A brain circuit controlling daily rhythms of stress hormone release

To evaluate whether the SPZ GABAergic neurons participate in regulating circadian Cort release, we placed injections in the SPZ of Vgat-ires-Cre::L10 mice of a viral vector that constitutively expresses mCherry in neurons lacking Cre-recombinase (Cre), but in cells that express Cre instead produces diphtheria toxin A (DTA), which induces cell death (AAV10-hSyn-mCherry-DIO-DTA). We placed injections of an AAV-EGFP virus (AAV8-DIO-EGFP) in the same site in control mice. After four weeks to allow the full expression of DTA or EGFP, we measured Cort levels in blood from DTA injected and control mice in a 12:12 light:dark (LD) cycle, by tail nicks at four time points (ZT 1, 7, 13, 19, taken in random order with at least 30 hrs between samplings), and at the same time points in constant darkness (DD) (CT 1, 7, 13, 19), starting on the third day in DD at CT13, and then every 30 h. Simultaneously, we recorded the circadian rhythms of locomotor activity (LMA) and body temperature (Tb) for 15 days in LD, and then 15 days under DD. From a total of 12 mice injected with DTA, 8 mice were identified by a blinded observer with bilateral DTA injection covering at least 70% of the SPZ; in mice that also expressed L10-GFP in the Vgat neurons, there was approximately 75% cell loss in the SPZ (Supplementary Fig. 1). Because some injections encroached on the borders of the SCN, we also did cell counts within the SCN in these mice and found less than 15% cell loss (Fig. 1a, b and Supplementary Fig. 1a). The mice with injections that missed the target were considered anatomical controls. The ablation of the SPZ^Vgat neurons reduced the Circadian Index (CI, comparing the levels during the light vs the dark period; for method of calculation of CI see Statistical Analysis in the Materials and Methods) of Cort secretion compared with the control animals by 66.6 ± 9.1% in LD (p = 0.008) and 97.7 ± 5.9% in DD (p < 0.001). The peak Cort level was reduced from control levels of 23.4 ± 1.7 ng/ml in LD and 28.1 ± 3.9 ng/ml in DD, to 12.8 ± 3.0 ng/ml in LD and 4.8 ± 1.1 ng/ml in DD in SPZ^Vgat ablated mice; (Fig. 1c–e). The SPZ^Vgat ablations also caused a dramatic reduction in the CI of locomotor activity (LMA) by 51.1 ± 3.5% in LD (p < 0.001) and by 79.7 ± 4.8% in DD (p < 0.001), mostly by reducing the general LMA during the (presumptive) dark period (Supplementary Fig. 1b–i). The CI of body temperature (Tb) was reduced by 50.42 ± 5.6% in LD (p < 0.001) and 58.34 ± 6.5% in DD (p < 0.001; Supplementary Fig. 1j–q). The anatomical controls (missed injections) did not show any changes in LMA or Tb.

To evaluate whether the temporal information was transmitted by GABA release from the SPZ^Vgat neurons, we placed injections in the SPZ of AAV8-SYN-EGFP-iCre (AAV8-EGFP-iCre) or a control virus (AAV8-DIO-EGFP) in Vgat^loxP/loxP (Vgat-flox) mice to delete expression of functional Vgat protein and so to abolish GABA transmission from the transfected neurons. From 12 injected mice, 7 received a bilateral injection of AAV8-EGFP-iCre covering at least 70% of the SPZ without extending into the SCN (Fig. 1f, g and Supplemenary Fig. 2a). Surprisingly, deletion of Vgat from the SPZ neurons did not affect the circadian release of Cort (Fig. 1h–j). By contrast, the circadian rhythms of LMA and Tb were reduced by the deletion of Vgat from SPZ^Vgat cells, although to a lesser degree than after ablation of those neurons. For instance, the CI of LMA was reduced by 39.75 ± 8.9% in DD (p < 0.001; Supplementary Fig. 2b–i), and the CI of Tb was reduced by 25.7 ± 4.6% in LD (p = 0.006) and 38.6 ± 7.3% in DD (p < 0.001; Supplementary Fig. 2j–q). Taken together, these data indicate that the SPZ^Vgat neurons are necessary to transmit temporal information for circadian regulation of Cort, LMA and Tb from the SCN, but that the release of GABA alone cannot account for their influence on circadian regulation.

DMH neurons are downstream of the SPZ and have been shown in rats to be necessary for circadian rhythms of Cort. As DMH ablations cause a reduction in overall Cort secretion, we hypothesized that circadian Cort secretion may be under the control of excitatory glutamatergic DMH neurons. We therefore next examined the role of the DMH^Vglut2 neurons in the circadian regulation of Cort. We first measured Cort levels in blood from 14 Vglut2-ires-Cre mice in a 12:12 light:dark (LD) cycle, at the same four time points (ZT 1, 7, 13, 19), and at the same time points in constant darkness (DD) (CT 1, 7, 13, 19). Simultaneously, we recorded the baseline circadian rhythms of LMA and Tb for 12 days in LD, and then 12 days under DD. We then placed injections into the DMH of AAV10-hSyn-mCherry-DIO-DTA, and after four weeks to allow full expression of DTA, we measured Cort levels in LD and DD as described, and recorded LMA and Tb in the mice for 12 days in LD and 12 days in DD. From a total of 14 mice, we identified 8 mice in which the injection covered at least 70% of the DMH bilaterally (Fig. 2a, b and Supplementary Fig. 3a) and these mice were used for further analysis. DMH^Vglut2 ablation reduced the Cort peak at the beginning of the active phase (ZT 13) from 39.4 ± 6 ng/ml in LD and 27.9 ± 3.5 ng/ml in DD before AAV10-hSyn-mCherry-DIO-DTA injections, to 23.4 ± 4.2 ng/ml in LD (p = 0.007, Fig. 2c) and 15.5 ± 4.3 ng/ml in DD after DMH^Vglut2 neuron ablation (p = 0.013, Fig. 2d). The CI of Cort secretion was reduced by 46.6 ± 14.2% in LD (p = 0.043) and 62.0 ± 17.3% in DD after the ablation (p = 0.014, Fig. 2e). To determine the specificity of the DMH^Vglut2 ablations for circadian increases in Cort, we measured the Cort response to stress of the mice with the ablation by subjecting them to 1 h of restraint at ZT3. However, we did not detect changes in the Cort levels induced by restraint stress, indicating that there was no impairment in overall Cort secretion and that DMH^Vglut2 neurons do not play a role in restraint stress-induced Cort secretion (Supplementary Fig. 3b). The ablation of DMH^Vglut2 neurons also reduced the peak in LMA at the transitions between active and inactive periods both in LD and DD. Because these transition periods spanned the light and dark cycles, there was no change in the CI of LMA, although the amplitude of the LMA rhythm in DD was reduced in cosinor analysis (Supplementary Fig. 3c–j). In contrast, ablation of DMH^Vglut2 neurons reduced Tb during the light phase in LD and in both presumptive light and dark phases in DD. As a result, the CI and cosinor amplitude of the Tb rhythm was increased during LD but not DD (Supplementary Fig. 3k–r).

We then evaluated whether the effect of ablation of DMH^Vgut2 neurons on Cort rhythms was due to the loss of glutamate transmission, as opposed to other possible transmitters expressed in DMH^Vgut2 neurons. We therefore injected the DMH of Vglut2^loxP/loxP (Vglut2-flox) mice with either AAV8-EGFP-iCre or a control virus AAV8-DIO-EGFP and recorded LMA, Tb and Cort levels in both LD and DD. Seven of the 11 mice had injections covering at least 70% of the DMH bilaterally and were compared with 5 mice with AAV8-DIO-EGFP injections in the DMH of Vglut2-flox mice as controls (Fig. 2f, g, Supplementary Fig. 4a). In these animals, the Cort levels at ZT13 were reduced in LD from 28.9 ± 1.9 ng/ml to 13.2 ± 2.6 ng/ml (p < 0.001, Fig. 2h) and in DD from 40.4 ± 7.5 ng/ml to 11.7 ± 3.5 ng/ml in DMH^Vglut2/flox-Control and DMH^Vglut2/flox-EGFP-iCre mice, respectively (p < 0.001, Fig. 2i). The CI of Cort was reduced by 68.2 ± 10.1% in LD (p < 0.001) and by 102.1 ± 14.6% in DD (p = 0.001, Fig. 2j). Both the LMA and Tb were reduced during the dark and subjective dark phase, and at the transition from the dark to the light phase during LD and at the presumptive transition in DD (Supplementary Fig. 4). However, the CI of neither LMA nor Tb was affected by the loss of glutamatergic signaling in the DMH.

