlap with the y axis.

(F) Left: PKCd + cell-evoked firing of L2/3 pyramidal cells in the IL, showing 5 separate trials. Scale bars, 20 mV × 50 ms. Right: summary of firing rate as a function of the stimulation pulse number (n = 6 cells, N = 5 mice).

Values are mean ± SEM (see also Figure S3).

To explore functional connections from the PT to IL and other areas, we next injected AAV-DIO-ChR2 into the aMT of PKCd-Cre mice. After waiting for expression and transport of virus, we prepared acute coronal slices of the IL, NAc, and BLA. We used optogenetics to stimulate inputs with trains of blue light, which trigger presynaptic release even from severed axons (5 pulses, 2 ms duration, 10 Hz frequency, 5–10 mW power, 470 nm).^46,47 We found that the PT inputs evoked prominent but depressing excitatory postsynaptic currents (EPSCs) at L2/3 pyramidal cells in the IL (EPSC₁ = 178.8 ± 21.4 pA, PPR [EPSC₅/EPSC₁] = 0.3 ± 0.0, n = 9 cells, N = 6 mice) (Figures 3E and S3F), reminiscent of MD inputs to the prelimbic cortex (PL).^21,45,48 In contrast, PKCd+ PT inputs evoked minimal EPSCs at medium spiny neurons (MSNs) in the NAc (EPSC₁ = 0.04 ± 0.03 pA, n = 8 cells, N = 6 mice) and pyramidal cells in the BLA (EPSC₁ = 0.09 ± 0.02 pA, n = 8 cells, N = 6 mice) (Figures 3E and S3F). Within the mPFC, optical stimulation evoked direct excitation at EGABA (60 mV) and feedforward inhibition at E_Glu (+20 mV) with a delay of 6.6 ± 1.2 ms (n = 7 cells) (Figure S3G). In separate current-clamp recordings, repetitive stimulation also evoked strong but transient firing of L2/3 pyramidal cells, similar to MD but not VM inputs (AP probability at pulse 1 = 0.6 ± 0.1; n = 6 cells, N = 5 mice) (Figure 3F). Together, these results indicate that PKCd+ cells in the PT are TC cells that contact and drive activity at L2/3 pyramidal cells in IL, but not cells in the subcortical areas like NAc and BLA.

Paratenial thalamus receives robust inputs from infralimbic mPFC

Neurons in aMT are known to receive long-range afferents from several other brain regions, but most studies have focused on the PVT.^15–17 To identify long-range inputs specifically onto PT cells, we next used Cre-dependent rabies tracing.^49–51 In PKCd-Cre mice, we first injected AAV helper viruses to express the G glycoprotein and the TVA receptor. After waiting 2–3 weeks for expression, we then injected EnvA-pseudotyped RV virus in the PT to express histone-tagged GFP in presynaptic cells (Figure 4A). Eight days later, we sectioned the brain, aligned to the CCF, and detected fluorescent starter cells in the PT. We used semi-automated software to detect presynaptic input cells residing in different locations across the brain (n = 57,883 ± 7,515 cells, N = 4 mice) (Figures 4B, S4A, and S4B). We found most presynaptic input cells in the frontal cortex (28.5% ± 3.4% of total presynaptic cells), with fewer cells in the hypothalamus (13.8% ± 1.7%), septum (9.4% ± 1.2%), and hippocampus (7.1% ± 1.0%) (Figure 4C). We observed no presynaptic input cells in the NAc or BLA. Within the cortex, most presynaptic input cells resided in the mPFC (31.5% ± 5.6% of total cortical inputs), including the PL and IL, with fewer cells in the orbitofrontal cortex (15% ± 1.5%) and anterior insular cortex (15.3% ± 2.2%) (Figure 4C). Within the IL, most presynaptic input cells were in layer 6 (L6), with a relatively small number of cells in layer 5 (L5) (Figures 4D and S4C). Together, these results indicate that the mPFC is a major input to the PKCd+ PT cells, suggesting the possibility of a reciprocal cortico-thalamo-cortical loop linking these two brain regions.

Figure 4. Identifying long-range inputs to PKCd+ cells in the paratenial thalamus.

(A) Left: schematic of injections of AAV helper virus and pseudotyped RV into the aMT of PKCd-Cre mice. Middle: images of coronal slice. Right: magnification of the aMT, showing GFP+ (green), tdTomato+ (red), and starter cells (yellow) in the PT but not the PVT. Scale bars: 1 mm, 200 μm, and 50 μm.

(B) Summary of locations of presynaptic input cells to PKCd+ cells in the PT across the anterior-posterior (A-P) axis of the brain (N = 4 mice).

(C) Left: summary of brain regions containing presynaptic input cells. Right: summary of cortical regions containing presynaptic input cells.

(D) Left: image of corticothalamic (CT) input cells to PKCd+ PT cells in the IL, with the cortical layers from CCF alignment. Middle: overlay of the position of all CT cells as a function of distance to the midline. Right: summary of the distribution of CT cells across the cortical layers in 50-μm bins. Scale bar, 200 μm.

Box and whisker plots represent median and minimum to maximum. Cell counts across cortical layers or A-P coordinates are presented as mean ± SEM (see also Figure S4).

