Somatostatin-expressing interneurons induce early NO-driven and late specific astrocyte-mediated vas

Somatostatin-expressing interneurons induce early NO-driven and late specific astrocyte-mediated vasodilation

Abstract

Somatostatin-expressing (SST) interneurons modulate hemodynamic responses both directly and indirectly, but their precise role remains unclear. Here, we investigated the influence of SST interneurons on hemodynamic control in response to optogenetic stimulation of SST neurons and somatosensory stimulation in both awake and anesthetized mice. Prolonged optogenetic stimulation of SST neurons induces fast vasodilation through nitric oxide synthase-expressing neurons that co-express SST, and slow vasodilation mediated by astrocytes. Similar neurovascular coupling mechanisms are observed during prolonged sensory stimulation, which also induces both fast and delayed vasodilation. The delayed vasodilation, mediated by the SST neuron-astrocyte pathway, enhances the specificity of cerebral blood volume (CBV)-weighted fMRI signals to cortical layer 4, as confirmed by chemogenetic inhibition of SST neurons. Our findings indicate that the SST neuron-astrocyte-vascular pathway shapes hemodynamic responses to prolonged stimulation and is critical for achieving high-specificity, laminar-resolution fMRI, which is increasingly pursued in human cognitive studies.

Introduction

Neurovascular coupling (NVC) refers to the close interplay between neural activity and the hemodynamic response. Understanding NVC is crucial for studying brain functions with hemodynamic-based functional magnetic resonance imaging (fMRI)1 and for elucidating the pathophysiology of neurological disorders such as stroke2,3, Alzheimer’s disease4,5, and epilepsy6. Since ~85% of neurons in rodent and 60–70% in primate are excitatory7,8,9,10,11, the role of excitatory neural activity in the hemodynamic response has been extensively investigated1,12,13, often assuming that hemodynamic responses are driven directly by excitatory neurons and indirectly through astrocytes14,15. The remaining 15% of neurons are inhibitory, which play an important role in the neural circuits by modulating the activity of neighboring neurons and maintaining the excitatory: inhibitory balance16,17,18,19. Among these inhibitory GABAergic interneurons, somatostatin-expressing (SST) neurons account for ~30% of cortical GABAergic interneurons20,21. These neurons release vasoactive nitric oxide (NO), vasoconstrictor neuropeptide Y (NPY), as well as GABA and SST22, complicating their direct and indirect roles in vascular control23,24,25. Activation of SST-positive cells induces vasodilation, presumably mediated by NO24,25. Moreover, prolonged vasodilation has been observed following optogenetic stimulation of SST-positive interneurons25. However, the underlying mechanisms responsible for this long-lasting vasodilation remain unknown. Hence, understanding the precise role of SST interneurons in the hemodynamic response is a crucial aspect of hemodynamic-based brain research.

In addition to their neuronal roles in hemodynamic responses, astrocytes also regulate NVC by releasing potassium ions and prostaglandin E2, which are triggered by changes in astrocytic calcium levels (Ca2+)26,27,28,29. The exact role of astrocytes to NVC is debated30 due to the delayed astrocytic calcium activity compared to evoked capillary responses14, and the absence of impact on evoked hemodynamic responses by astrocytic IP3R2-mediated Ca2+ changes31,32,33,34. Delayed astrocyte activity is observed in conjunction with sustained hemodynamic responses to prolonged sensory stimulation35,36,37. The late component of sustained functional hyperemia is eliminated by silencing astrocytic activity using CalEx or N-methyl-D-aspartate (NMDA) receptor inhibition35, indicating that excitatory neuron-astrocyte interaction is crucial for NVC. However, astrocyte calcium activity is also enhanced by SST interneuron activity38, which is suppressed by NMDA receptor blocker39,40. Thus, inhibitory somatostatin interneurons, which interact with excitatory neurons, may serve as intermediaries in mediating astrocyte calcium activity and the subsequent late vasodilative response. Interestingly, the late vasodilation in response to prolonged sensory stimulation tends to be more specific to layer 4, the primary target of sensory inputs from the thalamus, compared to the early hemodynamic response41,42. Hence, understanding the role of SST interneurons in astrocyte activity and late hemodynamic responses is crucial, particularly concerning their relevance to functional MRI responses to external sensory stimuli.

Here, we investigated the influence of SST interneurons on hemodynamic control in response to optogenetic stimulation of SST neurons and somatosensory stimulation, under both anesthesia and awake conditions. Neural, astrocytic, and hemodynamic activities were measured using a combination of electrophysiology, calcium imaging, intrinsic optical imaging (OIS), and fMRI, integrated with pharmacological interventions and chemogenetic inhibition. We found that the vascular response to optogenetic stimulation of SST neurons depends on frequency and duration of stimulation and the brain’s status. Specifically, during 20-s-long stimulation, hemodynamic responses exhibited three phases: rapid initial dilation, slow decrease, and later prolonged dilation. We investigated the underlying sources of these hemodynamic changes by imaging astrocytic calcium activities and applying receptor blockers. Additionally, we examined the role of SST neurons in spatiotemporal hemodynamic regulation in response to somatosensory stimuli by utilizing SST sensors in astrocytes, SST receptor blockers, SST inhibition and astrocyte inhibition. Furthermore, we analyzed CBV-weighted fMRI responses to somatosensory stimulation at the laminar level. We found that SST neurons induce early NO-driven and later specific astrocyte-mediated vasodilation in both optogenetic and sensory stimuli. This comprehensive approach revealed the impact of SST neurons on neuro-glio-vascular coupling, offering insight into their role in shaping vascular responses in both health and neurological disorders.

Results

Overview of experimental designs

To investigate the role of SST interneurons in NVC, various experiments were conducted using SST-cre×Ai32 mice, SST-Cre mice, Thy1-GCaMP6f mice, and wild-type mice (C57BL/6), under both anesthetized and awake conditions. Anesthesia was used to facilitate comparisons with previous NVC studies25,43,44 and to perform fMRI experiments. A summary of the experimental protocols corresponding to each figure is provided in Supplementary Table 1, including anesthetic status and the number of animals used in each experiment. Briefly, 1) hemodynamic responses, blood oxygenation-level dependent (BOLD) fMRI, and electrophysiological activity in response to optogenetic stimulation of SST interneurons were measured. 2) To characterize the source of hemodynamic responses to SST interneurons and sensory stimulation, hemodynamic responses were assessed following nitric oxide synthase (NOS) blockade. 3) To determine the causal role of SST interneurons in hemodynamic responses, SST or GABAB receptor blockade was performed in conjunction with calcium imaging and OIS during optogenetic stimulation of SST neurons. 4) The impact of SST neurons on hemodynamic responses to sensory stimulation was investigated using SST fluorescent sensors, calcium imaging, and OIS in combination with SST inhibition, SST receptor blockage, or astrocyte inhibition. 5) High-resolution CBV-weighted fMRI in response to forepaw stimulation was performed before and after chemogenetic inhibition of SST neurons at an ultrahigh field of 15.2 T.