As the DMH^Vglut2 ablation and the Vglut2 mRNA deletion are chronic lesions that might be affected by compensatory mechanisms, we decided to test the effect of acute inhibition of the DMH^Vglut2 neurons on Cort, LMA and Tb rhythms. We injected the DMH in 9 Vglut2-ires-Cre mice with a viral vector containing a modified human alpha-1 glycine receptor (AAV10-DIO-hGlyR-mCherry) which is insensitive to glycine but has 100-fold increased sensitivity to the antiparasitic drug ivermectin (IVM) compared to the unmutated glycine receptor¹⁶. Because the half-life of IVM is about 72 hrs, the injection of IVM induces long-lasting neuronal inhibition (4–5 d). To allow the drug to achieve a stable concentration in the CNS, we examined Cort levels between 24 and 48 hrs after drug administration¹⁷ compared with injection of vehicle. We administered vehicle (VEH) or IVM (5 mg/kg, ip) one hour after the light or subjective light onset and sampled the mice the next day at ZT1 and ZT13 to evaluate the Cort levels; one week later, they received the opposite treatment. In 5 mice in which expression of mCherry was seen in at least 70% of the DMH bilaterally (Fig. 2k, l, Supplementary Fig. 5a), the Cort levels at ZT13 under LD were reduced from 31.7 ± 2.3 ng/ml after VEH to 15.6 ± 4.8 ng/ml after IVM (p = 0.006), and under DD from 28.1 ± 6.9 ng/ml after VEH to 9.3 ± 1.8 ng/ml after IVM (p = 0.001, see Fig. 2m). The inhibition of the DMH^Vglut2 neurons with IVM reduced the Cort CI by 58.9 ± 15.9% in LD (p = 0.01) and by 92.3 ± 9.9% in DD (p = 0.019, see Fig. 2n). The average CI of LMA after IVM administration was reduced by 103.3 ± 21.1% in LD (i.e., the day-night difference was reversed due to higher levels of LMA during the light period) (p = 0.011, Supplementary Fig. 5b–d), and was reduced in DD by 81.3 ± 9.1% when compared with vehicle (p = 0.016, Supplementary Fig. 5e–g) mainly due to the shifting of much of the LMA peak from the early dark period to the light period, however, the total LMA was not significantly different at 24–48 h after IVM both in DD and LD (Supplementary Fig. 5b–g). IVM also induced a decrease in Tb during the dark period, with a reduction of 0.5 ± 0.2 °C observed in LD compared with vehicle administration (p = 0.024, Supplementary Fig. 5h–j). In DD conditions, the decrease in Tb during presumptive night was 0.4 ± 0.1 °C (p = 0.014, Supplementary Fig. 5k–m). Tb and LMA returned to normal around day 5 after IVM either in LD or DD. The Tb CI was reduced in LD by 38.4 ± 6.2% after administration of IVM (p = 0.025, Supplementary Fig. 5j). In all the 3 experiments (DTA, iCre and hGlyR), the anatomical controls (animals with injections that missed the target) failed to show the responses observed in LMA or Tb.

These data clearly show that glutamatergic transmission by the DMH^Vglut2 neurons plays an important role in the increase of Cort levels at the beginning of the active phase. The ablation or inhibition of DMH^Vglut2 neurons reduces the peak of Cort to the early morning levels. The DMH^Vglut2 neurons also increase LMA in a crepuscular pattern (at the transitions between the light and dark periods) and their loss causes a small (approximately 0.3–0.5 °C) decrease in Tb, but has little if any effect on the circadian rhythm of Tb.

As the ablation of DMH^Vglut2 neurons reduced the circadian peak of Cort levels, we hypothesized that activation of these neurons may elevate the Cort levels in preparation for the active phase. First, we injected the DMH of 15 Vglut2-ires-Cre mice with AAV10-EF1α-DIO-hM3Dq-mCherry (Fig. 3a, b). Five to six weeks later, the mice were injected i.p. with clozapine-N-oxide (CNO) 0.3 mg/kg at ZT3. The acute activation of DMH^Vglut2 neurons caused an increase in Cort levels from 7.4 ± 3.7 ng/ml just before the animals received the dose of CNO, to 162.5 ± 16.4 ng/ml 1 h after the administration of CNO (p < 0.001, see Fig. 3c). This increase is around three times higher than the Cort levels detected in the same mice that received saline or in WT mice injected with CNO. These results confirm that activation of the DMH^Vglut2 neurons boosts Cort release.

To determine whether the DMH^Vglut2 neurons make direct synaptic contacts onto the PVH^CRH neurons to activate them and so to increase Cort levels, we conducted conditional anterograde tracing and conditional monosynaptic retrograde rabies tracing studies. We first evaluated the projection pattern of DMH^Vglut2 neurons using Vglut2-ires-Cre mice crossed with reporter mice that express the fluorescent protein Venus (green) in the CRH expressing neurons (Vglut2-ires-Cre::CRH-Venus). Following injection of AAV8-DIO-ChR2-mCherry into the DMH, we found a dense projection from DMH^Vglut2 neurons to the PVH that formed appositions with the PVH^CRH neurons (Fig. 3d, e). To identify the locations of DMH^Vglut2 neurons potentially forming direct connections with the PVH^CRH cells, we injected the PVH of 5 CRH-ires-Cre mice with an AAV expressing TVA avian receptor (AAV8-Ef1a-DIO-TVA-mCherry) and an AAV expressing rabies glycoprotein necessary for rabies viral transfection (AAV8-CAG-DIO-rabiesG) specifically in the PVH^CRH neurons. Twenty-one days later, to label and map the direct monosynaptic inputs to the PVH^CRH neurons, we injected a glycoprotein (G)-deleted rabies virus (RVdG) that expresses EGFP and is enveloped with the avian ASLV type A protein (EnvA), which utilizes the TVA receptor for cell entry (EnvA-ΔG-rabies-EGFP). This results in PVH^CRH neurons infected by both AAV and rabies viruses (starter cells) displaying both green and red fluorescent signals, while neurons expressing only EGFP-Rabies (green) were retrogradely labeled by the viral particles produced in the starter cells (Supplementary Fig. 6a, b). Transneuronally infected neurons were seen in areas previously reported to project to the PVH such as the bed nucleus of the stria terminalis, preoptic area, arcuate nucleus, and DMH, and we also observed a large number of retrogradely labeled neurons within the PVH itself and its immediate surrounding areas (Supplementary Fig. 6c, d)¹⁸. We found few, if any, EGFP-Rabies infected neurons in the SCN, which supports the hypothesis that circadian control of Cort secretion by the SCN requires intermediate relays such as the SPZ and DMH.

We also combined rabies tracing from PVH^CRH neurons with in situ hybridization for Vglut2 mRNA. We found about half of retrogradely-EGFP labeled neurons in the DMH expressed Vglut2, confirming that DMH^Vglut2 neurons are likely to directly innervate PVH^CRH cells. In particular, we observe that the largest number of retrogradely labeled neurons expressing Vglut2 were located in the anterior portion of the DMH (Fig. 3f, g). By contrast, retrogradely labeled neurons that expressed Vgat mRNA were found throughout the DMH, but with a more caudal predominance (Supplementary Fig. 6e, f), indicating that there is also a direct GABAergic input predominantly from the caudal DMH to the PVH^CRH neurons (see below).

