Long-range inhibition from prelimbic to cingulate areas of the medial prefrontal cortex enhances network activity and response execution

Abstract

It is well established that the medial prefrontal cortex (mPFC) exerts top-down control of many behaviors, but little is known regarding how cross-talk between distinct areas of the mPFC influences top-down signaling. We performed virus-mediated tracing and functional studies in male mice, homing in on GABAergic projections whose axons are located mainly in layer 1 and that connect two areas of the mPFC, namely the prelimbic area (PrL) with the cingulate area 1 and 2 (Cg1/2). We revealed the identity of the targeted neurons that comprise two distinct types of layer 1 GABAergic interneurons, namely single-bouquet cells (SBCs) and neurogliaform cells (NGFs), and propose that this connectivity links GABAergic projection neurons with cortical canonical circuits. In vitro electrophysiological and in vivo calcium imaging studies support the notion that the GABAergic projection neurons from the PrL to the Cg1/2 exert a crucial role in regulating the activity in the target area by disinhibiting layer 5 output neurons. Finally, we demonstrated that recruitment of these projections affects impulsivity and mechanical responsiveness, behaviors which are known to be modulated by Cg1/2 activity.

Fig. 1. PrL GABAergic projection neurons target mainly L1 in the Cg1/2.

a Schematic of the injection for anterograde tracing. AAV1-CAG-Flox-eGFP-WPR6 and AAV1-CAMKIIa-hChR2(h134a)-mCherry were injected into the PrL of GAD^Cre mice. b Schematic of the injection site into the PrL and c the projection site in the Cg1/2 in a coronal section. d Representative confocal image of Cre-dependent eGFP (green) and CAMKII-promoter dependent mCherry (magenta) expression at the injection site in the PrL in a DAPI stained (blue) coronal section. e Cre-dependent eGFP expression in GABAergic neurons. f mCherry expression in excitatory neurons. g Representative confocal image of a DAPI stained coronal section showing eGFP and mCherry at the projection site, i.e., the Cg1/2. h PrL GABAergic neuron-derived GFP expressing axons in the Cg1/2. i PrL excitatory neuron-derived mCherry-expressing axons in the Cg1/2. j–l Magnified view of the boxed area in (g). m Density of GABAergic axons (L1 vs. L2/3, p = 0.0166, L1 vs. L5, p = 0.0122, L1 vs. L6, p = 0.0297, DF = 8, two-sided Dunnett’s test) and n excitatory axons in the indicated layers of the Cg1/2 (L1 vs. L2/3, p = 0.974, L1 vs. L5, p = 0.996, L1 vs. L6, p = 0.932, DF = 8, two-sided Dunnett’s test). Open circles represent mean values from individual mice. o GABAergic and excitatory axons calculated as percentage of total axons in L1a (p = 0.000216, DF = 4, two-sided t-test). Red asterisks and n.s. above bars refer to a one-sample t-test of the percentage of GABAergic or excitatory neuron axons in L1a versus 50% (GABAergic, p = 0.00337, DF = 2, glutamatergic, p = 0.0783, DF = 2). d–o Images and data were obtained from nine slices from three mice. m–o Data are presented as mean ± S.E.M. Scale bar in (d–f), 500 μm, in (g–i), 200 μm, in (j–l), 50 μm. PrL prelimbic cortex, Cg1/2 cingulate area 1 and 2, eGFP enhanced green fluorescent protein, DAPI 4’,6-diamidino-2-phenylindole. n.s. not significant, *p < 0.05, **p < 0.01, ***p < 0.001, DF degree of freedom.

Fig. 2. PrL GABAergic projection neurons preferentially target L1 INs and disinhibit L5 PNs in the Cg1/2.