Sustained optogenetic stimulation of somatostatin interneurons elicits prolonged vasodilation

Firstly, we examined the hemodynamic response induced by optogenetic activation of cortical somatostatin neurons at the primary somatosensory forelimb area (S1FL) in mice expressing channelrhodopsin-2 (ChR2) specifically in SST neurons (Supplementary Fig. 1a). Hemodynamic responses were monitored using a two-wavelength OIS system under lightly anesthetized (ketamine/xylazine) or awake condition (Fig. 1a), and then converted into changes in hemoglobin concentration (see Methods). The optogenetic stimulation, delivered at 1 Hz and 20 Hz, was administered for durations of 5 s and 20 s (Fig. 1a). As a reference experiment, the hemodynamic response evoked by forepaw stimulation (5 Hz) under anesthesia was also acquired.

Fig. 1: Hemodynamic responses evoked by sensory and optogenetic stimulation under anesthesia and awake conditions.

A brief 5-s forepaw somatosensory stimulus under anesthesia induced a rapid increase in total hemoglobin (HbT) concentration (often referred to as CBV) at the localized forelimb area, indicating vasodilation (red voxels, upper panel Fig. 1b). With sustained stimulation for 20 s, the functional vascular response to sensory stimulus peaked within the first 5 s of stimulation, followed by a slow increase (Fig. 1b, c), consistent with previous findings in awake mice35,45,46,47.

Compared to the hemodynamic response observed during somatosensory stimulation (Fig. 1b–d, and Supplementary Fig. 1b), optogenetic stimulation of SST neurons induced a much larger response at the stimulation site under both anesthesia and awake conditions. This difference is likely due to the activation of a larger number of SST neurons by optogenetic stimulation than by sensory stimulation, including NOS-expressing SST neurons. Notably, a widespread negative CBV response was evident in the surrounding area of the stimulation site under awake conditions (blue voxels, Fig. 1b and Supplementary Fig. 1b, c), consistent with prior findings in awake48 and isoflurane-anesthetized mice25. Sustained optogenetic stimulation of SST neurons showed a triphasic vascular response: initial rapid vasodilation, a trough, and prolonged late vasodilation, more evident at high frequency stimulation under awake state (Fig. 1b, c). Similar hemodynamic response behaviors were observed for 5 Hz and 40 Hz stimulation (Supplementary Fig. 1e, f).

Detailed analyses of the trimodal response to 20-s optogenetic stimulation of SST interneurons revealed interesting observations. Firstly, activating SST neurons quickly increased CBV under both anesthesia and awake conditions, referred to as an initial peak (Fig. 1d initial peak). Secondly, following the initial peak, the hemodynamic response underwent a subsequent reduction except 1 Hz stimulation under anesthesia (Fig. 1c, d). Specifically, 20 Hz stimulation elicited a more pronounced suppression under the awake condition (trough at 20 Hz stimulation: 15.25 ± 3.07 and 1.24 ± 0.82% under Ket/Xyl and awake condition, n = 5 mice, p = 0.0014, one-way ANOVA test with Turkey comparison) (Fig. 1b, c and Fig. 1d trough). The difference between the initial peak intensity and the subsequent trough is likely related to a balance between SST neuron-induced vasodilation and inhibition-driven vasoconstriction. The magnitude of the trough at the stimulation site (Fig. 1b) was strongly correlated with that of the negative CBV response in the surrounding area for 20 Hz stimulation (Supplementary Fig. 1d), indicating their interrelation. Thirdly, after the trough, the hemodynamic response increased greatly and persisted for about 10 s beyond the stimulation period in both anesthesia and awake conditions as a later peak. Despite similar initial peak intensities, the post-stimulus vasodilation after the 20-s stimulus offset was higher under Ket/Xyl compared to the awake condition (Ket/Xyl vs awake, for 1 Hz: 17.89 ± 2.11 vs 4.16 ± 1.09%, n = 5 mice, p = 0.0019; for 20 Hz: 25.67 ± 2.25 vs 13.26 ± 3.02%, n = 5 mice, p = 0.0072, one-way ANOVA test with Turkey comparison) (Fig. 1d post-stimulus). A post-stimulus peak was still observed even for 5-s 20-Hz stimulation in both anesthetized and awake conditions (Fig. 1c). Similar post-stimulus hemodynamic overshoot was also described in prior studies involving optogenetic stimulation of SST neurons25, but its underlying source remains unknown.

In summary, prolonged 20-Hz optogenetic stimulation of SST neurons induced a triphasic hemodynamic response at the stimulation site: an initial rapid vasodilation, a subsequent reduction, and a post-stimulus vasodilation.

Optogenetic stimulation of SST interneurons elicits prolonged positive BOLD fMRI responses at the stimulation site, but negative BOLD responses at downstream regions

To investigate the direct translation of OIS findings to fMRI, we subsequently conducted BOLD fMRI experiments at an ultrahigh field of 15.2 T in response to 20-Hz optogenetic stimulation of SST neurons in S1FL under a lightly anesthetized condition (Fig. 2a). During 20-s right forepaw stimulation, positive BOLD responses were observed in the left S1FL and its projection sites, including left S2, thalamus, and right S1 and M1 (Fig. 2c)49. These active regions of interest (ROIs) (Fig. 2b) were employed to extract the time course in response to optically stimulating SST neurons at S1FL.

Fig. 2: Ultrahigh-field BOLD fMRI responses to forepaw and 20-Hz optogenetic stimulation.

The 20-Hz optogenetic stimulation at the left S1FL induced prolonged positive BOLD responses beyond the 20-s stimulation period at the stimulation site (Fig. 2d, e), consistent OIS observations under Ket/Xyl (Fig. 1c). However, in all downstream areas (Fig. 2d, e), only negative BOLD responses were observed during the 20-s stimulation period (L-Tha/L-Stria/L-S1: −0.21 ± 0.09/−0.19 ± 0.04/−0.36 ± 0.04%, averaged over the 20-s stimulation period, n = 5 mice) and returned quickly to baseline. This suggests that the activation of SST neurons at S1FL inhibits local excitatory activity and subsequently reduces synaptic inputs to downstream areas, resulting in the negative BOLD response. These opposing observations between the stimulation site and downstream areas imply a dissociative role of SST cell activation on neuronal and hemodynamic activity: (1) suppression of excitatory activity in both local and downstream areas, and (2) a local increase in the hemodynamic response.

Optogenetic stimulation of SST interneurons suppresses local excitatory activity

To address whether the activation of SST neurons indeed inhibits local excitatory activity, we recorded electrophysiological signals in response to 20-s optogenetic stimulation with a 16-channel depth electrode under anesthesia (Fig. 3a). Each laser pulse significantly evoked changes in local field potential (LFP, <300 Hz) across channels (averaging from all channels, Fig. 3b). At high frequency domain (>300 Hz), neural activity was identified if the peak exceeded the threshold (Fig. 3c). The activation of SST neurons led to reduced multiunit activity (MUA), which was more pronounced during 20-Hz stimulation (Fig. 3d–g, and Supplementary Fig. 2a, b for 5 Hz and 40 Hz stimulation). Notably, during the baseline period, spontaneous neural activity exhibited a higher neuronal response in middle channels (Fig. 3e).

Fig. 3: Local neuronal activity responding to 20-s optogenetic activation of somatostatin (SST) neurons.