Prefrontal inputs selectively innervate the paratenial thalamus

Having identified mPFC as a major input to PT, we next examined the connections from CT cells in more detail. We first injected AAV-hSyn-eYFP into the mPFC of wild-type animals, waited for expression and transport, prepared coronal slices, and aligned to the CCF, using PKCd labeling to identify the PT (Figure 5A). Strikingly, we found that mPFC inputs almost exclusively innervated the PT, with little labeling of the neighboring PVT (normalized intensity: PVT = 0.3 ± 0.1, PT = 0.9 ± 0.1, N = 5 mice, Wilcoxon W = 21.0, p = 0.03) (Figure 5B). In principle, these connections could arise from either L5 or L6 CT cells in the mPFC, which have different intrinsic properties and send distinct connections to other parts of the thalamus.²¹ We, therefore, used transgenic mice that selectively label mPFC inputs from either layer 5 (L5) or layer 6 (L6) CT cells. Using Syt6-Cre mice that label deep layers of the mPFC,⁵² we found that L6 CT inputs again primarily innervated the PT, with minimal labeling of the PVT (normalized intensity: PVT = 0.2 ± 0.0, PT = 0.9 ± 0.0, N = 5 mice, Wilcoxon W = 21.0, p = 0.03) (Figure 5C). In addition, using Rbp4-Cre mice that label other layers of the mPFC,⁵³ we found that L5 CT inputs showed less overall labeling, but were still biased onto PT (normalized intensity: PVT = 0.1 ± 0.1, PT = 0.3 ± 0.1, N = 3 mice, Wilcoxon W = 6.0, p = 0.25) (Figure 5D). These findings suggest that the mPFC primarily innervates the PT, with minimal connections in the adjacent PVT.

Figure 5. mPFC inputs preferentially innervate the paratenial thalamus.

(A) Left: schematic of AAV-hSyn-eYFP injection into the mPFC of wild-type mice. Middle: representative image of injection site in mPFC. Right: labeling of mPFC axons in aMT. Scale bars, 1 mm.

(B) Left: confocal image of mPFC axons in the PT but not the PVT. Middle: average heatmap from multiple slices (n = 3 sections per animal) and animals (N = 5 mice). Right: summary of fluorescence intensity in the PT and adjacent PVT, normalized to maximum intensity across all thalamic areas. Scale bars, 200 μm.

(C) Similar to (B) for AAV-DIO-YFP injection into mPFC of Syt6-Cre mice.

(D) Similar to (B) for AAV-DIO-YFP injection into mPFC of Rbp4-Cre mice.

(E) Left: average Syt6+ cell-evoked EPCSs recorded at EGABA (60 mV) in TC cells in the PT (bottom, green) and PVT (top, blue). Scale bars, 50 pA × 100 ms. Right: summary of EPSC₁₀ amplitudes at TC cells in the PT and PVT (n = 6 pairs, N = 4 mice).

(F) Left: Syt6+ cell-evoked AP firing at TC cells in the PT (bottom, green) and subthreshold excitatory postsynaptic potentials (EPSPs) in the PVT (top, blue), with multiple traces overlaid. Scale bars, 20 mV × 100 ms. Right: summary of AP probability as a function of pulse number (n = 8 pairs, N = 6 mice).

Values are mean ± SEM. *p < 0.05 by Wilcoxon’s matched-pairs signed rank test (B, C, D, and E), or by two-way ANOVA (F) (see also Figure S5).

Previous studies demonstrate that the mPFC communicates in reciprocal loops with other higher-order thalamic nuclei, including the MD and VM.^21,54 We next asked if mPFC inputs also contact and drive TC cells in the PT by injecting both AAV-DIO-ChR2-YFP and AAVrg-H2B-tdTomato into the mPFC of Syt6-Cre mice. After waiting for expression and transport, we made whole-cell recordings of pairs of TC cells residing in the PT and PVT. To stimulate mPFC inputs, we used trains of blue light (10 pulses, 2 ms duration, 10 Hz frequency, 4–15 mW power, 470 nm). In voltage-clamp recordings, we found that mPFC inputs evoked prominent EPSCs at TC cells in the PT, but minimal responses at TC cells in the adjacent PVT (EPSC₁₀: PVT = 27.1 ± 4.8 pA, PT = 182.6 ± 47.4 pA, n = 6 pairs, N = 5 mice, Wilcoxon W = 21.0, p = 0.03) (Figures 5E and S5A). Responses built over stimulus trains, indicating presynaptic facilitation, consistent with L6 CT inputs to other thalamic nuclei (PPR₁₀: PVT = 9.5 ± 1.5, PT = 8.8 ± 1.5) (Figures S5B and S5C). In separate current-clamp recordings, we also found that mPFC inputs were sufficient to fire APs at TC cells in the PT but not the PVT (AP probability at pulse 10: PVT = 0.04 ± 0.01, PT = 0.3 ± 0.09, n = 8 pairs, N = 6 mice, Wilcoxon W = 28.0, p = 0.01) (Figure 5F). These results show that mPFC inputs preferentially contact TC cells in the PT over PVT, which is consistent with our anatomy and suggests a reciprocal loop between the mPFC and PT.