a Reconstructed pSBC (left) and pNGF (right) responsive to axonal stimulation of GABAergic projections. Somata and dendrites are indicated in black. Axonal arborization is indicated in green and blue. Normalized axonal density for pSBC (n = 6 cells, N = 3 mice) and pNGF (n = 7 cells, N = 3 mice) plotted against horizontal or vertical axes; measurements were performed in 10 μm intervals (black tick indicates location of the soma, horizontal axis: F(160,1760) = 2.673, vertical axis: F(160,1760) = 7.865, ***p < 0.0001, two-way RM ANOVA). Inset, typical firing patterns in response to current injections (−50, near threshold and 50 pA) and corresponding light-evoked IPSCs. Blue columns indicate laser illumination. Note that the light-evoked IPSCs are almost completely abolished (n = 5 cells, N = 3 mice) by the GABA_AR blocker SR95531 (SR, 1 μM). b Summary of IPSC peak amplitude (left; pNGF, n = 16 cells, pSGC, n = 11 cells) and connectivity (right). The number of connections per number of attempts for each cell type is indicated (N = 22 mice). c Experimental setting in a coronal section. Stimulating electrode is placed in L1. GABAergic axon stimulation occurs via blue light. The firing pattern belongs to a pSBC in L1, the approaching recording pipette (rec) is visible on the right. Pial surface is indicated as a dotted line. Inset, the firing patterns of the recorded pSBC. Scale bar, 40 mV, 0.25 s. d Cell-attached recording of the same pSBC shown in (c). Left and right, action currents induced by electrical stimulation (E-stim, triangle) only. Middle, optical stimulation (Opto, blue column) precedes electrical stimulation by 5 ms. e Summary of spiking probability of pSBC (n = 7 cells, N = 4 mice, p = 0.0025) and pNGF (n = 5 cells, N = 3 mice, p = 0.0342) before, during and after optical stimulations. **p < 0.005, *p < 0.05, Friedman test followed by post hoc two-sided Dunn’s test. f Experimental setting for whole-cell recording of L5 PNs. Stimulating electrode is placed in L1, GABAergic axons are activated by blue light. Inset, the firing pattern of a recorded PN. Scale bar, 40 mV, 0.25 s. g, h Electrically evoked responses from the same PN shown in (f). Top, compound PSCs (including EPSC-IPSC sequences and putative pure EPSCs) recorded at −50 mV. Individual traces are superimposed and shown in gray. Averaged traces are shown in black. Two EPSC-IPSC sequences are highlighted in orange. Shaded areas indicate the inward charge. Note that the stimulation not always elicited EPSC-IPSC sequences as the spike probability of L1 INs in (d, e) was always below 100%. Middle, EPSCs recorded at −92 mV (E_GABA). Bottom, feedforward (FF) IPSCs isolated by subtraction (see “Methods”). The onset of the EPSC and of the FF IPSC are indicated in red and black dotted line, respectively. In (h), optical stimulation (blue column) precedes electrical stimulation (triangle) by 5 ms. Stimulation artifacts were removed for clarity sake. i Summary of inward charge recorded at near −50 (blue, n = 12 cells, p = 0.0015) or −92 mV (E_GABA, red, n = 10 cells, p = 0.1055) with and without optical stimulation (N = 4 mice). **p < 0.005, two-sided Wilcoxon signed rank test. j Summary of normalized inward charge recorded at near −50 (blue) or −92 mV (red) (n = 10 cells, N = 4 mice, p = 0.0039, **p < 0.005, two-sided Wilcoxon signed rank test). b, e, i, j Data are presented as means ± S.E.M. pSBC putative single-bouquet cell, pNGF putative neurogliaform cell, pSOM⁺ putative somatostatin-expressing cell, pPV⁺ putative parvalbumin-expressing cell, p5HT₃R⁺ putative serotonin 3A receptor-expressing cell, PN pyramidal neuron, IPSC inhibitory postsynaptic current, EPSC excitatory postsynaptic current.

Fig. 3. Optogenetic activation of of PrL GABAergic projections in the Cg1/2 increases impulsivity.

a AAV-DIO-ChR2-mCherry (ChR2 group, N = 9 mice) or AAV-DIO-YFP (YFP control group, N = 10 mice) was injected into the PrL of GAD^Cre mice and an optic fiber was implanted into the Cg1/2. b Schematic of the operant task. Mice were trained in an operant task where a briefly illuminated [3 s] bright light signaled that pressing a lever would lead to delivery of a reward. The illumination of the bright light was preceded by a dim light for an unpredictable duration [5–8 s] during which time lever pressing was punished with darkness [10–20 s]. Therefore, the mice had to attend to the light and, in order to earn a food pellet reward, rapidly respond when the intensity increased. The attentional demands of the task could be manipulated by changing the intensity of the dim light. Thus, in the low demand version of the task there was a greater difference between the dim and bright light states compared to the high demand version of the task. The task has four response types: Correct responses = lever pressing during the bright light state; premature responses = lever pressing during the dim light state; omissions = failure to lever press during either the dim or bright light states and dark = pressing when the stimuli light was not illuminated. Once trained to ~60% correct responses, mice were tested in a sham session (Sham) where no light was delivered into the Cg1/2 and a stimulation session (Stim) where blue light pulses (10-ms pulses at 20 Hz, 473-nm wavelength) were delivered during the entire session. c Correct responses during sham and stimulation conditions. Stimulation has no effect during the low demand task (top, all p > 0.05), but significantly reduces correct responses during the high demand task (bottom, group × demand interaction F(3,49) = 5.237, p = 0.0032) with YFP vs. ChR2 in the high demand task with stimulation p = 0.0035). d Premature responses during sham and stimulation conditions. Stimulation has no effect during the low demand task (top, p > 0.05), but significantly increases premature responses during the high demand task (group × demand interaction F(3,49) = 5.489, p = 0.0025, with YFP vs. ChR2 in the high demand task with stimulation p = 0.0085) indicating increased impulsivity. e Stimulation has no significant effect on omissions during either the low or high demand versions of the task (all p > 0.05). All data were analyzed with Geisser-Greenhouse corrected RM ANOVA followed by two-sided multiple comparisons using the two-stage method of Benjamini, Krieger and Yekutieli. Lines on graphs show group means ± S.E.M., dots (YFP) and squares (ChR2) show individual data points. **p ≤ 0.01).