To better characterize MUA during the 20-Hz optogenetic stimulation period for 20 s, changes in MUA (ΔMUA), as shown in Fig. 3f and Supplementary Fig. 2b, were separated into the laser pulse (ON) and the inter-pulse period (OFF) (Fig. 3h and Supplementary Fig. 2c for 5 Hz and 40 Hz stimulation). During the ON period for both 1 Hz and 20 Hz, optogenetic stimulation transiently increased ΔMUA with short latency, indicating the direct activation of ChR2-expressing SST neurons. Conversely, during the OFF period, ΔMUA decreased below baseline levels, more pronounced for 20-Hz stimulation (OFF-MUA at 1/20 Hz: −9.60 ± 1.59/−21.24 ± 3.12, n = 4 mice, p = 0.0049, one-way ANOVA repeated test with Bonferroni comparison) (Fig. 3i and Supplementary Fig. 2d for 5 Hz and 40 Hz stimulation), suggesting the inhibition of local excitatory activity. Overall, these findings suggest that optogenetic activation of SST neurons induces a dissociation, where neuronal activity decreases despite an increase in the hemodynamic response.

Mechanism of the initial hemodynamic response to SST interneuron activation: NO-mediated initial vasodilation

Our next question was how SST interneurons elicit vasodilation despite suppressed excitatory activity. Since a large portion of SST interneurons co-express neuronal nitric oxide synthase (nNOS)50,51,52,53, we first tested the possibility that NO, a potent vasodilator, mediates the observed vasodilation. To investigate this, we applied L-NG-Nitro arginine methyl ester (L-NAME), a non-selective NOS inhibitor, via intracortical injection in the S1FL area (Fig. 4a). Interestingly, L-NAME administration significantly abolished the initial peak of the triphasic pattern (1–5 s, HbT in pre/post-L-NAME: 7.04 ± 1.90/−4.73 ± 0.55%, n = 4 mice, p = 0.0089, paired sample, two-tailed t test), and slightly reduced, albeit insignificantly, the late vasodilation under awake conditions (26–35 s, HbT in pre/post-L-NAME: 19.11 ± 3.04/11.16 ± 2.10%, n = 4 mice, p = 0.1293, paired sample, two-tailed t test) (Fig. 4b, c). With the removal of the initial peak by the NOS inhibitor, vasoconstriction became more apparent, indicating that the SST interneuron-induced GABA-mediated inhibition of excitatory neurons contributes to a negative hemodynamic response at the stimulation site. Notably, NOS inhibitor also reduced the HbT response evoked by forepaw somatosensory stimulation under anesthesia, with a more pronounced effect during the initial period (Fig. 4d–f). As a control using Thy1-GCaMP6f mice that express neuronal Ca²⁺ indicators, we confirmed that saline injection did not modify functional hyperemia or neuronal calcium activity in response to forepaw stimulation (Fig. 4g, h). In summary, the initial hemodynamic responses to both SST and sensory stimulation are at least partially driven by NO release.

Fig. 4: Nitric oxide (NO) mediates the fast vasodilation induced by somatostatin (SST) optogenetic and sensory stimulation.

Mechanism of the delayed hemodynamic response to SST interneuron activation: astrocyte-driven late vasodilation via SST peptide

Then, we questioned whether the late, prolonged vasodilation in response to optogenetic activation of SST interneurons is driven by astrocyte activity. To selectively express Ca2+ indicators in astrocytes, we injected the AAV5-gfaABC1D-cyto-GCaMP6f virus in SST-cre×Ai32 mice at S1FL (Fig. 5a). We simultaneously measured HbT and astrocyte calcium activity responding to 20-s SST stimulation under awake conditions and found that optogenetic stimulation of SST neurons induced a slow astrocytic Ca2+ elevation in conjunction with prolonged vasodilation (Fig. 5b, black arrows indicate the astrocyte Ca2+ bursts). The time course extracted from the stimulation site indicated that SST interneuron-induced post-stimulus hemodynamic response was highly coupled with astrocytic Ca2+ elevation (Fig. 5c; Supplementary Fig. 3c for normalized HbT and Ca2+ responses), which was barely observed in the control mouse group (Supplementary Fig. 3a, b).

Fig. 5: Astrocytes mediate the late vasodilation induced by optogenetic stimulation of somatostatin (SST) neurons through both SST neuropeptide and GABAB signaling.

Subsequently, we tested how SST interneurons elicit astrocytic activity by infusing an SST receptor (SSTR) antagonist (CYN 154806) (Fig. 5d). With CYN 154806 administration, the astrocytic Ca2+ activity was completely removed (26–35 s, astrocytic Ca2+ in pre/post-CYN 154806: 1.11 ± 0.15/−0.12 ± 0.07%, n = 4 mice, p = 0.0022, paired sample, two-tailed t test) (Fig. 5e, f), coupled with a nearly abolished post-stimulus vasodilation (26–35 s, HbT in pre/post-CYN 154806: 13.59 ± 2.08/2.73 ± 0.40%, n = 4 mice, p = 0.0085, paired sample, two-tailed t test). In addition, astrocyte Ca²⁺ activity in response to SST interneuron signaling has been reported to occur through the co-activation of astrocytic SST and GABAB receptors in slice studies38,54. To test this hypothesis, we next blocked GABAB receptors (GABABR) using SCH 50911 infusion, which also suppressed both astrocytic Ca2+ activity and post-stimulus vasodilation (26–35 s, pre/post-SCH 50911: 2.05 ± 0.35/0.02 ± 0.13%, p = 0.0064 for astrocytic Ca2+ and 24.59 ± 3.38/7.07 ± 1.75%, p = 0.0073 for HbT; n = 4 mice, paired sample, two-tailed t test) (Fig. 5g). However, the initial HbT response remained unaffected in both experiments involving SST and GABAB receptor antagonists, indicating that the initial HbT response to optogenetic stimulation of SST interneurons is not related to the astrocyte-mediated process.

An interesting observation is that application of either the SSTR or GABABR antagonist increased vasodilation during the trough period (6–15 s after stimulus onset) at the stimulation site, indicating a reduction in inhibition (see Fig. 5f, g). We also examined whether the surrounding negative response was affected (Fig. 5h). Following antagonist application, HbT in the surrounding area increased (6–15 s, pre/post-CYN 154806: –1.38 ± 0.42%/2.03 ± 0.23%, p = 0.0171, n = 3 mice; pre/post-SCH 50911: –0.23 ± 0.47%/4.12 ± 0.15%, p = 0.0027, n = 3 mice, paired sample, two-tailed t test). Despite the small sample size, the disinhibitory effect induced by the GABAB receptor antagonist SCH 50911 appears larger than that induced by the SSTR antagonist CYN 154806 (change in HbT [post–pre]: 4.35 ± 0.32% for SCH 50911 vs. 3.41 ± 0.65% for CYN 154806, p = 0.2624, two-sample, two-tailed t test). In both studies, changes in HbT responses at both the stimulation site and the surrounding region were similar following antagonist application, suggesting a shared neural source.

In summary, optogenetic stimulation of SST interneurons releases SST neuropeptides and GABA, inducing SSTR- and GABABR-mediated astrocytic calcium activity, which is essential for late, prolonged vasodilation. Unlike the well-investigated excitatory neuron-astrocyte-vasodilation pathway55,56, SST interneuron-driven astrocyte activity represents a distinct and previously unexplored neurovascular mechanism, which may play an important role in functional hyperemia.