To test the input from the DMH^Vglut2 neurons to the PVH^CRH neurons (DMH^Vglut2 → PVH^CRH) functionally, we conducted in vitro channelrhodopsin-2(ChR2)-assisted circuit mapping (CRACM) recordings. We injected the DMH with an AAV8-DIO-ChR2-mCherry in a new cohort of 4 Vglut2-ires-Cre::CRH-Venus mice, and then three to four weeks later we recorded from Venus-labeled PVH^CRH neurons in brain slices while photo-stimulating ipsilateral axons and terminals of DMH^Vglut2 neurons within the PVH (Fig. 4a, b). All recordings were performed in the PVH ipsilateral to the injection site during the daytime (ZT3-8), at the nadir of the daily Cort cycle¹⁹. Optogenetic stimulation of the DMH^Vglut2 input evoked excitatory synaptic responses in 72% of the PVH^CRH neurons recorded (n = 18). This effect was mediated by the release of glutamate and AMPA receptor signaling as the opto-evoked excitatory postsynaptic currents (oEPSCs) were blocked by the AMPA receptor antagonist DNQX (n = 4; Fig. 4c–f). Furthermore, the probability of oEPSCs in the PVH^CRH cells peaked about 5 msec after the photostimulation of the DMH^Vglut2 terminals and the oEPSCs persisted in the presence of TTX (TTX 1 µM) and 4-aminopyridine (4-AP), 200–500 µM) indicating monosynaptic connectivity (n = 6 out of 8; Fig. 4f, g). These experiments demonstrate that DMH^Vglu2 neurons have robust direct synaptic input to PVH^CRH neurons (Fig. 4h, i) that is consistent with DMH^Vglut2 neurons driving the elevation of Cort levels during the early active period (Fig. 2).

Although inputs from DMH^Vglut2 neurons to PVH^CRH cells are necessary to drive the daily Cort increase, their ablation does not completely eliminate the circadian rhythm of Cort secretion, as is seen with non-specific DMH ablation¹². It has been estimated that half of the synapses to the PVH^CRH neurons are GABAergic²⁰, and our rabies virus experiment showed that about half of the DMH cells innervating the PVH^CRH neurons are GABAergic. Thus, we decided to evaluate whether the DMH^Vgat neurons could play a complementary role in the circadian release of Cort. We hypothesized that the DMH^Vgat neurons could contribute to the circadian rhythms of Cort secretion either by inhibiting PVH^CRH neurons during the inactive cycle to suppress Cort secretion, or by dis-inhibiting PVH^CRH neurons (i.e., inhibiting inhibitory inputs to them) during the active cycle. Thus, we used Vgat-ires-Cre mice crossed with a Cre-dependent GFP reporter mouse line R26-loxSTOPlox-L10-GFP to generate Vgat-ires-Cre::R26-loxSTOPlox-L10-GFP mice (Vgat-ires-Cre::L10), in which Vgat neurons expressing Cre-recombinase show green fluorescence. We collect blood samples at 4 temporal points in LD (ZT 1, 7, 13, 19), and then in the same temporal points in DD (CT 1, 7, 13, 19), as we did with the Vglut2-ires-Cre mice. Also, we recorded LMA and Tb for 12 days in LD, and then 12 days in DD. Then, we injected the DMH of Vgat-ires-Cre::L10 mice with an AAV10-hSyn-mCherry-DIO-DTA to ablate the DMH^Vgat neurons. Four weeks after the injections, we recorded and took blood samples from the mice in LD and DD conditions. From a total of 13 Vgat-ires-Cre::L10 mice injected, we analyzed data from 8 mice in which a blinded investigator found that the injection sites covered at least 70% of the DMH bilaterally (Fig. 5a, b, Supplementary Fig. 7a, b). DMH^Vgat ablation did not alter the daily rhythm of Cort release in LD but reduced the CT13 peak under DD from 26.5 ± 2.9 ng/ml in Ctrl to 14.1 ± 4.4 ng/ml (p = 0.014, Fig. 5c, d). Similarly, there was no statistically significant change in the CI of Cort in LD, but it was reduced by 72.4 ± 21.3% in DD after ablation of DMH^Vgat neurons (p = 0.01, Fig. 5e). Importantly, ablation of DMH^Vgat neurons did not change the levels of Cort at its nadir (ZT1 or CT1), which might have been expected if DMH^Vgat neurons with direct inputs to PVH^CRH cells contributed to circadian rhythms of Cort by suppressing secretion during the daily nadir. To test the effect of ablation of DMH^Vgat neurons on other aspects of Cort secretion, we also measured Cort levels after one hour of restraint stress, but these were unaffected by the ablations (Supplementary Fig. 7c). In contrast to the effects on Cort rhythms, LMA was dramatically reduced during the dark phase of LD after DMH^Vgat ablation, the total count over the dark period decreasing from 873.5 ± 45 counts to 429.5 ± 52.5 counts (p < 0.001); LMA was reduced even further in DD during the presumptive dark phase from 771.2 ± 40.6 counts to 401.7 ± 44.9 counts (p < 0.001, Supplementary Fig. 7h), resulting in a reduction in the Circadian Index of LMA by 38.23 ± 9.2% in LD (p = 0.005) and by 50.61 ± 12.9% in DD (p = 0.003, Supplementary Fig. 7i, j), and a significant reduction in the amplitude of the peak at 24 h in the periodogram and in the cosinor analysis of LMA in both LD and DD (LD: p < 0.001; DD: p < 0.001, Supplementary Fig. 7e, g). Thus, most of the reduction in LMA after DMH lesions appears to be due to loss of DMH^Vgat neurons. In the mice with ablation of DMH^Vgat neurons, Tb was about 0.3 °C lower throughout the day in LD and during the subjective dark period in DD, but there was no statistically significant change in Tb in either the CI or cosinor analysis (Supplementary Fig. 7l–s).

We interpreted these results as indicating that it was unlikely that DMH^Vgat neurons played a role in suppressing Cort secretion during the inactive (light) phase, but that they potentially accounted for the other half (not caused by the DMH^Vglut2 neurons) of the surge in Cort levels during the active (dark) phase. We next evaluated whether this effect was due to GABA release from these neurons (vs. some other neurotransmitter in the same neurons). Thus, we placed AAV8-SYN-EGFP-iCre or AAV8-DIO-EGFP injections in the DMH in Vgat^loxP/loxP (Vgat-flox) mice to delete expression of functional Vgat protein and so to abolish GABA transmission from the GABAergic neurons of the DMH. From 12 mice injected with AAV-EGFP-iCre in the DMH, 7 with bilateral injections covering at least 70% of the DMH were included in the analysis (Fig. 5f, g, Supplementary Fig. 8a). Vgat deletion from DMH neurons did not change the circadian regulation of Cort release in LD, but again reduced the CI of Cort release in DD by 76.6 ± 13.6% (p = 0.008, Fig. 5j), by reducing the CT13 peak of Cort levels in DD from 27.0 ± 4.5 ng/ml in controls to 11.7 ± 2.1 ng/ml in Vgat-deleted mice (p = 0.006, Fig. 5h). Thus, most if not all of the effect of the DMH^Vgat neurons on the circadian release of Cort is mediated by GABA. Interestingly, although the total levels of LMA during the dark period were decreased by Vgat deletion in the DMH (from 979.2 ± 66.8 counts in the control group to 681.5 ± 55.6 counts in the deletion group during LD; p < 0.001, Supplementary Fig. 8f), the reduction was not nearly as profound as seen after the ablation of the DMH^Vgat neurons, suggesting that other transmitters released from those neurons could play a role in regulating LMA. Results in DD were similar (from 979.2 ± 66.8 total counts in DMH^Vgat/flox-Control group to 681.5 ± 55.6 counts in the DMH^Vgat/flox-EGFP-iCre during DD; p < 0.001, Supplementary Fig. 8f), so that the CI of LMA was reduced by only 25.7 ± 6% in LD (p = 0.001) and 33.3 ± 12.6% in DD (p = 0.021, Supplementary Fig. 8g), similar to the effect observed in the amplitude in the cosinor analysis (Supplementary Fig. 8g, h). Tb of DMH^Vgat–deleted mice was also reduced during the dark phase by 0.2 °C in LD (mostly between ZT14-18; p = 0.001, Supplementary Fig. 8j–n), and in the subjective dark by 0.3 °C in DD (p < 0.001, Supplementary Fig. 8l–n). As a result, the Circadian Index of Tb was reduced in the DMH^Vgat–deleted mice by 20.8 ± 8.8% in LD (p = 0.037) and by 26.6 ± 8.9% in DD (p = 0.012, Supplementary Fig. 8o), similar to the reduction in the amplitude in the cosinor analysis (Supplementary Fig. 8o, p).