Prefrontal inputs preferentially contact subcortically projecting cells

Our results indicate that the mPFC makes connections onto TC cells in the PT, forming a reciprocal loop between these structures. Given that the cortex- and subcortex-projecting populations within PT are non-overlapping, we next asked if the mPFC also innervates the TS and TA cells that project to the NAc and BLA, respectively. In this case, we injected both AAV-DIO-ChR2 and a retrograde tracer into the mPFC, along with a second retrograde tracer into either the NAc (N = 6 mice) or the BLA (N = 7 mice) in Syt6-Cre mice. In voltage-clamp recordings from pairs of TC cells and neighboring TS or TA cells, we found that mPFC inputs made facilitating connections onto both TS and TA cells (Figure 6A). However, the responses were much larger at subcortically projecting cells (EPSC₅: TC = 153.4 ± 41.9 pA, TS = 492 ± 114.8 pA, n = 7 pairs, Wilcoxon W = 28.0, p = 0.01; TC = 174.5 ± 95.11 pA, TA = 384.7 ± 111.7 pA, n = 7 pairs, Wilcoxon W = 28.0, p = 0.01) (Figures 6B and S6). The amount of bias was similar for TS and TA cells (EPSC ratio: TS/TC = 3.5, 95% CI = 1.5–7.8, TA/TC = 3.3, 95% CI = 1.6–6.7, Mann-Whitney U = 23, p = 0.9) (Figure 6C), although the amount of facilitation given by the paired-pulse ratio was similar (Figure 6D). In separate current-clamp recordings, we found that mPFC inputs also drive activity in TS and TA cells much more effectively than in TC cells (AP probability: TC = 0.1 ± 0.1, TS = 0.7 ± 0.1, p = 0.008, n = 7 TS/TC pairs, N = 7 mice, Wilcoxon W = 21.0, p = 0.03; TC = 0.2 ± 0.1, TA = 0.8 ± 0.04, p = 0.001, n = 7 TA/TC pairs, N = 6 mice, Wilcoxon W = 28.0, p = 0.01) (Figures 6E and 6F). These findings indicate that while mPFC makes reciprocal connections with TC cells in the PT, it can even more effectively engage distinct cell types that project to subcortical brain areas.

Figure 6. mPFC inputs are stronger onto subcortically projecting cells.

(A) Average Syt6+ cell-evoked EPCSs recorded at EGABA (60 mV) in TC (green), TS (blue), and TA (yellow) cells in the PT. Scale bars, 100 pA × 100 ms.

(B) Summary of EPSC₅ amplitudes in pairs of TC and TS cells (left) (n = 7 pairs, N = 6 mice) and pairs of TC and TA cells (right) (n = 7 pairs, N = 6 mice) in PT.

(C) Summary of EPSC ratio for TC vs. TS and TC vs. TS in PT.

(D) Summary of PPR as a function of stimulation pulse number. Dotted line indicates PPR = 1.

(E) Syt6+ cell-evoked AP firing in TC, TS, and TA cells in the PT, with multiple traces overlaid. Scale bars, 20 mV × 100 ms.

(F) Summary of Syt6+ cell-evoked AP probability as a function of stimulation pulse number for TC vs. TS (left) (n = 7 pairs, N = 7 mice), and TC vs. TA (right) (n = 7 pairs, N = 6 mice).

Values are mean ± SEM (B, D, and F) or geometric mean ±95% CI (C). *p < 0.05 by Wilcoxon’s matched-pairs signed rank test (B), Mann-Whitney’s test (C), or by two-way ANOVA (F) (see also Figure S6).

Polysynaptic circuits link mPFC to NAc and BLA via paratenial thalamus

Together, our findings suggest that CT cells in the mPFC can indirectly access both the NAc and BLA via cells in the PT. However, testing this idea is complicated by the lack of specific genetic markers for TS and TA cells within PT. As an alternative strategy, we took advantage of our finding that mPFC inputs selectively target PT over PVT. We injected AAV1-Cre in mPFC, which at high titer travels anterogradely and expresses Cre in postsynaptic cells.⁵⁵ In the same animals, we also injected AAV-DIO-ChR2-eYFP into aMT, which we found selectively labeled PT over PVT (Figure 7A). In this case, we observed green axons in both the cortical and subcortical brain regions, including the cortex, striatum, and amygdala (Figures 7B and 7C). In the striatum, we found that labeling was concentrated in the core and medial shell of the NAc (Figure 7B), similar to inputs from the PVT.^19,32,36 In the amygdala, we found that labeling was mainly in the medial BLA (Figure 7B) and very sparse in the central amygdala. These findings indicate that while mPFC makes direct contact with NAc and BLA,⁵⁶ it also has a previously unknown indirect path through PT.

Figure 7. Paratenial thalamus routes cortical signals to limbic areas.

(A) Left: schematic of injections of AAV1-Cre into the mPFC and AAV-DIO-eYFP into the PT of wild-type mice. Right: image of CCF-aligned coronal section, showing viral expression in the PT.

(B) Top: coronal slices showing PT axons in the IL (left), NAc (middle), and BLA (right). Bottom: magnification of the IL, NAc, and BLA. Scale bars, 1 mm and 500 μm.

(C) Summary of PT axon fluorescence intensity normalized to maximum intensity across all areas (N = 3 mice).

(D) Left: average PT-evoked EPSCs at the IL L2/3 pyramidal cells (top; green), NAc MSNs (middle; blue), and BLA pyramidal cells (bottom; yellow). Scale bars, 100 pA × 100 ms. Right: summary of EPSC₁ amplitudes in the IL, NAc, and BLA (n = 7–8 cells per area, N = 5 mice).

(E) Left: PT-evoked AP firing in the IL L2/3 pyramidal cells (top), NAc MSNs (middle), and BLA pyramidal cells (bottom), with multiple traces overlaid. Scale bars, 20 mV × 100 ms. Right: summary of evoked AP probability as a function of stimulation pulse number (n = 8–9 cells each, N = 5–6 mice). Values are mean ± SEM (see also Figure S7).