The SST interneuron-astrocyte-vascular pathway is crucial in mediating functional hyperemia during prolonged sensory stimulation in awake conditions

Next, we wondered whether the SST neuron-astrocyte-vasodilation mechanism is applicable to the hemodynamic response to sensory stimulation, given that somatosensory inputs recruit diverse types of cortical neurons, including SST interneurons19. For this, we adopted naturalistic whisker stimulation under awake conditions. The stimulation parameters for whisker air puff stimulation were based on a prior astrocyte calcium study35 for comparison: 30 s at 5 Hz, 100 ms duration, and 3.5 psi (Fig. 6a).

Fig. 6: The SST neuron-astrocyte pathway mediates late-phase vasodilation in response to prolonged whisker stimulation in awake mice.

Initially, we examined whether SST neuropeptide is involved in astrocytic activity via G-protein–coupled receptor binding during somatosensory stimulation, as suggested by our SST-ChR2 studies (Fig. 5d–g). To detect elevated SSTR activity in astrocytes, we injected astrocyte-specific genetically encoded fluorescent sensor, AAV5-gfaABC1D-SST1.057, into the barrel field (Fig. 6b). Simultaneous recordings of HbT and SST1.0 fluorescence signals were performed during prolonged somatosensory stimulation. Sustained whisker stimulation induced both SST1.0 signal and biphasic HbT responses (Fig. 6c, d). The HbT response consists of an initial vasodilation followed by a second phase of vasodilation, consistent with previous studies35. This confirms that our implementation of whisker stimulation is appropriate under awake conditions. Interestingly, the increase of HbT response was accompanied by the increase of the SST fluorescent response (time to 30% of peak for SST1.0 and HbT: 1.80 ± 0.23 and 1.05 ± 0.05 s, p = 0.019, n = 4 mice, two-sample, two-tailed t test), which was more localized and prolonged than the HbT response (decay time to 30% of peak for SST1.0 and HbT: 53.50 ± 4.40 and 39.8 ± 1.71 s, p = 0.027, two-sample, two-tailed t test). These findings suggest that astrocytic SSTR activity may play a role in sustaining the late phase of the hemodynamic response evoked by prolonged whisker stimulation.

To investigate the causal role of the SST interneuron-astrocyte-vessel pathway, the following three experiments with whisker stimulation were performed in awake conditions: (1) chemogenetic inhibition of astrocytes (Fig. 6e–h); (2) application of SSTR blockers (Fig. 6i–l), and (3) chemogenetic inhibition of SST interneurons (Supplementary Fig. 4a–c). In the first two experiments, the astrocytic calcium indicator gfaABC1D-cyto-GCaMP6f was injected into the barrel field of wild-type mice for imaging astrocyte Ca2+ activity. Additionally, in the first experiment with astrocyte inhibition, the designer receptors exclusively activated by designer drugs (DREADDs), specifically astrocyte-specific AAV5-GFAP-hM4D(Gi)-mCherry virus, were co-injected to chemogenetically inhibit astrocytes while observing astrocyte calcium activity. In the third experiment, the AAV5-hSyn-hM4Di-DIO-mCherry virus was injected into SST-Cre mice to inhibit SST interneurons. Chemogenetic inhibition was performed by the intraperitoneal (IP) injection of the widely used DREADD actuator clozapine N-oxide (CNO). Whisker stimulation elicited a rapid increase in astrocyte Ca2+ (Astrocyte at pre-CNO, Fig. 6g; Astrocyte at pre-CYN 154806, Fig. 6k), accompanied by a biphasic hemodynamic response (HbT at pre-CNO, Fig. 6h and Supplementary Fig. 4c; HbT at pre-CYN 154806, Fig. 6l).

The direct impact of astrocytes on vasodilation in response to whisker stimulation was examined using chemogenetic inhibition of astrocytes. While the duration and concentration of CNO have been shown to differentially affect astrocytic Ca2+ activity58, we observed that Gi DREADD activation significantly reduced astrocyte Ca2+ signals (AUC of astrocytic Ca2+ activity pre/post-CNO: 165.45 ± 42.63/53.94 ± 13.72, n = 5 mice, p = 0.0204, paired sample, two-tailed t test) (Fig. 6g), with a greater reduction observed in the initial phase of calcium response. As expected, the second phase of the CBV increase in response to whisker stimulation was suppressed, with no further dilation observed (21–30 s, HbT pre/post-CNO: 35.55 ± 4.56/23.35 ± 1.65%, n = 5 mice, p = 0.0199, paired sample, two-tailed t test) (Fig. 6h), although a slight reduction was noted in the initial phase of the hemodynamic response. This observation is consistent with previous astrocyte-driven hemodynamic studies during prolonged whisker stimulation35.

To examine whether astrocyte-driven vasodilation is induced by SSTR/GABABR activity, SSTR antagonists were used (Fig. 6i–l). Similar to chemogenetic inhibition of astrocytes (Fig. 6e–h), blocking SSTRs significantly reduced astrocyte Ca2+ activity (astrocyte Ca2+ AUC pre/post-CYN 154806: 209.26 ± 30.11/100.17 ± 21.82, p = 0.0002, paired sample, two-tailed t test) (Fig. 6k) and the second phase of vasodilation (21–30 s, HbT pre/post-CYN 154806: 22.76 ± 1.24/13.66 ± 4.95%, n = 6 mice, p = 0.0003, paired sample, two-tailed t test) (Fig. 6l). Additionally, we repeated chemogenetic inhibition of SST neurons under awake conditions during whisker stimulation (Supplementary Fig. 4a–c). Blocking SST neurons reduced the second phase of vasodilation (HbT pre/post-CNO: 35.61 ± 7.43/27.81 ± 6.68%, averaged over 21–30 s, n = 5 mice, p = 0.0015, paired sample, two-tailed t test), while the initial vasodilation remained mostly intact.

To further determine the causal effect of the SST interneuron-astrocyte pathway on vasodilation, we calculated the reduction rate of CBV responses due to manipulation in all three experiments in three 10-s time windows (Supplementary Fig. 4d). In the first 10 seconds, there were slight reductions similarly in all three experiments. The strongest suppression occurred in the last 10 seconds of the 30-second stimulation, with no significant differences between groups (one-way ANOVA test, p = 0.2077). The similar reduction observed across three different manipulations suggests that sensory-driven SST activity induces astrocytic calcium activity, which in turn leads to vasodilation.

In summary, our findings indicate that the SST interneuron-astrocyte-vessel pathway plays a critical role in amplifying the hemodynamic response to prolonged sensory stimulation.

The SST interneuron-astrocyte-vascular pathway is essential for inducing the delayed hemodynamic response to sustained sensory stimulation under anesthesia

To compare findings with existing hemodynamic literature on the commonly used forepaw stimulation in the fMRI community under anesthesia (Fig. 2)43, we measured both hemodynamic response and Ca2+ dynamics either in astrocytes or excitatory neurons under anesthesia. Additionally, we examined the effects of an SSTR antagonist (CYN 154806) on these responses during forepaw stimulation. For this, we used two mouse groups: (1) the Astrocyte-GCaMP6f, in which wild-type mice were injected with AAV5-gfaABC1D-cyto-GCaMP6f to express astrocyte Ca²⁺, and (2) the Neuron-GCaMP6f with Thy1-GCaMP6f transgenic mice expressing neuronal Ca²⁺ (Fig. 7a and Supplementary Fig. 5e for histology).