To determine whether these chronic conditional lesions or impairments of GABA transmission might be affected by compensatory mechanisms, we examined the effect of acute inhibition of the DMH^Vgat neurons on the circadian release of Cort. We placed injections of AAV-DIO-hGlyR-mCherry in the DMH of Vgat-ires-Cre mice. From 7 mice injected, 5 mice with bilateral injection covering at least 70% of the DMH were included for analysis (Fig. 5k, l, Supplementary Fig. 9a). After 4 weeks allowing the full expression of hGlyR, mice were injected with either vehicle or IVM (5 mg/Kg, ip) at ZT1 in LD, or CT1 in DD, and Cort levels were sampled 24 h and 36 h later. The inhibition of the DMH^Vgat neurons did not change the Cort levels under LD, but reduced the level at CT13 from 22.8 ± 6.1 ng/ml after vehicle to 6.9 ± 1.1 ng/ml after IVM administration (p = 0.006, Fig. 5m) and reduced the CI by 95.7 ± 6.9% in DD (p = 0.006, Fig. 5n). Similar to the effects of their ablation, the inhibition of DMH^Vgat neurons reduced LMA during the dark or subjective dark period between 24–48 h after IVM administration, reducing the CI in LD by 77.7 ± 5.7% (VEH p = 0.001, Supplementary Fig. 9b–d), and in DD by 80.7 ± 5% (p = 0.002, Supplementary Fig. 9e–g). The Tb during the dark period or presumptive dark period was lower after IVM administration when compared vs VEH (which includes the effect of the handling in contrast to the baseline), therefore, the CI of Tb was significantly reduced in both LD and DD, 24–48 h after IVM administration by 77.7 ± 8.1% in LD (p = 0.005, Supplementary Fig. 9h–j), and by 68.3 ± 14% in DD (p = 0.004, Supplementary Fig. 9k–m). In all three experiments (ablation, deletion, and inhibition), the changes observed in levels or rhythms of LMA or Tb were not observed in the anatomical controls (animals with injections that missed the DMH). These data suggest that DMH^Vgat neurons participate in the circadian peak of Cort release at the beginning of the active period (as seen in DD), likely reducing the inhibitory tone to the PVH^CRH neurons at this temporal point, but that in LD light can act as a cue for elevation of Cort at the onset of the active period, presumably through another pathway. The ablation of DMH^Vgat neurons did not, however, increase the levels of Cort secretion at ZT1 or CT1, indicating that they do not shape circadian secretion of Cort by suppressing levels at their nadir. Also, the DMH^Vgat neurons play a larger role than DMH^Vglut2 neurons in the circadian regulation of LMA, as the ablation of DMH^Vgat but not the DMH^Vglut2 neurons dramatically reduced LMA during the dark or subjective dark periods.

As the ablation of the DMH^Vgat neurons prevented the peak in DD conditions, we tested whether the activation of those neurons can induce Cort release. We placed injections of AAV8-DIO-hM3Dq-mCherry in the DMH of Vgat-ires-Cre mice. From 8 mice injected, 6 mice were confirmed to have injection sites covering at least 70% of the DMH bilaterally (Fig. 6a, b). CNO administration at the beginning of the light phase (ZT2) elevated the Cort levels to 36.8 ± 9 ng/ml one hour after administration, while Cort was only 10.2 ± 2 ng/ml after vehicle (p = 0.016, Fig. 6c). Although activation of DMH^Vglut2 neurons produced a larger increase in Cort levels ( > 150 ng/ml) than activation of DMH^Vgat neurons, in both cases the increase in Cort levels caused by CNO was at least as great as the circadian peak of Cort in undisturbed littermates. Thus, activation of either DMH^Vgat or DMH^Vglut2 neurons is capable of elevating Cort levels within the range achieved by circadian rhythms.

Because activation of the DMH^Vgat neurons is expected to release GABA and inhibit their postsynaptic targets, we hypothesized that the DMH^Vgat neurons induce Cort release by disinhibiting the PVH^CRH neurons through inhibitory relay neurons, similar to the mechanism by which GABAergic inputs from the arcuate nucleus increase Cort secretion during food deprivation¹⁸. To understand the circuit through which DMH GABAergic neurons promote CRH secretion, we crossed Vgat-ires-Cre mice with CRH-Venus reporter mice to generate Vgat-ires-Cre::CRH-Venus mice. We then injected the DMH of 4 Vgat-ires-Cre::CRH-Venus mice with AAV8-DIO-ChR2-mCherry. These injections were relatively small (3–9 nl compared to 24–45 nl for the other experiments) and covering mostly the rostral part of the DMH (rDMH). We found that axons and terminals labeled anterogradely (expressing mCherry, in red) from rDMH^Vgat neurons largely avoided PVH^CRH neurons (green). The bulk of the labeled terminal field was in areas surrounding the PVH, as well as in the region medial and ventral to the PVH^CRH neurons and extending into the caudal ventral PVH (Fig. 6d, e, j, k). Although the PVH contains very few GABAergic cells, many are found in the region surrounding the PVH, known as the peri-PVH area, as well as in the adjacent caudal ventral part of the PVH. As terminals from the DMH^Vgat neurons blanket the caudal ventral PVH as well as the adjacent peri-PVH just outside it, for the sake of simplicity we will refer to this region as the cvPVH. Local inhibitory inputs within the PVH have been described from this region^21–23, so we decided to explore whether the cvPVH^Vgat neurons might mediate disinhibition of the PVH^CRH neurons during the circadian peak of Cort. To explore whether the cvPVH^Vgat neurons might make monosynaptic contacts with PVH^CRH neurons, we combined conditional retrograde rabies tracing from PVH^CRH neurons with in situ hybridization for Vgat mRNA using the CRH-ires-Cre mice (n = 2). We observed doubly labeled neurons in the cvPVH suggesting monosynaptic GABAergic inputs from this region to the PVH^CRH neurons (Fig. 6f, g). We next injected 3 Vgat-ires-Cre::CRH-Venus mice with AAV8-DIO-ChR2-mCherry in this part of the PVH. The injections covered the ventral and lateral region of the PVH including the cvPVH, and labeled axons ramified inside the PVH, including ones making appositions with the CRH-Venus neurons (Fig. 6h, i). We then investigated whether the DMH^Vgat neurons innervate the cvPVH^Vgat neurons. Following injections of AAV8-DIO-ChR2-mCherry into the rostral DMH of Vgat-ires-Cre::L10-GFP mice (n = 5) we found dense terminal labelling in apposition to the cvPVH^Vgat neurons, suggesting direct connectivity (Fig. 6j, k). These neuroanatomical findings support the hypothesis of a circuit where the DMH^Vgat neurons disinhibit PVH^CRH neurons through direct inhibition of the cvPVH^Vgat neurons (DMH^Vgat → cvPVH^Vgat → PVH^CRH).