To examine functional connections, we prepared slices and recorded from pyramidal cells in the IL, MSNs in the NAc, and pyramidal cells in the BLA. In voltage-clamp recordings, we found that PT inputs evoked prominent EPSCs in each of these brain regions (EPSC₁: IL = 263.5 ± 47.5 pA, n = 8 cells, NAc = 213.2 ± 33.5 pA, n = 6 cells, BLA = 235.7 ± 75.61 pA, n = 7 cells, N = 8 mice) (Figure 7D), all of which showed short-term depression (PPR₅: IL = 0.5 ± 0.1, NAc = 0.6 ± 0.07, BLA = 0.7 ± 0.1) (Figure S7A). In each case, we also found that PT inputs evoke feedforward inhibition, suggesting engagement of local inhibitory circuits (Figures S6B–S6D). Last, in separate current-clamp recordings, we found that PT inputs could evoke transient action potential firing in each of these brain regions, with highest firing earlier in stimulation trains (AP probability at pulse 1: IL = 0.7 ± 0.1, NAc = 0.5 ± 0.1, BLA = 0.5 ± 0.1) (Figure 7E). Together, these findings indicate that TC, TS, and TA cells in the PT make func tional, potent connections onto both cortical and subcortical targets.

DISCUSSION

We show that the PT and PVT are genetically, anatomically, and functionally distinct nuclei in the aMT. We find that PKCd is a selective genetic marker for TC cells in the PT, while intermingled TS and TA cells are a smaller and distinct population of subcortically projecting neurons. We show that PKCd+ TC cells primarily project to superficial layers of the IL cortex, where their axons overlap with the endogenous PKCd in thalamo-recipient layers and drive transient activity in L2/3 pyramidal cells in the IL. We also find that L6 CT cells in mPFC primarily target PT over PVT, where they contact TC cells in a cortico-thalamo-cortical loop. However, these inputs are stronger onto TS and TA cells, where they effectively drive activity as part of a cortico-thalamo-limbic broadcast circuit. Together, our data reveal PT as a key node in the limbic network, enabling mPFC inputs to be routed to the IL, NAc, and BLA.

The borders between the PT and PVT have been difficult to delineate, historically drawn using peptidergic staining or cytoarchitecture, and consequently are often grouped together. Recent studies have identified selective markers for PVT,^22,35,36 but there are no equivalent studies for PT. We find that previously reported, PVT-specific markers such as Ntrk1³⁶ and Nts²⁶ are not expressed in the PT. While Drd2 is expressed in the posterior PVT,^22,33 we find minimal expression in the anterior PVT or PT. While Calb2 is expressed in the PVT,^15,37,38,40 we find it is also expressed in the PT, indicating it is not a nucleus-specific marker. In contrast, our data show that Prkcd is selectively enriched in the PT, approximately 10-fold higher labeling than the adjacent PVT. Thus, our data indicate that the PT and PVT are genetically separate, with distinct connectivity, and likely different functions.

Our retrograde anatomy data indicate differential distribution of TC, TS, and TA cells in the PT and PVT. We find that TC cells are more abundant in the PT compared to the PVT, consistent with previous studies in both primates⁵⁷ and rodents.¹⁹ TS and TA cells are found across aMT, but we find that they are more abundant in the PVT compared to the PT. TC cells are also highly co-labeled with PKCd, which does not label TS or TA cells, confirming cell-type specificity. In contrast, CALB2 labels TC, TS, and TA cells, confirming a lack of specificity. These different cells in the aMT have sometimes been treated as a homogenous population that projects to multiple areas.^34,37 However, our data indicate there are at least two distinct streams of PT outputs: one targeting the mPFC and the other projecting to the NAc and BLA. We find that all the projection neurons in PT appear to share similar intrinsic physiological properties, including tonic firing at depolarized potentials, burst firing following hyperpolarization, and compact dendrites.^21,41

Using PKCd-Cre mice allowed us to restrict viral expression to PT and avoid PVT, which has been challenging due to the small size and close proximity of these nuclei. In agreement with our retrograde tracing data, we find that PKCd+ TC cell axons are densest in the IL, with few axons in the NAc and BLA. Interestingly, the PKCd+ TC cell axons in the IL resemble those of MD inputs to PL, which terminate in superficial layers 1 and 2/3.^21,45,48 Similarly, the PKCd+ TC inputs display short-term depression and can drive transient activation of L2/3 pyramidal cells. These inputs can also evoke feedforward inhibition in IL, indicating that they engage local inhibitory networks. In future studies, it will be interesting to further characterize the influence of PKCd+ TC synapses on specific microcircuits in the IL. For example, MD inputs activate PV+ interneurons in the anterior cingulate cortex⁴³ but preferentially activate CCK+ interneurons in the PL.⁴⁸

Our anatomy indicates that PKCd expression also closely overlaps with TC axons in both the mPFC and other cortical areas. For example, PKCd is robustly expressed in thalamocortical recipient layers in the visual (L1 and L4), somatosensory (L4 barrels), and motor cortices (L3), suggesting it may serve as a common marker for TC connections. Intriguingly, PKC is also a key effector in many Gq-mediated signaling pathways,⁵⁸ suggesting it may play a role in the regulation of TC synapses. In the future, it will thus be interesting to assess how PKCd influences both modulation and plasticity at TC cells in the thalamus and TC axons in the cortex.

The aMT is often considered an integrative hub that receives convergent inputs from limbic areas involved in arousal and reward. Our rabies-based tracing experiments of presynaptic inputs to PKCd+ cells confirmed that PT samples input from a wide array of cortical and subcortical areas for subsequent routing to the cortex and subcortex. The largest subcortical inputs come from a collection of hypothalamic nuclei, many of which do not send direct connections to the IL.⁵⁹ PKCd+ cells may, therefore, act as a relay center of interoceptive and homeostatic information to the IL for top-down control of behavior. Interestingly, ventral hippocampus sends abundant inputs to PKCd+ cells but in this case, is also directly connected with the IL.^60,61 Our results suggest that the PT thalamus may serve as an additional waystation for the routing of hippocampus signaling to the prefrontal cortex, similar to circuits via the nucleus reuniens.⁶² In future studies, it will be interesting to compare the connectivity and function of these direct and indirect hippocampal-prefrontal connections.