Fig. 7: SST neuron-astrocyte mechanism mediates the late vasodilation observed in response to prolonged forepaw stimulation under anesthesia.

In both groups, 20-second forepaw stimulation elicited a bimodal hemodynamic response with initial and later components (HbT response in Fig. 7c, e), as well as fast neuronal and astrocytic calcium activity in the localized S1FL (Fig. 7b). During the 20-second forepaw stimulation, astrocytic Ca2+ levels increased gradually (Astrocyte Ca2+ in Fig. 7c), with a time to peak of 14.9 ± 1.7 seconds (n = 6 mice, Supplementary Fig. 5a), while neuronal Ca2+ responded quickly (Neuron Ca2+ in Fig. 7e) and peaked at 5.05 ± 2.05 seconds (n = 6 mice, Supplementary Fig. 5a).

After applying an SSTR antagonist in Astrocyte-GCaMP6f mice, forepaw stimulation-evoked astrocytic Ca2+ activity was strongly suppressed (11–20 s, astrocytic Ca2+ in pre/post-CYN 154806: 0.45 ± 0.04/0.09 ± 0.11%, n = 6 mice, p = 0.0124, paired sample, two-tailed t test) (black traces in Fig. 7c). Along with the decrease in the astrocytic Ca2+, the later component of the hemodynamic response to the 20-second stimulation was predominantly suppressed (11–20 s, HbT in pre/post-CYN 154806: 10.4 ± 1.10/4.47 ± 1.41%, n = 6 mice, p = 0.0272, paired sample, two-tailed t test) (black traces in Fig. 7c), similarly around the peak (see also Supplementary Fig. 6a, b). In contrast, the initial component, as well as the hemodynamic response induced by the 5-second stimulation (Supplementary Fig. 5b), were still preserved.

Since the SSTR antagonist blocks SSTRs in both astrocytes and neurons, the decrease in astrocytic Ca2+ activity may be attributed either to the direct inhibition of astrocytic SSTR38 or to an indirect pathway involving excitatory neurons59. If the excitatory neuron-to-astrocyte pathway contributed to this effect, blocking SSTRs in excitatory neurons would be expected to increase neuronal calcium activity60. To assess this possibility, we measured excitatory neuronal calcium activity in response to forepaw stimulation, both with and without the presence of SSTR antagonists (Fig. 7b). Although the reduction in excitatory neuronal calcium activity was not statistically significant (Fig. 7e, f), a trend was observed at a later time point (11–20 s, neuronal Ca2+ in pre/post-CYN 154806: 0.20 ± 0.05/0.10 ± 0.06%, n = 6 mice, p = 0.0837, paired sample, two-tailed t test, see also Fig. 7e and Supplementary Fig. 5c). This suggests that the SST interneuron → excitatory neuron → astrocyte pathway is unlikely, whereas the SST interneuron → astrocyte → excitatory neuron pathway remains feasible.

By comparing calcium and HbT responses with and without SSTR antagonist treatment, the relationship between excitatory/astrocytic calcium activity and cell-type-specific HbT responses to sensory stimulation can be assessed. In the presence of the SSTR antagonist, the remaining astrocytic Ca2+ activity closely follows the post-CYN neuronal Ca2+ activity (Astrocyte Ca2+, Neuronal Ca2+ in post-CYN 154806, Fig. 7g, black time course). Consequently, the corresponding hemodynamic response primarily reflects excitatory neuronal activity (HbT in post-CYN 154806, Fig. 7g, black time course). The difference in astrocytic Ca²⁺ activity before and after SSTR antagonist administration reflects SSTR-driven astrocyte activation, which in turn contributes to the differential HbT responses to forepaw stimulation observed pre- and post-SSTR antagonist treatment (Fig. 7g, pink time courses). The hemodynamic response function (HRF) was determined from HbT and astrocyte/neuronal calcium activity (Fig. 7h). We found that the HRF of SST interneuron-astrocyte activity was slower and broader than that of excitatory neurons (peak time = 6.0 s vs 2.0 s and full width at half maximum (FWHM) = 7.0 s vs 5.0 s for HRF of SST interneuron-astrocyte activity vs excitatory neurons, respectively).

Overall, the SST interneuron-astrocyte process is critical for inducing the late hemodynamic response to prolonged somatosensory stimulation, both under awake and anesthesia conditions, similar to the observation made with the optogenetic stimulation of SST interneurons under awake conditions.

SST interneuron activity in response to somatosensory stimulation induces an early fMRI response, followed by a delayed, enhanced laminar-specific fMRI response

Our next question was whether and how the SST interneuron-astrocyte-vasodilation mechanism is related to the fMRI signal. High-resolution fMRI is gaining popularity due to the increasing availability of ultrahigh magnetic fields that enhance sensitivity and spatial resolution. A key issue is how accurately submillimeter-resolution fMRI can localize sites of neuronal activity, which has been investigated using well-established cortical column and layer models42,61,62,63,64,65. Contrary to the expectation that early vasodilation would be the most specific, the specificity of CBV-weighted response actually improves over time41,66,67,68. An interesting observation during sensory stimulation is that the early CBV fMRI response appears in the upper cortical layers, while the later response becomes more specific to layer 4, the thalamocortical input layer62,67,69,70,71,72. However, the neuronal source underlying this spatial specificity has remained unclear. Thus, we hypothesize that the SST interneuron-astrocyte-vessel pathway is the origin of the layer-specific CBV response.

To test our hypothesis, we conducted high-resolution fMRI experiments during 20-second forepaw somatosensory stimulation at an ultrahigh field of 15.2 T, both before and after chemogenetic silencing of SST interneurons, under anesthesia (Fig. 8a). The AAV5-hSyn-hM4Di-DIO-mCherry DREADD virus was injected in the left somatosensory cortex of SST-cre mice, and its transfection was confirmed by histology (Fig. 8b). As a control, the AAV-hSyn-DIO-mCherry virus without hM4Di was similarly injected (Supplementary Fig. 7f). Based on scout multislice BOLD fMRI studies (Fig. 8c), a single 500-μm-thick slice was selected for high-resolution CBV-weighted fMRI studies with a reduced field of view, in-plane resolution of 125 × 125 μm2, and temporal resolution of 1 s following intravenous (IV) injection of iron oxide nanoparticles (MION). Note that we switched anesthetics from ketamine/xylazine to dexmedetomidine with a low dose of isoflurane (Dex/Iso) due to interactions between ketamine and DREADD actuator CNO. Under Dex/Iso anesthesia, the post-stimulus prolonged hemodynamic response was shortened, but the reduction of the late hemodynamic response by the SSTR blockade was confirmed (Supplementary Fig. 7a).

Fig. 8: Chemogenetic inhibition of somatostatin (SST) neurons reduces late-phase laminar specificity of CBV-weighted fMRI responses to prolonged somatosensory stimulation.