Fig. 4. Optogenetic activation of PrL GABAergic projections in the Cg1/2 increases the response to mechanical stimulation.

a Mice were injected with virus (AAV-DIO-ChR2-mCherry, N = 9 mice, AAV-DIO-YFP, N = 9 mice) into the PrL and an optic fiber was implanted into the Cg1/2. They were subsequently tested in a mechanical stimulation assay using von Frey filaments (session 1, pre-optical stimulation; session 2, peri-optical stimulation; session 3, post-optical stimulation) each with 6 different vF filament forces. b There is no difference in basal mechanical sensitivity between ChR2 and YFP groups prior to optical stimulation (all p > 0.05). c During optical stimulation of PrL GABAergic axon terminals in the Cg1/2 (10-ms pulses at 20 Hz, 473-nm wavelength), there is a group v force interaction (F(5,80) = 3.57, p = 0.0058) and an increased sensitivity to the 0.16-g (p < 0.0001) and 0.4-g filaments (p = 0.0003). d Following optical stimulation, there is again no difference between the ChR2 and YFP groups in response to vF stimulation (all p > 0.05). All data were analyzed by RM ANOVA followed by two-sided Bonferroni corrected multiple comparisons. Lines on graphs show group means ± S.E.M., ***p ≤ 0.001.

Fig. 5. Optogenetic activation of PrL GABAergic projections in the Cg1/2 enhances vF-evoked responses in L5.

a Schematic of the approach for Ca²⁺ imaging of Cg1/2 neurons and optogenetic stimulation of axon terminals of PrL GABAergic projection neurons in head-fixed GAD^Cre mice that received GRIN lens and optic fiber implantation. The animals were placed on a grid and received von Frey (vF) stimulation to the hind paw contralateral to the implantation site. b AAV-FLEX-tdTomato (control, N = 13 mice) or AAV-Syn-FLEX-ChrimsonR-tdTomato (ChrimsonR, N = 15 mice) was injected into the PrL, and AAV-CaMKIIa-GCaMP6s was injected into the Cg1/2 of GAD^Cre mice. c Virus-induced expression of tdTomato (control) or tdTomato-tagged ChrimsonR in GABAergic cells of the PrL. Virus-induced expression of GCaMP6s in excitatory cells, implantation of GRIN lens and illumination with red light from an optic fiber tip implanted in the Cg1/2. d Representative field of view through the implanted GRIN lens using the two-photon microscope showing GCaMP6 expressing neurons. Scale bar, 50 μm. e Schematic of the procedure used for Ca²⁺ imaging experiments, indicating stimulation details and temporal profile of vF and red-light stimulation in the three sessions. f Average neuronal responses to vF stimuli in Cg1/2 neurons during session 1 (stippled line), 2 (solid line) and 3 (dotted line) in control mice expressing only td-Tomato (N = 13 mice, n = 307 cells) and in ChrimsonR-expressing mice (N = 15 mice, n = 397 cells). ΔF/F0 traces were smoothed with a Gaussian filter (s = 100 ms) for visualization purpose. Data are shown as means ± S.E.M. Black vertical line indicates the onset of the vF stimulus. Black stippled vertical line indicates the end of the stimulus. g Area under the curves (AUCs) of the vF-evoked responses in control (p = 0.00872, Friedman test, session 1 vs. 2, p = 0.00239, session 1 vs. 3, p = 0.533, DF = 306, two-sided Wilcoxon signed-rank test followed by Bonferroni correction) and h in ChrimsonR mice (p = 6.05e−05, Friedman test, session 1 vs. 2, p = 1.07e−05, session 1 vs. 3, p = 0.00348, DF = 396, two-sided Wilcoxon signed-rank test followed by Bonferroni correction). i Fold change of AUC for session 2 and 3 relative to session 1 in all layers (session 2, p = 3.94e−07, session 3, p = 0.00378, DF = 306, two-sided Mann–Whitney U test followed by Bonferroni correction), j Fold change of AUC for session 2 and 3 relative to session 1 in L2/3 (control, n = 38 neurons, ChrimsonR, n = 94 cells, session 2, p = 1.00, session 3, p = 0.132, DF = 37, two-sided Mann–Whitney U test followed by Bonferroni correction) and k Fold change of AUC for session 2 and 3 relative to session 1 in L5 neurons (control, n = 127 cells, ChrimsonR, n = 112 neurons, session 2, p = 2.44e−06, session 3, p = 0.0538, DF = 111, two-sided Mann–Whitney U test followed by Bonferroni correction). Box and whisker plots in (g–k) indicate median, interquartile range and 10th to 90th percentiles of the distribution. n.s. not significant, *p < 0.05, **p < 0.01, ***p < 0.001, DF degree of freedom.