To test this model, we ablated the Vgat expressing neurons in the cvPVH by placing injections of AAV10-hSyn-mCherry-DIO-DTA in the cvPVH of Vgat-ires-Cre mice (n = 5). In all cases, the injection sites marked by mCherry expression involved the cvPVH, where Vgat cells projecting to the PVH^CRH neurons are found, and the body of the PVH, where few Vgat cells are seen. (Fig. 6l, m, Supplementary Fig. 10a). After four weeks to allow full expression of DTA, we recorded LMA and Tb in the mice for 12 days in both LD and DD, and we collected blood samples for Cort measurements toward the end of both the LD and DD periods. The cvPVH^Vgat ablated mice showed higher levels of Cort but only at the beginning of the active phase, from 21.3 ± 2.3 ng/ml in controls to 59.8 ± 11.3 ng/ml in cvPVH^Vgat ablated mice in LD (p < 0.001, Fig. 6n), and from 25.6 ± 5.4 ng/ml to 51.3 ± 8.7 ng/ml in DD (p < 0.001, Fig. 6o). The CI of Cort also increased by 180.2 ± 52.9% in LD (p = 0.01) and 100.4 ± 33.9% in DD (p = 0.036, Fig. 6p). At the end of the protocol, we also exposed the mice to restraint stress and found that the Cort levels were significantly higher in the cvPVH^Vgat ablated mice (p = 0.047, Supplementary Fig. 10b). No major changes were observed in LMA or Tb with ablation of cvPVH^Vgat neurons (Supplementary Fig. 10c–r). These data indicate that the cvPVH^Vgat neurons play an important role in inhibiting Cort secretion, even limiting its peak during the daily surge in Cort secretion at CT13 and in response to restraint stress. However, these neurons do not appear to play a role in suppressing Cort levels at their daily nadir. DMH^Vgat inhibition of cvPVH^Vgat neurons partially disinhibits the PVH^CRH neurons, allowing the surge in Cort secretion at CT13, but completely eliminating the cvPVH^Vgat neurons allows even greater stimulation of the PVH^CRH cells at that time by DMH^Vglut2 inputs.

To dissect the synaptic mechanisms by which the DMH^Vgat neurons control the PVH^CRH neurons, we injected the DMH of a new cohort of seven Vgat-ires-Cre::CRH-Venus mice with AAV8-DIO-ChR2-mCherry. We then performed patch clamp recordings in brain slices from green-labelled PVH^CRH neurons on the same side of the brain while photoactivating the red DMH^Vgat terminals in the PVH (Fig. 7a). Our rabies virus tracing data indicated that PVH^CRH neurons receive a direct GABAergic input from the DMH but mainly from its caudal portion (Supplementary Fig. 6e, f). In support of these results, we observed that in three mice where the injection site was restricted mostly to the rostral part of the DMH (rDMH), the photostimulation of the DMH^Vgat input evoked inhibitory synaptic responses in only 4 out of 17 recorded PVH^CRH neurons (23%, n = 17;Fig. 7a–e, r), whereas in four mice in which ChR2 expression involved the more caudal portion of the DMH, the photostimulation of the DMH^Vgat input evoked oIPSCs in nearly all PVH^CRH neurons (96%, n = 23; Supplementary Fig. 11a–g). As the direct GABAergic inputs from the caudal DMH to PVH^CRH neurons do not appear to play a role in inhibiting the PVH^CRH neurons at the circadian nadir of Cort, we focused on whether the rDMH GABAergic inputs could disinhibit the PVH^CRH neurons at the daily peak of Cort secretion.

To investigate whether rDMH^Vgat neurons directly inhibit the cvPVH^Vgat neurons, we expressed AAV8-DIO-ChR2-mCherry in the rDMH of 5 Vgat-ires-Cre::L10-GFP mice. We then recorded from GFP-labelled cvPVH^Vgat neurons while stimulating the ipsilateral putative rDMH^Vgat input to them (rDMH^Vgat →cvPVH^Vgat) (Fig. 7f). Photostimulation evoked opto-inhibitory postsynaptic currents (oIPSCs) in 14 out of 15 recorded cvPVH^Vgat neurons. These oIPSCs depended on GABA_A receptor-mediated signaling, as they were blocked by bicuculline (n = 4; Fig. 7g–m). Furthermore, oIPSCs persisted in the presence of TTX (1 µM) and 4-AP (200 µM) indicating monosynaptic connectivity (n = 6, Fig. 7n).

When we activated the rDMH^Vgat terminals at different frequencies while recording from the PVH^CRH neurons in which single light pulses did not evoke oIPCS, we found that photostimulation of the ipsilateral rDMH^Vgat input at 5 or 10 Hz caused a reduction in the number of spontaneous IPSCs, which became statistically significant at 20 Hz (n = 7; Fig. 8a–e). These results indicate that the rDMH^Vgat neurons can disinhibit most PVH^CRH neurons by reducing their GABAergic afferent input.

Our conditional tracing study and CRACM recordings suggest that the rDMH^Vgat input disinhibits the PVH^CRH neurons at least in part through the GABAergic neurons located in the cvPVH. We therefore tested whether the cvPVH^Vgat neurons can directly inhibit the PVH^CRH neurons (cvPVH^Vgat → PVH^CRH). We placed AAV8-DIO-ChR2-mCherry injections in the cvPVH of 3 Vgat-ires-Cre::CRH-Venus mice. We then recorded from labelled PVH^CRH neurons while photostimulating cvPVH^Vgat neurons and terminals. Photostimulation of cvPVH^Vgat neurons evoked oIPSCs in all PVH^CRH neurons recorded (n = 14). These oIPSCs were mediated by GABA_A receptor signaling as they were blocked by bicuculline (n = 4) and were resistant to TTX + 4-AP, indicating monosynaptic connectivity (n = 6) (Fig. 8f–l). Interestingly, in four cases where the ChR2 expression was more dorsal and lateral involving the portion of the pPVH just lateral to the PVH only about half of the PVH^CRH neurons we recorded from showed oIPSCs (n = 4) (Fig. 8m–r).

Taken together, these results indicate that rostral DMH^Vgat (rDMH^Vgat) neurons can disinhibit PVH^CRH neurons through GABAergic interneurons located in the cvPVH (rDMH^Vgat → cvPVH^Vgat → PVH^CRH), promoting the circadian increase of Cort levels at CT13.

It is also important to point out that the roles of the DMH^Vgat and DMH^Vglut2 neurons in the circadian regulation of Cort secretion appear to depend upon release of GABA and glutamate, respectively, because deletion of just the Vgat or Vglut2 genes from neurons in the DMH had effects similar to abating the entire Vgat or Vglut2 neuron population. The ablation of the SPZ^Vgat neurons also caused almost complete loss of the circadian rhythm of Cort secretion and reduced the amplitude of the LMA rhythm by about 80% and reduced the Tb rhythm by about 60%. These effects on LMA and Tb were similar to a previous report using nonspecific SPZ ablation in rats¹¹. However, the deletion of the Vgat gene in the SPZ had almost no effect on the circadian rhythm of Cort and only modestly reduced the of the amplitudes of LMA and Tb by about half as much as Vgat cell ablation. This disparity suggests that SPZ^Vgat neurons cannot rely upon GABA alone to regulate these circadian rhythms. Information about the expression of peptides or other transmitters by SPZ neurons is limited so far, and will likely depend upon single cell mRNA expression profiling. On the other hand, many DMH neurons also express peptides, including galanin, dynorphin, neuropeptide Y, CART, orexin, and many others^33–39. While the different patterns of gene expression by DMH neurons may help us to classify them and provide key marker genes for neurons that participate in particular roles or have specific projections, it does not appear that the DMH neurons depend upon peptidergic transmission to produce the circadian pattern of Cort secretion.

In this work, we also were able to dissect the role of the DMH^Vglut2 vs DMH^Vgat neurons in the global reduction of LMA and Tb and dramatic reduction in the circadian rhythm of LMA reported after non-specific ablation of the DMH in rats¹². The DMH^Vgat neurons appear to be particularly important for the circadian increase of LMA during the active period, as their ablation reduced the amount of LMA by about half across the entire dark or presumptive dark cycle. By contrast, ablation of the DMH^Vglut2 neurons caused a much smaller decrease in LMA which was predominantly confined to the transitions between the light and dark periods. The combination would explain the low levels of LMA with almost complete loss of circadian rhythms of LMA reported in rats after non-specific DMH cell ablation¹². The ablation of either the DMH^Vgut2 or DMH^Vgat neurons reduced Tb mainly during the dark and presumptive dark phase when the animals are most active. This effect is consistent with the known increase in Tb that occurs during LMA. In addition, dorsal hypothalamic/DMH glutamatergic neurons also play a direct role in promoting thermogenesis by projections to the raphe pallidus^29,40.