The group-averaged activation map showed that the CBV response to somatosensory stimuli before the injection of CNO (pre-CNO CBV response) was more specific to middle cortical layers compared to post-CNO CBV response (Fig. 8c). CBV time courses extracted from the identical active ROI showed that chemogenetic inhibition of SST interneurons reduced the CBV response to forepaw stimulation (Fig. 8c). To visualize laminar-dependent spatiotemporal CBV responses, cortical flattening was applied to the forelimb somatosensory cortex (Fig. 8d and Supplementary Fig. 7b, comparing histology and laminar demarcation), and group-averaged time-dependent flattened functional maps were plotted (Fig. 8e, f). Before the CNO injection, time-dependent CBV maps and profiles (Fig. 8g, h) show that forepaw stimulation quickly induces CBV responses (time to 10% peak at layer 2/3 & 4 in pre-CNO: 0.4 s & 0.3 s, see also Supplementary Fig. 7c, d for layer-specific CBV time courses and Supplementary Movie 1). During the initial period (1–10 s), the CBV response in layer 2/3 exhibits a clear initial peak (peak value: 2.3 ± 0.7%, n = 8 mice, time to peak: 6.0 s, Supplementary Fig. 7d), which closely resembles optical imaging data due to limited depth penetration in superficial layers. In this phase, the CBV response in layer 4 (Supplementary Fig. 7e for hM4Di pre-CNO group, 1.54 ± 0.28%, n = 8 mice; Supplementary Fig. 7f for control pre-CNO group, 1.13 ± 0.64%, n = 3 mice) is slightly higher than in layer 2/3 (0.89 ± 0.17%, n = 8 mice for hM4Di pre-CNO group; 0.78 ± 0.59%, n = 3 mice for control pre-CNO group).

At later time points, such as 10 s after the onset of forepaw stimulation, CBV responses increased further and were predominantly localized in layer 4, rather than the superficial layers (late CBV response in layers 2/3 and 4: 1.98 ± 0.51% and 3.43 ± 0.73%, respectively, averaged over 11–20 s, n = 8 mice for hM4Di group, one-way ANOVA repeated test, multiple comparison with turkey, p = 0.0009) (pre-CNO, Fig. 8g, i). A similar pattern was observed in the control group (Supplementary Fig. 7f), with late CBV responses of 1.30 ± 0.16% in layers 2/3 and 2.38 ± 0.50% in layer 4 (n = 3 mice, one-way ANOVA repeated test with Turkey comparison, p = 0.0065). These time-dependent CBV responses were consistent with previous layer-dependent fMRI studies in anesthetized cats41.

After inhibition of SST interneurons with CNO injection, the CBV responses were reduced and less localized at layer 4 (post-CNO, Fig. 8f, see also Supplementary Movie 2). The initial peak of the CBV response in layer 2/3 was reduced (see +5 s map, Fig. 8f and 1–5-s black-colored profile, Fig. 8h), consistent with forepaw stimulation-evoked HbT responses in the presence of a NOS inhibitor (L-NAME) (Fig. 4f), likely due to inhibition of SST cells that are nNOS-positive. More importantly, after CNO injection, the later peak response in layer 4 was strongly suppressed (late CBV response in layer 4 pre/post-CNO: 3.43 ± 0.73/2.34 ± 0.26%, averaged over 11–20 s, n = 8 mice, p = 0.0261), while the response in layer 2/3 remained similar (late CBV response in layer 2/3 pre/post-CNO: 1.98 ± 0.51%/2.21 ± 0.36%, n = 8 mice, not significant) (Fig. 8g–i). As a result, the peak of the post-CNO CBV response shifted toward the upper cortical layers (black profiles in Fig. 8h). In non-DREADD control experiments with CNO injection, high CBV specificity was preserved at layer 4 (Supplementary Fig. 7f), indicating that the DREADD actuator CNO did not interfere with hemodynamic responses to sensory stimulation. These findings suggest that SST interneuron activity is crucial for inducing laminar-specific vasodilation in thalamic input layers during prolonged somatosensory stimulation.

In summary, activation of SST interneurons in response to prolonged sensory stimulation is essential for both the early hemodynamic response and the late, laminar-specific vasodilation via astrocytes, which shapes the specificity of CBV response in active cortical layers.

Discussion

Our investigations reveal the underlying mechanisms by which SST interneurons control the hemodynamic response. Both optogenetic activation of SST neurons and somatosensory stimulation engage two processes in vascular control: an initial response mediated by NO and/or excitatory neurons, and a later response involving astrocytes (Fig. 9 for schematic representations of the proposed vasodilation pathways). The SST neuron-astrocyte process plays a key role in mediating slow, specific vasodilation during prolonged sensory stimulation, enhancing the laminar specificity of CBV-weighted fMRI signals in layer 4. These findings address previously unanswered questions regarding interneuron-astrocyte-vessel regulation and the cellular basis of CBV-weighted fMRI specificity, especially in the increasingly used human studies62,69,73,74,75.

Fig. 9: Schematic summary of vascular control during neural activation.

Short vs. prolonged stimulation for NVC

In most NVC research, short stimulation durations (<5 s) have been commonly used23,25,76. For short (5-second) SST optogenetic stimulation, vasodilation was primarily driven by NOS neurons that also co-expressed SST during the stimulation period (Fig. 1), consistent with previous findings24,25. However, in fMRI studies, longer stimulation durations (>10 s) are often used due to the limited temporal resolution of fMRI. Unlike the CBV response to 5-second stimulation, triphasic responses were observed with longer stimulation. Consequently, findings from short stimulation are insufficient to explain the sustained vasodilation observed during 20-second stimulation. Although it is reported that the later vasodilation is mediated by astrocyte activity via the glutamate pathway35, evidence also supports that SST or GABAB receptor-mediated signaling evokes astrocyte activity38,54,60,77,78,79 (see also Fig. 5). However, the role of the SST neuron-astrocyte-vessel pathway in functional hyperemia remains unexplored until now.

Initial vasodilation is induced by NOS neuron activity

NO, a potent vasodilator released by neuronal NO synthase (nNOS, also known as Nos1), is widely recognized as a major contributor of early hemodynamic responses to sensory stimulation80,81. NOS neurons are primarily GABAergic interneurons that co-express SST50, with a minor population being glutamatergic excitatory neurons (<13%)82. Optogenetic activation of NOS neurons increases blood flow and volume24,25,83, while pharmacological inhibition of nNOS suppresses the initial vasodilation induced by optogenetic stimulation of SST interneurons (Fig. 4). During sensory stimulation, nNOS neurons, which have close contact with arterial vessels, are activated by the binding of glutamate to NMDA receptors84. Thus, feedforward excitatory activation induces nNOS neurons, leading to dilation of adjacent arterioles via diffused NO83. The initial hemodynamic responses upon forepaw stimulation were significantly diminished by NOS inhibitor (Fig. 4f). Additionally, a slight suppression of late vasodilation was observed following nNOS inhibition during both sensory and optogenetic stimulation (Fig. 4), indicating that NO may also contribute to sustained hemodynamic responses. Further investigation is warranted to elucidate the precise role of NO in these processes.