While subcortical inputs are present, the major input to the PT is from the frontal cortex, with the densest presynaptic cell labeling in the IL. Strikingly, anterograde labeling of CT inputs from the IL primarily labels the PT, with minimal labeling of PVT. Similarly, CT inputs preferentially engaged TC cells in the PT, with minimal functional connections onto TC cells in the PVT. These inputs primarily derive from L6 CT cells, which make facilitating synapses that can robustly drive TC cells, with little input from L5 CT cells. These findings are reminiscent of a similar reciprocal loop that links the mPFC with both the MD and VM thalamus.²¹ However, our findings appear to contrast with the prevailing view that the mPFC strongly and reciprocally engages PVT.^6,22–24 One explanation is the difficulty in distinguishing the PT from PVT, as our findings indicate that the commonly used marker CALB2 labels both nuclei. Another explanation is the location within the aMT, particularly on the anterior-posterior axis, as the posterior PVT has distinct genetic identities and connectivity and appears to engage more with the mPFC.^22,35,36

Previous studies in sensory thalamus have focused on reciprocal connections between CT and TC cells,⁶³ including in somatosensory,⁶⁴ visual,⁶⁵ and auditory⁶⁶ systems. We find that PT contains TC cells, which project to IL, as well as intermingled TS and TA cells, which project to the subcortical brain regions. These three populations are also found in the PVT, but there is a bias in distributions with more TC cells in the PT, and more TS and TA cells in the PVT. mPFC inputs can evoke firing of all three populations in the PT, but are most robust at TS and TA cells, indicating preferential drive of subcortical outputs. Therefore, mPFC inputs engage in both cortico-thalamo-cortical loops and cortico-thalamo-limbic broadcast circuits within the PT. This contrasts with higher-order nuclei like MD, which appear to primarily form cortico-thalamo-cortical loops that amplify and sustain cortical activity.^20,21 Therefore, our findings highlight how mPFC connections with different thalamic nuclei may perform distinct computational and functional roles.

Our transneuronal labeling experiments using AAV1 capitalized on our finding that mPFC axons are heavily localized to the PT over PVT. This viral strategy proved useful not only for isolating PT but also for labeling PT cells that project to the NAc and BLA. We observed that mPFC-recipient PT cells project densely to the core and medial shell of the NAc, as well as the medial wall of the BLA. These projections resemble those of PVT inputs to the NAc, which are not labeled using this transneuronal strategy. PT inputs display short-term depression, as seen in other TC,⁶⁷ TS,⁶⁸ and TA synapses,⁶⁹ and can drive firing in these areas. In each case, PT inputs also evoke feedforward inhibition, suggesting they recruit local interneurons. In future studies, it will be interesting to identify the principal cells and interneurons engaged by PT inputs in NAc and BLA.

Our anatomy and physiology data all suggest that L6 CT cells are the main input to TC, TS, and TA cells in the PT. In sensory systems, L6 CT cells in the primary cortex project back to first-order thalamus, where they make facilitating synapses on TC cells.⁷⁰ These connections have been labeled “modulators,” due to their weak influence on thalamic relay cells and relatively low firing rates in vivo.^63,71,72 While the activation of L6 CT cells in the mPFC are less explored, they are engaged in many behaviors.^6,10,23,73 Our results suggest that when L6 CT cells are activated in the IL, they will engage the PT, and subsequently the NAc and BLA. In the future, it will, therefore, be important to identify the behavioral and physiological states in which L6 CT cells are active, and how L6 CT inputs modulate the activity of PT circuits in vivo.

Our experiments suggest that the mPFC can access limbic structures through both direct and indirect routes. The mPFC directly projects to NAc and BLA via intratelencephalic pyramidal cells in layers 2/3 and 5.^56,74 In contrast, mPFC projects to PT via separate CT pyramidal cells located in layer 6. These different populations of cells have distinct genetic, physiological, and morphological properties.^75,76 By driving TS and TA cells in the PT, mPFC can broadcast signals to multiple subcortical structures. In the dorsal striatum, cortical and thalamic synapses display distinct properties at postsynaptic targets.⁶⁸ Similarly, in the amygdala, cortical and thalamic inputs undergo different forms of synaptic plasticity under stress and threat learning.⁷⁷ In the future, it will be interesting to compare the circuits engaged by direct cortico-limbic and indirect cortico-thalamo-limbic pathways in the NAc and BLA.

In summary, we identify the PT as a molecularly distinct nucleus in the aMT that participates in both reciprocal circuits with the mPFC and broadcast circuits with the NAc and BLA. The aMT contributes to a variety of motivated and emotional behaviors, feeding,¹ arousal,² salience encoding,³ reward learning,^4–6 aversive learning,^7–9 and social interactions.¹⁰ It is also implicated in many neuropsychiatric disorders, including stress,²⁷ pain,^28,29 and substance use disorders.^30–33 Together, our characterizations of cells, synapses, and circuits will provide a foundation for understanding the role of PT to these behaviors and associated disorders.

Limitations of the study

We identified PKCd as a molecular marker for TC cells in PT, allowing us to demarcate this nucleus from the adjacent PVT, but we were unable to identify markers for the TS and TA cells. Although our transneuronal labeling strategy using AAV1-Cre allowed us to label TS and TA cells in the PT, and their projections in the NAc and BLA, it also labeled TC cells and their projections to the IL. In the future, it will also be important to identify specific markers for TS and TA cells in aMT, which will enable characterization of their circuit properties, as well as their in vivo roles during behavior.