Previous work in rats had shown that non-specific ablation of the DMH in rats prevented the daily increase of Cort levels associated with the active cycle, reducing Cort secretion to the level normally measured during the inactive period when the animals are mainly asleep¹². Because the effect of the DMH lesions was to eliminate the daily peak in Cort levels without affecting basal (nadir) levels, we initially hypothesized that the circadian input from the DMH to the PVH^CRH neurons was likely to be excitatory, i.e., glutamatergic. In this study we tested that hypothesis and observed that the DMH^Vglut2 neurons play an important role in the circadian elevation of Cort at CT13 but that ablation of the DMH^Vglut2 neurons did not completely eliminate the circadian surge in Cort. We therefore tested the role of DMH^Vgat neurons and were surprised to find that they also participate in the circadian elevation of Cort, particularly in DD, at CT13 suggesting that the DMH^Vgat neurons would be likely to disinhibit PVH^CRH cells at this time.

These observations caused us to examine the origins of the GABAergic inputs to PVH^CRH neurons. The PVH contains very few GABAergic neurons within its borders, whereas the pPVH is rich in GABAergic neurons and it has been proposed that the pPVH^Vgat neurons could provide tonic inhibition of the PVH^CRH neurons and function as a “brake” upon their secretion of CRH²³. On the other hand, other studies have found that some of the GABAergic inputs to PVH^CRH neurons derive from neurons that are distant from the PVH. For example, a recent study from Douglass and colleagues reported that the elevation of Cort levels during fasting depends upon AgRP neurons in the arcuate nucleus inhibiting tonically active GABAergic afferents from the bed nucleus of the stria terminalis to the PVH^CRH neurons, thus disinhibiting them¹⁸. In the case of circadian secretion of Cort, we found that the rDMH^Vgat neurons project to cvPVH^Vgat neurons, and that cvPVH^Vgat neurons provided a particularly rich source of inhibitory input to the PVH^CRH cells. Furthermore, activation of DMH^Vgat terminals in the PVH disinhibited PVH^CRH cells. Interestingly, the surge in Cort levels at CT13 was greater after the ablation of the cvPVH^Vgat neurons than in control animals, suggesting that the tonic activity of cvPVH^Vgat neurons places a brake on CRH secretion that is only partially lifted during the circadian daily peak. This interaction suggests that disinhibition of the PVH^CRH neurons may be a common motif in their regulation.

Although the evidence is strong for DMH^Vglut2 neurons directly exciting PVH^CRH cells and DMH^Vgat neurons disinhibiting PVH^CRH cells via their inputs to the cvPVH^Vgat neurons, we cannot rule out other polysynaptic contributions driven by the DMH^Vglut2 and DMH^Vgat neurons. In addition, both neuron types in the DMH may contribute to regulation of Cort under other physiological conditions. Exploring these alternatives will require resolving the DMH glutamatergic and GABAergic neurons into discrete genetically identified cell types and determining their connections and physiological roles.

Although the activities of the DMH^Vglut2 and DMH^Vgat circadian inputs to the PVH^CRH neurons are temporally aligned in control of the circadian rhythm of Cort, neither is in phase with the transitions between the light and dark periods. In fact, in mice the activity of PVH^CRH neurons peaks about 6 h and the Cort levels rise about 3 h prior to the onset of the active (dark or presumptive dark) cycle, and reach a peak shortly after the active cycle begins, only to fall back to low levels by halfway through the active cycle^41,42. Similarly, in humans the cortisol levels begin to rise about halfway through the habitual sleep cycle and are about twice as high at 8am as they are at 4 pm⁴³. As indicated by our rabies virus tracing experiments (Supplementary Fig. 6b–d) and previous observations by others^4,5,42, there are few if any direct projections from the SCN to the PVH^CRH neurons. It has been suggested that the sparse SCN^VIP inputs to regions of the PVH that are nearby the CRH neurons may influence their firing via volume transmission⁴². However, our data indicate that such inputs are insufficient to cause the daily circadian surge in Cort secretion in the absence of relays in the SPZ and DMH. The fact that the timing of the increase in Cort secretion is offset by about 6 h from the light-dark cycle suggests that there is further circuitry downstream of the SCN that converts its signal, which is tied to the light-dark cycle, into a timing signal that starts the increase in Cort several hours prior to the onset of the active period. In addition, in nocturnal (mice) and diurnal (humans) species the onset of Cort secretion bears the same time relationship to onset of the active phase, even though their relationships to the light cycle (and the peak activity of SCN neurons) are 180 degrees out of phase with each other.

Here, our data suggest that the circadian pattern of Cort secretion depends upon a multi-synaptic circuit from the SCN to the SPZ and then the DMH. The ventromedial SPZ receives a large portion of the SCN output and projects heavily to the DMH in both rats and mice^6,15,44,45. Most SPZ neurons have activity patterns that are in antiphase to the SCN (i.e., are more active during the dark period) and they are almost uniformly GABAergic^9,10, but our results point to a different mechanism than GABA release in relaying the circadian rhythm of Cort secretion. The SPZ also receives inputs from a wide range of other hypothalamic cell groups. Thus, it is possible that various configurations of polysynaptic connections in the SPZ might allow it to produce timing signals that are driven by the SCN but not in phase with it. Such a multi-synaptic pathway is consistent with the smooth and gradual increase in Cort secretion that occurs prior to the active phase, allowing the SPZ^Vgat neurons to drive the firing of DMH^Vgat and DMH^Vglut2 neurons that are responsible for the circadian rhythm of Cort secretion. Further study of the intrinsic circuitry of the SPZ and the dynamics of its interaction with the DMH will be important for understanding how the SCN signal is used to optimally time various physiological processes.

Because adult female mice lose their circadian rhythm of Tb every 4-5 days during estrus and we needed to record circadian rhythms at specific times across many days, we used only male adult mice, age 12–16 weeks old, in these experiments. The strains of the mice were Vgat-ires-Cre⁴⁶ (JAX: 016962), Vglut2-ires-Cre⁴⁶ (JAX: 016963), CRH-ires-Cre⁴⁷ (JAX: 012704), CRH-Venus∆Neo⁴⁸, Vglut2^loxP/loxP (JAX:036439), Vgat^loxP/loxP49 (JAX: 012897) and R26-loxSTOPlox-L10-GFP mice⁴⁷. Mice were individually housed under 12–12 h light/dark cycle unless the protocol specified otherwise. Room temperature was controlled in a range of 22 ± 2 °C and free access to food and water was provided. All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and formal approval of our protocols was obtained from the Institutionary Animal Care and Use Committee at Beth Israel Deaconess Medical Center. All precautions were taken to minimize pain and discomfort in the mice.

AAV10-hSyn-mCherry-DIO-DTA, which conditionally expresses the subunit A of diphtheria toxin in a Cre-dependent fashion and the mCherry protein in non-Cre cells (acquired from Patrick M Fuller)^50,51 was injected ( ∼ 24nL) in Vgat-ires-Cre or Vglut2-ires-Cre mice. AAV8-eSYN-EGFP-T2A-iCre containing the genes for Cre and EGFP under the neuronal eSYN promoter (VB1089, Vector Biolabs, PA, US; 8.6 × 10¹¹ viral genomes ml⁻¹) was bilaterally injected ( ∼ 24nL) in Vglut2^loxP/loxP and Vgat^loxP/loxP mice. AAV8-Ef1a-DIO-ChR2(H134R)-mCherry injections (UNC, addgene #AV9080, 1.4 × 10¹³ viral genomes ml⁻¹) were used for the neuroanatomical and CRACM experiments ( ∼ 3–9 nL) in Vgat-ires-Cre or Vglut2-ires-Cre mice. AAV10-DIO-hGlyR-mCherry (acquired from Patrick M Fuller)^16,29,45 injections were made in the DMH ( ∼ 45 nL) either in Vglut2-ires-Cre or Vgat-ires-Cre mice. AAV10-EF1α-DIO-hM3Dq-mCherry (UNC, addgene #50460 virus core; 2 × 10¹³ viral genomes ml⁻¹) was injected ( ∼ 30nL) either in Vglut2-IRES-Cre or Vgat-IRES-Cre mice. AAV8-CAG-DIO-EGFP ( ∼ 24 nl) was injected as a control virus in specified experiments (UNC; addgene #59331 7 × 10¹² viral genomes ml⁻¹). ∼ 45 nL of AAV8-Ef1a-DIO-TVA-mCherry (UNC Vector Core; 1.13 × 10¹² viral genomes ml⁻¹) mixed 1:1 with AAV8-CAG-DIO-rabiesG (Stanford Vector Core; 3.4 × 10¹² viral genomes ml⁻¹) was injected unilaterally, and 21 days later, ∼ 55nL of EnvA-ΔG-rabies-EGFP (Salk Viral Vector Core; 1.51 × 10⁸ viral genomes ml⁻¹) were administrated in the PVH of CRH-ires-Cre mice. All the experiments and recordings started four weeks after the injections to allow complete transfection with the AAVs.