CBV-weighted fMRI studies under Dex/Iso conditions (Fig. 8c) revealed that the reduction of CBV responses to somatosensory stimulation following chemogenetic inhibition of SST interneurons was depth-dependent, being negligible in layer 2/3 but substantial in layer 4 (Supplementary Fig. 7c, d). Additionally, under awake conditions, chemogenetic inhibition of SST neurons did not reduce the initial vasodilation evoked by whisker stimulation (Supplementary Fig. 4c). These slight discrepancies can be likely explained by differences in brain state and the distinct properties of imaging modalities: OIS is heavily weighted toward upper cortical layers without depth discrimination, whereas CBV-weighted fMRI cannot detect pial vessel dilations85. Overall, further investigation is needed to determine the cellular source of the initial hemodynamic response to sensory stimulation.

Role of NPY vs. inhibition induced by SST interneurons

NPY induces vasoconstriction in brain slices22 and in the somatosensory cortex after the offset of sensory stimulation23. A subset of SST neurons is NPY-positive53,86, suggesting that the NPY release from optogenetic stimulation of SST neurons may lead to vasoconstriction specifically at the stimulation site, rather than in downstream areas. The reduction in vasodilation following the initial vasodilation during SST neuron stimulation may be attributed to vasoconstrictive NPY. In our data, activation of SST neurons induces local inhibition, resulting in a reduction in MUA response, a negative BOLD response in downstream areas, and a negative hemodynamic response in the surrounding area of the stimulation site. The reduction in hemodynamic responses at the stimulation site after the initial peak depends on the frequency and duration of optogenetic stimulation, as well as the brain’s state, with a greater reduction observed at higher stimulation frequencies under awake conditions (Fig. 1 and Supplementary Fig. 1). The vasoconstriction at downstream sites can be explained by reduced synaptic inputs resulting from the inhibited excitatory activity at the upstream stimulation site. Widespread vasoconstriction was observed after inhibiting nNOS (Fig. 4b). The reduced hemodynamic responses at the stimulation site after the initial peak likely reflect local inhibition rather than local vasoconstrictive NPY release from SST neurons. However, the role of NPY released by the subpopulation of SST neurons still requires further investigation.

Activation of SST neurons elevates astrocytic Ca2+, accompanied by prolonged vasodilation

SST neuropeptides released from SST interneurons bind to G-protein-coupled receptors (Gi/o) in neurons, similar to GABAB receptors (GABABR), reducing cellular calcium activity in neurons87,88,89,90. However, in contrast to their inhibitory role in neurons, these Gi/o activation leads to Ca2+ elevation in astrocytes60. Indeed, both SST and GABAB enhance astrocytic calcium activity (Fig. 5d–g), which is consistent with previous studies on astrocyte calcium dynamics38,78,79. In our previous studies of PV interneurons43, only a slight reduction in astrocytic calcium activity, rather than an increase, was observed, indicating that GABABR activation alone is insufficient to drive astrocyte activity. The specific sensitivity of astrocytes to SST, rather than to PV neuron activation38, suggests a unique SSTR-mediated mechanism. The co-release of GABA and SST peptides activates GABABR-SSTR complexes on astrocytes38, synergistically enhancing Ca2+ elevation. Late astrocyte activity via SST/GABAB receptors induces the late component of the hemodynamic response to prolonged sensory stimulation (Figs. 6f–l and 7b, c).

Early astrocyte calcium activity in response to somatosensory stimulation is not likely induced by SST neurons under anesthesia, as indicated by the persistence of this response after SSTR blockage (Fig. 7c). The rapid astrocyte response is generally assumed to be mediated by neuron-to-astrocyte signaling via mGluRs or NMDA receptors during sensory stimulation35,55,91,92. Our observations suggest that after SSTR blockade, astrocytic responses are likely mediated through the mGluR pathway, given the similar temporal dynamics of excitatory neuronal and astrocytic calcium activity (Astrocyte Ca2+ at post-CYN 154806, Fig. 7g). However, in our awake measurements using whisker stimulation (Fig. 6k, l), SSTR inhibition reduced both astrocytic Ca2+ and HbT responses during both early and later phases, with a more pronounced reduction in the late phase. The reason for the difference in the reduction factor of early HbT responses following suppression of the SST-astrocyte pathway during anesthetized vs. awake conditions remains unclear, which may be related to neuromodulator influences associated with brain status. Further investigation is warranted to determine the dependency of the SST neuron-astrocyte pathway on neuromodulators and to elucidate the extent to which excitatory neuron-astrocyte interactions contribute to vasodilation in a spatiotemporally distinct manner.

Increased hemodynamic layer-specificity at a later time to the synaptic input layer

During prolonged sensory stimulation, the hemodynamic response is characterized by an initial fast vasodilation followed by a secondary delayed vasodilation. The fast vasodilation, which is less specific to thalamocortical input (layer 4), is also detected in layer 2/3 of laminar fMRI studies62,63 and is potentially mediated by NO-related pathways and endothelial hyperpolarization81,93 (Fig. 9b). These mechanisms are known to strongly impact the dilation of penetrating and branching arterioles15,81,84,94. The secondary, and delayed vasodilation becomes more specific at layer 4, likely involving capillary dilation primarily controlled by astrocytes15,35 (Fig. 9b).

When SST neurons were suppressed during sensory stimulation, the CBV response was delayed and peaked in the upper cortical layers during the 20-second stimulation period, disrupting the typical layer 4-specific localization. The spatiotemporal differences in CBV responses observed before and after chemogenetic inhibition of SST interneurons suggest a critical role for SST cell-related vascular control. The later response supports a pathway in which SST neurons amplify both the amplitude and spatial localization of the hemodynamic response through astrocyte-mediated mechanisms. Delayed layer 4-specific hemodynamic regulation may be interpreted as excitatory neurons with the highest synaptic inputs driving nearby SST neurons in a facilitative manner, which leads to astrocyte activity and consequently a hemodynamic response. The spatial localization to layer 4 may also be related to the laminar-dependent distribution and regulation of SST neurons and astrocytes in the somatosensory cortex95,96. Further studies are necessary to determine whether the SST neuron-astrocyte-vessel pathway is universally involved in laminar CBV specificity across different cortical areas and layers.

Implications of SST-driven hemodynamic responses to clinical neuroimaging research

Unlike other subtypes of inhibitory neurons, SST interneurons mainly receive inputs from nearby excitatory neurons in a facilitating manner and inhibit these excitatory neurons by targeting their distal dendrites17,19,97,98,99,100, thereby shaping sensory responses21,101. Thus, SST cells are recruited more effectively by sustained sensory stimulation compared to phasic sensory stimulation.

SST neurons are reduced in various neurodegenerative and neuropsychiatric disorders, including Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, major depressive disorder, bipolar disorder, and schizophrenia21,102. For example, SST expression in the frontal cortex and hippocampus is reduced to <30% in AD patients compared to control subjects103. A detailed review of SST interneurons in neurological diseases can be found in the paper by Song et al.102. When the overall function of SST interneurons is altered in disease conditions or different brain states, hemodynamic responses to prolonged external stimuli are likely to be affected. Therefore, investigating the role of SST neurons in vascular responses within neurological disorders will be crucial.