Our rabies-based tracing of monosynaptic inputs to PKCd+ PT cells identified many cortical and subcortical inputs to these cells (Figure 4). We characterized the synaptic physiology of IL mPFC inputs to the PT, but the sign, strength, and dynamics of subcortical inputs remain unexplored. In the future, it will be important to examine how subcortical areas, including the hypothalamus, lateral septum, and hippocampus, influence PT cells and their output to the mPFC, NAc, and BLA.

STAR★METHODS

EXPERIMENTAL MODEL AND SUBJECT PARTICIPANT DETAILS

All experimental procedures were approved by the New York University Animal Welfare Committee (protocol 07–1281). P28-P70 wild-type C57BL/6J (JAX #000664), Syt6-Cre (RRID: MMRRC_037416-UCD), Rbp4-Cre (RRID: MMRRC_037128-UCD), and PKCd-Cre (RRID: MMRRC_011559-UCD) mice of both sexes were used for experiments as noted. PKCd-Cre mice were generously provided by Eric Klann. All animals were held in reverse light cycle, and all recordings were made in the dark cycle. All experiments were replicated in at least 3 animals. No formal method for randomization was used, and experimenters were not blind to experimental groups. No pre-test analyses were used to estimate sample sizes. No data were excluded from final analyses.

METHOD DETAILS

Stereotaxic injections

Mice were anesthetized with isoflurane (1–2%) and head fixed in a stereotax (Kopf Instruments). A small craniotomy was made over the injection site, at the following coordinates relative to bregma (mediolateral, dorsoventral, anteroposterior): mPFC = ±0.35, −2.5, +2.1 mm; PT = ±1.5, −3.9, +0.4 mm at a 18° angle; NAc = ±1.8, −4.3, +1.4 mm at a 14° angle; BLA = ±3.1, −5.1, −1.9 mm. Borosilicate pipettes with 5–10 μm tip diameters were backfilled with mineral oil. Injection volumes of 100 nL (into PT) or 150–300 nL (into mPFC, NAc, and BLA) of virus were pressure-injected using a Nanoject III (Drummond) with 30–45 s inter-injection intervals. For retrograde labeling, mice were injected with either G-deleted rabies virus RV-CVS-N2c(ΔG)-H2B-GFP (UNC Vector Core, titer = 3.13E+08 GC/mL), AAVrg-CAG-H2B-tdTomato (UNC Vector Core, NT-231075.1, titer = 2.18E+13 GC/mL), Alexa 647-tagged cholera toxin subunit B (CTB647, Thermo Fisher, C34778) or a 1:1:1 mixture of rabies virus, AAVrg, and CTB. For tracing of monosynaptic inputs to PKCd+ cells in PT, PKCd-Cre mice were first injected with a helper virus AAV1-hSyn-DIO-TVA66T-dTom-CVS-N2cG (UNC Vector Core, titer = 2.81E+13 GC/mL) in PT. Three weeks later, pseudotyped rabies virus EnVA-CVS-N2c(ΔG)-H2B-GFP (UNC Vector Core, titer = 1.0E+9 GC/mL) was injected into the same location. Cre-dependent anterograde labeling was achieved using AAV1-EF1a-DIO-eYFP-WPRE.hGH (Addgene #27056-AAV1, titer = 2.5E+13 GC/mL) for anatomical tracing, and AAV1.EF1a.DIO.hChR2(H134R)-eYFP.WPRE.hGH (Addgene #20298-AAV1, titer = 2.3E+13 GC/mL) for optogenetic stimulation. For transneuronal labeling of PT, AAV1-hSyn-Cre-WPRE.hGH (Addgene #105553-AAV1, titer = 1.9E+13 GC/mL) was injected into mPFC, and AAV1.EF1a.DIO.hChR2(H134R)-eYFP.WPRE.hGH was injected into PT. Following injections, the pipette was left in place for an additional 5–10 min before being slowly withdrawn from the brain. Animals received post-op analgesics (ketoprofen, 5 mg/kg, 3 days) and returned to their home cages, where they remained for 2–4 weeks before experiments.

Slice preparation

Mice were deeply anesthetized with isoflurane and perfused intracardially with an ice-cold cutting solution containing (in mM): 65 sucrose, 76 NaCl, 25 NaHCO₃, 1.4 NaH₂PO₄, 25 glucose, 2.5 KCl, 7 MgCl₂, 0.4 Na-ascorbate, and 2 Na-pyruvate (bubbled with 95% O₂/5% CO₂). Coronal sections (300 μm thick) were cut in this solution and transferred to ACSF containing (in mM): 120 NaCl, 25 NaHCO₃, 1.4 NaH₂PO₄, 21 glucose, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 0.4 Na-ascorbate, and 2 Na-pyruvate (bubbled with 95% O₂/5% CO₂). Slices were recovered for 30 min at 35°C and stored for at least 30 min at room temperature. All recordings were performed at 30°C–32°C.