Mice were deeply anesthetized with a mixture of ketamine/xylazine (100/10 mg/kg, i.p.). Radiotelemetry sensors (TA-F10, DSI, US) for body temperature (Tb) and locomotor activity (LMA) were implanted in the peritoneal cavity with a small abdominal incision. The coordinates for the stereotactic microinjections of viral vectors in the brain, from bregma were: DMH, AP = −1.75 mm, ML = ± 0.25 mm, DV = −4.9 mm; PVH, AP = −0.8 mm, ML = + 0.25 mm, DV = −4.8 mm; cvPVH, AP = −0.8 mm, ML = +0.25 mm, DV = −4.85 mm⁵². Mice receive analgesic treatment with meloxicam or buprenorphine post-surgery. All procedures were performed under aseptic conditions.

Average Tb and LMA were recorded every 5 min during all the protocols using the radiotelemetry DSI system. The signal from the telemetry probes, previously implanted, was received and converted with the PhysioTel HD and PhysioTel (DSI) hardware. The LMA and Tb data were analyzed using the software ClockLab Analysis version 6 (Acticmetrics).

Blood samples ( ~ 20 µl) were obtained in a microvette tube (CB300, Sarstedt, US) after a small incision in the tail in less than 1 min after onset of handling the mouse, centrifuged at 9,240 × g for 10 min, and the plasma was collected and stored at −40 °C. Samples were collected with a minimum interval of 30 h between samples. Under the DD protocol, the first sample was obtained on the third day after the lights turned off at CT13 and then every 30 h until we completed the 4 time points. An ELISA assay for corticosterone was performed as per the manufacturer instructions (ADI-901-097, Enzo life science, US) and the corticosterone signal read with a 405 nm filter (iMark, Bio-Rad, US). Each sample was read in duplicates.

At the end of the LMA and Tb protocols, mice were physically restrained for 1 h starting at ZT3 in a plastic tube with small holes to allow normal respiration. Blood samples were obtained at the end of the stress protocol by a small incision in the tail. The mice were placed in their home cage after the protocol.

At the end of each protocol, animals were deeply anesthetized with 7% chloral hydrate (0.015 ml/gr of body weight i.p.) and perfused transcardially with 30 ml of PBS followed by 30 ml of 10%-buffered formalin (Thermo Fisher Scientific, US). Brains were extracted and postfixed in 10%-buffered formalin overnight, cryoprotected in 30% sucrose solution, sectioned into 40 µm coronal sections (three series) and stored at 4 °C in PBS with sodium azide. Brain sections were washed three times (5 min each) and then blocked for 1 h with 3% normal horse serum (diluted in 0.4% Triton X-100 in PBS). Sections were incubated in primary antibody overnight in the same solution as blocking, at room temperature with continuous agitation. Primary antibody used were rabbit anti VIP (ImmunoStar; 1:1000, #20077), Chicken anti-GFP (Invitrogen; 1:5000, A10262), Rat anti-mCherry (Invitrogen; 1:4000, M11217↗), Mouse anti-NeuN (Millipore; 1:3000, MAB377). Then, the sections were washed six times (5 min each) in PBS, and incubated with secondary antibodies for 2 h in blocking solution at room temperature and continuous agitation. The secondary antibodies used were donkey anti rabbit antibody conjugated with CY5 fluorophore (Jackson Immuneresearch; 1:1000, # 711-175-152), Alexa fluor 488-conjugated Donkey anti-Chicken (Jackson; 1:200, AB2340375), Alexa fluor 555-conjugated Goat anti-Rat (Invitrogen; 1:200, A21434), Alexa fluor 488-conjugated Goat anti-Mouse (Jackson; 1:200, AB2338840). Then, sections were washed three times (10 min each) in PBS, mounted on electrostatically treated slides and coverslipped with VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories) for microscope visualization and image acquisition.

Two brains from the EnvA-ΔG-rabies-EGFP experiments were sectioned at 20 µm and tissue was mounted on glass slides in RNAase-free conditions. Slices were dried and in situ hybridization was performed using RNAScope multiplex fluorescent reagent kit V2 (catalog #323100, Advanced Cell Diagnostics, US). Slices were pretreated with hydrogen peroxide for 10 min at room temperature and subjected to target retrieval step for 5 min in a steamer (˃99 °C), This was followed by dehydration in 90% alcohol and air-dried for 5 min. Sections were then treated with protease III at 40 °C for 30 min. Then, the slices were rinsed with sterile water and incubated for 2 hrs either with the Vglut2 probe (RNAscope Probe- Mm-Slc17a6; catalog #319171, Advanced Cell Diagnostics, US) or Vgat Probe (RNAscope Probe- Mm-Slc32a1; catalog # 319191, Advanced Cell Diagnostics, US) at 40 °C for RNA hybridization. This is followed by incubation with amplification reagents, AMP1, AMP2 (30 min each) and AMP3 (15 min) at 40 °C. Sections were then incubated with HRP -C1 for 15 min and Cy5 fluorophore (catalog #NEL741001, PerkinElmer) for 30 min at 40 °C. Finally, HRP blocker was added to the sections for 15 min at 40 °C. After each step sections were washed with 1 x wash buffer provided in the kit. Slides were then dried and coverslipped with Vectashield antifade mounting medium (catalog #H-1400, Vector Laboratories) for microscope visualization and image acquisition.

Coronal sections were scanned at 20X magnification using an Olympus VS120 slide-scanning microscope or at 63X magnification using a confocal microscope (Leica Stellaris 5). Fluorescence images were analyzed using Olympus OlyVIA software (3.4.1) or Leica application suite (version 4.2.1). Photomicrographs were adjusted for brightness and contrast to compensate for over- or undersaturation due to differences in fluorescence brightness. The injection site of each animal was confirmed by the expression of mCherry or EGFP. Microphotographs were transformed to a gray scale and the background were subtracted. The images at the same anatomical level were superimposed with 30% transparency to build a gradient map using GIMP software (version 2.10.34). In the brain sections from the rabies experiment, we counted cell bodies in at least 3 levels for each nucleus (anterior, middle and posterior). In the sections used for in situ hybridization of rabies-infected cells, we counted cell bodies and doubly-stained cells in 6 levels of the DMH. Cell body counting of the rabies infected cells was conducted manually using the multipoint tool in FIJI (Image J software, version 1.54). Because the number of infected cells varied considerably from animal to animal and we were not trying to establish the absolute number of cells that project to the PVH^CRH neurons, we did not attempt to correct the numbers for cell size.