Fig. 6. Optogenetic activation of PrL GABAergic projections differentially modulates responses of Cg1/2 L5 neurons during and after vF stimulation.

a Schematic of k-means clustering, illustrating data transformation of vF-evoked responses. Example of a neuron which became activated upon vF stimulation. Neuronal responses to vF stimulation during session 1–3 are binned and then concatenated before k-means clustering is performed. b Heatmap indicating mean vF-evoked responses from representative L5 neurons of control and c Chrimson mice in each session. Data are assigned to two ensembles based on k-mean clustering. White vertical line and stippled line indicate onset and end of the vF stimulus. The responses were sorted based on the maximum value during vF stimulation in session 1. d Average neuronal responses to vF stimuli of L5 vF-activated neurons during session 1 (stippled line), 2 (solid line) and 3 (dotted line) in control (n = 81 cells) and in ChrimsonR mice (n = 81 neurons). Gray vertical and stippled lines indicate onset and end of vF stimulus, respectively. e Mean ΔF/F0 of L5 vF-activated neurons during (0–1 s) and post (1–5 s) vF stimulation in control (open box, during vF stimulation in session 1 vs. 2, p = 1.28e−06, session 1 vs. 3, p = 1.00, post vF stimulation in session 1 vs. 2, p = 5.04e−05, session 1 vs. 3, p = 1.00, DF = 80, two-sided Wilcoxon signed-rank test followed by Bonferroni correction) and f in ChrimsonR-expressing mice (filled box, during vF stimulation in session 1 vs. 2, p = 0.00167, session 1 vs. 3, p = 0.179, post vF stimulation in session 1 vs. 2, p = 0.0449, session 1 vs. 3, p = 0.0826, DF = 80, two-sided Wilcoxon signed-rank test followed by Bonferroni correction). g Fold change of ΔF/F0 means for session 2 and 3 relative to session 1 during and post vF stimulation in control (open box) and ChrimsonR mice (filled box, during vF stimulation in session 2, p = 5.58e−10, session 3, p = 0.238, post vF stimulation in session 2, p = 1.24e−05, session 3, p = 0.139, DF = 80, two-sided Mann–Whitney U test followed by Bonferroni correction). h Average neuronal responses to vF stimuli of L5 vF-inactivated neurons during session 1 (stippled line), 2 (solid line) and 3 (dotted line) in control (n = 46 neurons) and in ChrimsonR mice (n = 31 neurons). Gray vertical and stippled vertical lines indicate onset and end of vF stimulus, respectively. i Mean ΔF/F0 of L5 vF-inactivated neurons during (0–1 s) and post (1–5 s) vF stimulation in control (open box, during vF stimulation in session 1 vs. 2, p = 1.00, session 1 vs. 3, p = 0.438, post vF stimulation in session 1 vs. 2, p = 1.00, session 1 vs. 3, p = 0.668, DF = 45, two-sided Wilcoxon signed-rank test followed by Bonferroni correction) and j in ChrimsonR-expressing mice (filled box, during vF stimulation in session 1 vs. 2, p = 1.00, session 1 vs. 3, p = 0.0243, post vF stimulation in session 1 vs. 2, p = 1.00, session 1 vs. 3, p = 1.00, DF = 30, two-sided Wilcoxon signed-rank test followed by Bonferroni correction). k Fold change of ΔF/F0 means for session 2 and 3 relative to session 1 during and post vF stimulation in control (open box) and ChrimsonR mice (filled box, during vF stimulation in session 2, p = 1.00, session 3, p = 0.820, control vs. ChrimsonR, post vF stimulation in session 2, p = 1.00, session 3, p = 1.00, DF = 30, two-sided Mann–Whitney U test followed by Bonferroni correction). d, h ΔF/F0 traces were smoothed with a Gaussian filter (s = 100 ms) for visualization purpose. Data are shown as means ± S.E.M. e–g, i–k Box and whisker plots indicate median, interquartile range and 10th to 90th percentiles of the distribution. n.s. not significant, *p < 0.05, **p < 0.01, ***p < 0.001, DF degree of freedom.