Slice electrophysiology

Whole-cell recordings were made from neurons using infrared-differential interference contrast microscopy. In the PT and PVT, retrogradely labeled neurons were identified by virally or CTB-expressed fluorophores. Sequential pairs of TC cells in PT vs. PVT, TC vs. TS cells in PT, or TC vs. TA cells in PT, were recorded in alternated orders to eliminate biases. For voltage-clamp experiments, borosilicate pipettes (3–5 MΩ) were filled with the following (in mM): 135 Cs-gluconate, 10 HEPES, 10 Na-phosphocreatine, 4 Mg₂-ATP, 0.4 NaGTP, 10 TEA, 2 QX-314, and 10 EGTA, pH 7.3 with CsOH (290–295 mOsm). For current-clamp recordings, borosilicate pipettes (3–5 MΩ) were filled with the following (in mM): 135 K-gluconate, 7 KCl, 10 HEPES, 10 Na-phosphocreatine, 4 Mg₂-ATP, 0.4 NaGTP, and 0.5 EGTA, pH 7.3 with KOH (290–295 mOsm). In some cases, 30 μM Alexa Fluor 594 (Thermo Fisher, A10438) was added to visualize morphology with two-photon microscopy. In all optogenetic stimulation experiments, 10 μM (R)-CPP was used to block NMDA receptors, in order to be able to detect the presence of polysynaptic inhibition. All reagents were purchased from Sigma or Tocris Bioscience.

Slice electrophysiology data were collected with a Multiclamp 700B amplifier (Axon Instruments) and National Instruments boards using custom software in MATLAB (MathWorks). Signals were sampled at 10 kHz and filtered at either 5 kHz for current-clamp recordings or 2 kHz for voltage-clamp recordings. Series resistance was 10–25 MΩ and not compensated.

Slice optogenetics

Glutamate release was triggered by activating channelrhodopsin-2 (ChR2) present in presynaptic terminals of cortical or thalamic axons.²¹ Channelrhodopsin-2 was activated using 2 ms pulses of 473 nm light from a blue light-emitting diode (LED; 473 nm; Thor-labs) delivered via a 10X 0.3 NA objective centered over the recorded neuron (Olympus), with a power range of 2–15 mW, routinely measured and calibrated at the back focal plane of the objective.

Two-photon microscopy

Two-photon imaging was performed using a custom microscope and software, as previously described.⁷⁸ A titanium-sapphire laser (Coherent) tuned to 810 nm was used to excite Alexa Fluor 594 and image dendrite morphology. z stack images were acquired with a 60× 1.0 NA objective (Olympus) and processed with NIH ImageJ and Neurolucida (MBF Bioscience).

Histology and immunohistochemistry

Mice were deeply anesthetized with isoflurane and perfused intracardially with 0.01 M PBS followed by 4% PFA. Brains were stored in 4% PFA overnight at 4°C before being washed 3x in 0.01 M PBS. Slices were cut on a VT-1000S vibratome (Leica) at 60 μm thickness. For immunohistochemistry, slices were washed once in PBS (0.01 M), once in PBS-T (0.2% Triton X-100). Non-specific binding was blocked with PBS-T with 1% w/v bovine serum albumin (BSA) and 2% normal goat serum (Abcam #ab7481) for 1 h at room temperature. Slices were incubated in a mixture of primary antibodies (mouse anti-calretinin, MA5–47457, Thermo Fisher, 1:2000; rabbit anti-protein kinase c delta, ab182126, Abcam, 1:2000) at 4°C overnight. After washing 4x in PBS, slices were incubated with secondary antibodies (goat anti-mouse IgG H&L Alexa 488-conjugated, A28175, Abcam, 1:500; goat anti-rabbit Alexa 647-conjugated, A21244, Abcam, 1:500) for 1 h at room temperature. Slices were washed a further 3x in PBS before being mounted on gel-coated glass slides and cover-slipped with VectaShield with DAPI (Vector Labs). Fluorescent images were taken with an Olympus VS120 microscope using 10X 0.25 NA objective (Olympus), or a Leica TCS SP8 confocal microscope using 20X 0.75 NA objective (Olympus).

Fluorescence in situ hybridization (RNAscope)

Mice were anesthetized with isoflurane and then perfused intracardially with chilled 0.01 M PBS. Brains were immediately submerged in cold isopentane on dry ice after dissection and stored in an airtight container at −80°C until sectioning. Sectioning was performed on a cryostat at −13°C, and 10 μm slices were mounted on Superfrost Plus microscope slides (Fisher Scientific) and stored at −80°C until staining. The standardized RNAscope protocol for flash-frozen tissue from ACDBio was followed, using Mm-nts (420441), Mmdrd2-C2 (406501-C2), Mm-calb2-C2 (313641-C2), Mm-ntrk1-C2 (435791-C2), and Mm-prkcd_C3 (441791-C3) probes. Slides were mounted using ProLong Gold antifade reagent with DAPI (Invitrogen) before being covered under coverslips.

Data analysis

Offline analysis of slice physiology recordings was performed using Igor Pro (WaveMetrics). For current-clamp recordings, resting membrane potential (Vrest) was measured at break-in. Input resistance (Rin) was measured using the steady-state response to 500 ms current steps of 25 or 50 pA. Voltage sags indicating h-current were calculated by taking the minimum voltage in the first 200 ms, subtracting the average voltage over the final 50 ms, and dividing by the steady-state value. Optically evoked firing probabilities were calculated as the numbers of the action potentials within a 100 ms window divided by the numbers of trials. For voltage-clamp recordings, PSC amplitudes were measured as the average value across 1 ms around the peak response.