For electrophysiology experiments, AAV8-DIO-ChR2-mCherry (~3–9 nL) was injected into the DMH of Vglut2-ires-Cre::CRH-Venus mice (n = 4) or Vgat-ires-Cre::CRH-Venus mice (n = 3) or Vgat-ires-Cre::L10-GFP mice (n = 5) or into the cvPVH of Vgat-ires-Cre::CRH-Venus mice (n = 7) or into the cneurons neurons

DMH of Vgat-ires-Cre::CRH-Venus mice (n = 4). Mice with AAV injections placed outside of the target were excluded from the analysis. Four to six weeks following the AAV injections, mice were deeply anaesthetized with isoflurane (5% in oxygen) via inhalation and transcardially perfused with ice-cold ACSF (N-methyl-D-glucamine, NMDG-based solution described below). The mouse brains were then quickly removed and sectioned coronally (250 µm-thickness) in ice-cold NMDG-based ACSF using a vibrating microtome (VT1200S, Leica). We first incubated the slices containing the PVH for 5 min at 37 °C, then transferred them into a holding chamber at 37 °C containing ACSF (Na-based solution) for 10 min. We let the brain slices gradually return to room temperature ( ~ 1 h) before starting recording. Brain slices were recorded in a recording chamber, where were submerged, and perfused with Na-based ACSF (described below, 1–1.5 ml/min). We recorded PVH^CRH and cvPVH^Vgat neurons identified by Venus or L10-GFP (green) fluorescence, respectively, using a combination of fluorescence and infrared differential interference contrast microscopy (IR-DIC). Both the Venus-expressing CRH neurons and the L10-GFP neurons were numerous and clear under the microscopy without any enhancement protocol. We used a fixed stage upright microscope (BX51WI, Olympus America) equipped with a Nomarski water immersion lens (Olympus 40X / 0.8 NAW) and IR-sensitive CCD camera (ORCA-ER, Hamamatsu) to acquire real-time images using Micro-Manager software. We recorded the neurons in whole-cell voltage clamp and current clamp configurations using a Multiclamp 700B amplifier (Molecular Devices), a Digidata 1322 A interface, and Clampex 9.0 software (Molecular Devices). Neurons showing a greater than 10% change in input resistance over the duration of the recording were excluded from the analysis. We photostimulated the DMH or cvPVH axons in the PVH using a full-field ( ~ 10 mW/mm², 1 mm beam width) 5 W LUXEON blue light-emitting diode (470 nm wavelength; #M470L2-C4; Thorlabs), coupled to the epifluorescence pathway of the microscope. We stimulated photo-evoked excitatory postsynaptic currents (oEPSCs) or inhibitory postsynaptic currents (oIPSCs) with 10 ms light pulses (0.1 Hz, for a minimum of 30 trials). In the DMH^Vgat injected mice, we also tested the effects of photostimulation on both the action potential firing of cvPVH^Vgat neurons and the spontaneous IPSC (sIPSCs) of PVH^CRH neurons. For sIPSC recordings, we used train stimulations with a duration of 60 s at frequencies of 5, 10, and 20 Hz, and a light pulse duration of 10 ms. For action potential recordings, we tested train stimulations of 10 and 60 s at the same frequencies (5, 10, and 20 Hz) with a light pulse duration of 10 ms. Action potential firing and oEPSCs (holding potential = −70 mV) were recorded in ACSF and using a K-gluconate-based pipette solution. oIPSCs (holding potential = 0 mV) were recorded in ACSF containing 1 mM kynurenic acid and using a Cs-methane-sulfonate-based pipette solution. To record photo-evoked synaptic events in the presence of synaptic blockade, we bath applied tetrodotoxin (TTX; 1 µM) and the potassium channel blocker, 4-AP (200–500 µM). For all recordings we added 0.5% biocytin in the pipette solutions to mark the recorded neurons. The recorded slices containing the PVH and cvPVH and the slices containing the DMH, where the AAVs were injected, were fixed overnight in 10%-buffered formalin for post hoc histological and anatomical assessment. To label and map the recorded neurons filled with biocytin, immediately after the in vitro recordings, we fixed the recorded slices in 10% buffered formalin, washed them, and incubated them overnight in streptavidin-conjugated Alexa Fluor 405 (1:500; Cat#: S32351↗; Invitrogen, Thermo Fisher Scientific Waltham, MA)^53,54. We acquired images using a Leica Stellaris 5 confocal microscope using a 63X oil immersion objective.

NMDG-based ACSF solution containing (in mM): 100 NMDG, 2.5 KCl, 1.24 NaH₂PO₄, 30 NaHCO₃, 25 glucose, 20 HEPES, 2 thiourea, 5 Na-L-ascorbate, 3 Na-pyruvate, 0.5 CaCl₂, 10 MgSO₄ (pH 7.3: 95% O₂ and 5% CO₂; 310–320 mOsm). Na-based ACSF solution contained (in mM): 120 NaCl, 2.5 KCl, 1.3 MgCl₂, 10 glucose, 26 NaHCO₃, 1.24 NaH₂PO₄, 4 CaCl₂, 2 thiourea, 1 Na-L-ascorbate, 3 Na-pyruvate (pH 7.3–7.4 in 95% O₂ and 5% CO₂; 310–320 mOsm). Cs-methane-sulfonate-based pipette solution containing (in mM): 125 Cs-methane-sulfonate, 11 KCl, 10 HEPES, 0.1 CaCl₂, 1 EGTA, 5 Mg-ATP and 0.3 Na-GTP (pH adjusted to 7.2 with CsOH, 280 mOsm). K-gluconate-based pipette solution containing (in mM): 120 K-Gluconate, 10 KCl, 3 MgCl2, 10 HEPES, 2.5 K-ATP, 0.5 Na-GTP (pH 7.2 adjusted with KOH; 280 mOsm). We purchased tetrodotoxin and kynurenic acid from Cayman Chemical (Ann Arbor, MI) and bicuculline methiodide from Tocris Bioscience (Ellisville, MO). We purchased all other chemicals from Fisher Scientific (Waltham, MA) or Sigma-Aldrich (Saint Luis, MO).

Recording data were analyzed using Clampfit 10 (Molecular Devices), MiniAnalysis 6 software (Synaptosoft), customized Python scripts (Python 3, www.python.org↗) and MatLab (version R2020B; MathWorks; Natick, MA) software. Figures were generated using Igor Pro version 6 (WaveMetrics), Prism 7 (GraphPad, La Jolla, CA), Inkscape (GitLab) and Photoshop (Adobe) software. To ensure unbiased detection of synaptic events, the EPSCs and IPSCs were detected and analyzed automatically using MiniAnalysis. We considered EPSCs or IPSCs in PVH^CRH neurons and cvPVH^Vgat neurons to be photo-evoked if their probability during the first 50 ms following the light pulses was greater than the IPSC probability + five times the standard error of the mean (SEM) before the light stimulation and if their latency was within ± 1 ms of the median value of E/IPSC latencies calculated for each neuron within the first 50 ms following photostimulation. We calculated the latency of the photo-evoked EPSC and IPSCs as the time difference between the start of the light pulse and the 5% rise point of the first synaptic event⁵⁴. Group means were compared using paired t-tests. Cumulative distributions were compared using the Kolmogorov-Smirnov test. Values indicating p < 0.05 were considered significant.

Data sets were tested to see whether they fulfilled the parametric criteria of Brown-Forsythe (homo/heteroscedasticity) and Shapiro-Wilk (fit to normal distribution). Corticosterone plasma levels and circadian distribution of LMA and Tb were analyzed with a two-way ANOVA with a factor for group and a factor for time, followed by a post-hoc multiple comparisons Sidak test. CI for Cort was calculated by the difference between the mean ZT13 and ZT1 levels and divided by the ZT13-ZT1 mean in each mouse, then normalized to 100% by the mean difference in the Control group. CI for LMA and Tb were calculated by the difference between the mean dark (or presumptive dark) and light (or presumptive light) levels divided by the 24 h total counts for LMA and 24 h mean for Tb, then normalized to 100% against the mean for the Control group. Light or Dark mean for LMA and Tb were evaluated with two-way ANOVA with a factor for group and a factor for LD vs DD, followed by a post-hoc multiple comparisons Tukey test. When the same mice were evaluated under different conditions as their own controls for LMA and Tb (pre- and post-DTA experiments, and the hGlyR experiments), we evaluate them with a Repeated Measures [RM] two-way ANOVA with a factor for group and a factor for LD vs DD, followed by a post-hoc multiple comparisons Tukey test. CI and amplitude data were compared by unpaired t-test analysis, or one-way ANOVA, followed by a post-hoc multiple comparisons Tukey test in the cases where included baseline in the analysis. Actograms, periodograms, and amplitude of the cosinor fitting were built using ClockLab ActiMetrics software version 6.1.02. Statistical analyses and graphs were performed with the GraphPad Prism software version 8. Data are represented as mean ± SEM. Significance values α were set at p < 0.05.

Further information on research design is available in thelinked to this article. Nature Portfolio Reporting Summary

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

The data generated in this study are provided in the Source Data file.are provided with this paper. Source data