Coronal brain sections were semi-automatically registered to the Allen Mouse Brain Common Coordinate Framework (CCF)³⁹ with NeuroInfo software (MBF Bioscience). Images were first automatically registered using the DAPI channel, and were then visually compared with anatomical landmarks, including corpus callosum, stria medullaris and third ventricles, as reference points. Retrogradely labeled, immunohistochemically labeled, or FISH-labeled cells were automatically counted using NeuroInfo software. For analysis of co-localization of PKCd and CALB2 in retrogradely-labeled neurons, cells were manually counted in ImageJ. Retrogradely labeled and immunohistochemically labeled cell quantification are presented as total cell counts across the anterior-posterior axis of the PT and adjacent PVT, while FISH-labeled cells are averaged cell counts from at least 3 images per animal. Thalamic axon distributions across the brains were quantified using NeuroInfo software. The average fluorescence intensity for each region was subtracted from within-slice background fluorescence and normalized to the peak fluorescence intensity across all measured areas.

For analysis of two-photon image stacks, all visible dendrites of each neuron were reconstructed in Neurolucida (MBF Bioscience). For Sholl analysis, 10 μm concentric circles were centered on the soma, and dendritic intersections and total dendritic lengths were calculated.

QUANTIFICATION AND STATISTICAL ANALYSIS

Sample sizes are reported within the Results and Supplemental Figure legends as the number of recorded cells (n) and animals (N) (for physiology) or number of animals (N) (for anatomy). Summary data are reported in the Results and figures as arithmetic mean ± SEM. Ratio data are displayed in figures on logarithmic axes and reported as geometric mean ±95% confidence interval (CI). In some graphs with three or more traces, SEM waves are omitted for clarity. Statistical analyses were performed using Prism 9.0 (GraphPad Software). Comparisons between unpaired data were performed using two-tailed Mann-Whitney tests. Comparisons between data recorded in pairs were performed using two-tailed Wilcoxon matched-pairs signed rank tests. Comparisons of more than two unpaired groups were performed using two-tailed Kruskall-Wallis tests with Dunn’s multiple comparisons post-hoc tests. Comparisons between groups across a range of variables (FI curves, PSC amplitudes and action potential probability per stimulation pulse, or cell types per area) were made using repeated-measures two-way ANOVA with Šídá k’s multiple comparison test. For all comparisons, no assumptions were made regarding data distributions, and significance was defined as p < 0.05.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2026.117865.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-Protein Kinase C Delta	Abcam	Cat #ab182126; RRID: AB_2892154
Mouse anti-Calretinin	Thermo Fisher	Cat #MA5-47457; RRID: AB_2942439
Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 647	Thermo Fisher	Cat #A-21244; RRID: AB_2535812
Goat anti-Mouse IgG (H + L) Superclonal™ Secondary Antibody, Alexa Fluor™ 488	Thermo Fisher	Cat #A-28175; RRID: AB_2536161
Bacterial and virus strains
AAV1-EF1a-DIO-hChR2-eYFP	Addgene	20298-AAV1
AAV1-hSyn-hChR2(H134R)-eYFP	Addgene	26973-AAV1
AAV1-hSyn-Cre-WPRE-hGh	Addgene	105553-AAV1
AAVrg-CAG-H2B-tdTomato	UNC Neurotools	NT-231075.1
AAV1-Syn-DIO-TVA66T-dTomato-CVS-N2cG	UNC Neurotools	N/A
Rbv-CVS-N2c(dG)-H2B-GFP (EnvA)	UNC Neurotools	N/A
Rbv-CVS-N2c(deltaG)-H2B-EGFP	UNC Neurotools	N/A
Chemicals, peptides, and recombinant proteins
Alexa Fluor 594	Thermo Fisher	Cat #10438
Cholera toxin subunit B – 647	Thermo Fisher	Cat #C34778
(R)-CPP	Tocris Bioscience	Cat #0247
TEA	Tocris Bioscience	Cat #3068
QX-314	Tocris Bioscience	Cat #2313
Prolong Gold with DAPI	Invitrogen	Cat #P36931
VectaShield with DAPI	Vector Labs	Cat #H-1200-10
RNAScope Multiplex Fluorescent Assay	ACDBio	Cat #323370
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Lab	RRID: IMSR_JAX:000664
Mouse: PKCd-Cre	MMRRC	RRID: MMRRC_011559-UCD
Mouse: Syt6-Cre	MMRRC	RRID: MMRRC_037416-UCD
Mouse: Rbp4-Cre	MMRRC	RRID: MMRRC_037128-UCD
Sequence-based Reagents
RNAscope Probe- Mm-Calb2 -C2	ACDBio	Cat #313641-C2
RNAscope Probe- Mm-Drd2-C2	ACDBio	Cat #406501-C2
RNAscope Probe- Mm-Ntrk1-C2	ACDBio	Cat #435791-C2
RNAscope Probe- Mm-Nts	ACDBio	Cat #420441
RNAscope Probe- Mm-Prkcd-C3	ACDBio	Cat #441791-C3
Software and Algorithms
Igor Pro 9	WaveMetrics	RRID: SCR_000325
ImageJ	NIH	RRID:SCR_003070
NeuroInfo	MBF Bioscience	RRID:SCR_017346
MATLAB	MathWorks	RRID: SCR_001622
Prism 11	GraphPad Software	RRID: SCR_002798

Highlights.

• Protein kinase C delta labels paratenial thalamus (PT) in the anterior midline thalamus

• Protein kinase C delta-positive thalamocortical cells project to the infralimbic cortex

• Infralimbic cortex contacts thalamocortical cells in the PT to form reciprocal loops

• Cortical inputs preferentially excite thalamostriatal and thalamoamygdala cells in the PT

RESOURCE AVAILABILITY

Lead contact

Further information and requests for materials and data should be directed to and will be fulfilled by the lead contact, Adam G. Carter (agc5@nyu.edu).

Materials availability

This study did not generate any new materials.

Data and code availability

• All electrophysiology data reported in this paper will be made available on request by the lead contact.

• This study did not generate any original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.