Potential limitations

In our studies, several potential issues need to be discussed. (1) In our neural data with 40 Hz optogenetic stimulation of SST neurons (Supplementary Fig. 2), a neural rebound was observed after stimulus offset. Since MUA is slightly delayed relative to light stimulation and persists104,105, the exact source of the delayed rebound needs to be further investigated. (2) Calcium imaging and optogenetic stimulation use similar wavelengths; thus, 470-nm illumination for calcium imaging can inadvertently activate SST neurons and increase HbT throughout the imaging session (Fig. 5). However, HbT responses to 473-nm optogenetic stimulation are correctly detected by measuring HbT changes relative to baseline. (3) Astrocyte activity induced by optogenetic stimulation of SST interneurons (Fig. 5) was much slower than that triggered by SST neurons during somatosensory stimulation (Fig. 6). The delay in astrocyte activity in response to SST optogenetic stimulation might be due to SST neuron-induced inhibition of local excitatory activity (see Fig. 9a), potentially lowering intracellular Ca2+ stores via metabotropic glutamate receptors (mGluRs). Consequently, higher concentrations of SST peptides may be required to induce astrocytic calcium activity. These dynamic differences in SSTR-driven astrocytic calcium activity warrant further investigation. (4) Activation of inhibitory G-protein-coupled receptors (Gi) by SST neuropeptides increases astrocytic Ca²⁺ activity under normal conditions (Figs. 5–7). However, when baseline Gi activity is elevated via astrocytic DREADD activation by an agonist, subsequent Ca²⁺ responses to somatosensory stimulation may be attenuated due to saturation effects (see Fig. 6g). However, we did not directly compare baseline Ca²⁺ activity before and after DREADD activation. (5) Blocking SSTRs reduced neuronal Ca2+ activity at a later time point, but not at an earlier time point under anesthesia (Fig. 7e). Given that SSTR blockade is generally expected to enhance neuronal activity rather than suppress it (see Fig. 9a), the observed reduction in neuronal Ca²⁺ activity is more likely due to a disruption in astrocyte-neuron feedback at the later time point36 (Fig. 9b, gliotransmitters). This astrocyte–excitatory neuron pathway may play a role in modulating HbT responses to somatosensory stimulation. Further investigation is needed to clarify its precise contribution. (6) Different anesthesia mixtures (ketamine/xylazine, dexmedetomidine/isoflurane) were used in addition to awake conditions. Hemodynamic responses to sensory stimulation under ketamine/xylazine anesthesia (Figs. 1 and 7) were prolonged compared to those under dexmedetomidine/isoflurane anesthesia or awake conditions (see Fig. 6 for awake conditions and Supplementary Fig. 7a for dexmedetomidine/isoflurane anesthesia). Variations in hemodynamic responses across different brain states can be attributed to multiple factors, including the effects of anesthesia on neural activity and NVC, as well as neuromodulatory influences. The role of neuromodulators such as acetylcholine and serotonin should be further investigated22. (7) Unlike optical imaging studies with various manipulations, high-resolution CBV-weighted fMRI was only performed with chemogenetic inhibition of SST interneurons. It would be interesting to measure depth-dependent BOLD and CBV fMRI responses with cell-type-specific manipulations under awake conditions to further elucidate the role of different neuronal populations in laminar-dependent NVC. (8) In our somatosensory stimulation, SST activity contributes to laminar-specific vasodilation. However, our findings in the primary somatosensory cortex may not be directly translatable to other cortical regions or cognitive tasks. Further studies are needed to determine whether this mechanism is a general feature of prolonged tasks across different brain regions and functional contexts.

Methods

Animals

All animal experiments were approved by the Institutional Animal Care and Use Committee of Sungkyunkwan University (protocols SKKUIACUC2023-03-39-1, SKKUIACUC2024-07-51-1, and SKKUIACUC2025-04-106-1), and were conducted in compliance with the Animal Welfare Act (AWA), National Institutes of Health Guide for the Care and Use of Laboratory Animals, and ARRIVE guidelines 2.0 (Animal Research: Reporting in Vivo Experiments)106. Animals were kept on a normal 12–12 h light and dark cycle (9 am: on, 9 pm: off). Room temperature and humidity were set to 22–24 °C and 48–62%, respectively. Animals had access to food and water ad libitum, and were group housed. SST-cre and Ai32 mice (both male and female, 20–25 g, 8–12 weeks old) were bred in-house from breeding pairs originally acquired from Jackson Laboratory (Bar Harbor, ME, USA) (strains: SST-cre, SSTtm2.1(cre)Zjh/J, Stock number: 013044; Ai32, B6.Cg-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFPHze)/J, Stock number: 024109). Ai32 mice with expression of ChR2/EYFP were crossed with SST-cre mice, resulting in SST-ChR2 mice. For measurement of neuronal Ca2+ activity, we acquired C57BL/6J-Tg(Thy1-GCaMP6f)GP5.17Dkim/J (both male and female, 20–25 g, 8–12 weeks old, Stock number: 025393) from Jackson Laboratory and bred them in-house. Wild-type mice (C57BL/6, male, 25–30 g, 8 weeks old) were purchased from a local vendor (Orient Bio, Seongnam, Korea). Both male and female adult mice (6–8 weeks) were used for experiments. In this study, we used a total of 93 mice (SST-ChR2: 39; SST-Cre: 16; C57BL/6: 29; Thy1-GCaMP6f: 9). The number of animals used for each experiment is provided in Supplementary Table 1.

Injection of AAV virus for astrocytic calcium imaging, chemogenetic inhibition, and SSTR sensing

The mouse was anesthetized with isoflurane (2.5%), placed in a stereotaxic frame (SR-10R-HT, Narishige, Tokyo, Japan). Body temperature was maintained at 37 °C and the scalp was removed after injection of analgesics. The craniotomy (0.2 mm diameter) was performed on the left primary somatosensory forelimb area (AP: +0.05, ML: +2.2 mm) or barrel field (AP: −2, ML: 3 mm). For virus delivery into the primary somatosensory cortex, a small volume of virus (0.5 µL) was infused at a rate of 50 nL/min through a glass pipette connected to a micro-injection pump (Harvard Apparatus, United States) at a depth of 300 and 600 µm about 3–4 weeks before experiments. For astrocytic calcium imaging, AAV5-gfaABC1D-cyto-GCaMP6f (Addgene, United States; catalog number: #52925) was injected in SST-ChR2 or wild-type mice. For chemogenetic inhibition of SST neurons, AAV5-hSyn-DIO-hM4D(Gi)-mCherry virus (Addgene, #44362) or AAV5-hSyn-DIO-mCherry (control) was injected in SST-cre mice. For chemogenetic inhibition of astrocytes, AAV5-GFAP-hM4D(Gi)-mCherry (Addgene, #50479) was injected in wild-type mice. For detecting astrocytic SST neuropeptide signals, we injected AAV5-gfaABC1D-SST1.0 (BrainVTA, China; catalog number: PT-7454, with kind approval by Prof. Yulong Li, Peking University) in wild-type mice57.

Thinned-skull procedure and optic fiber implantation

For optical imaging studies, the thinned-skull procedure was performed by a handheld drill under isoflurane (2%). A thin layer of cyanoacrylate glue (Loctite 401, Henkel) was applied to the entire skull surface. Meloxicam (1 mg/kg) was injected after surgery to avoid inflammation. Then, the mouse was recovered from anesthesia and returned to its cage. For optogenetic fMRI studies, an optic fiber (105 µm in diameter; Thorlabs) was implanted in the left somatosensory area (L-S1) at a depth of 100–200 µm. Then, a biocompatible silicone elastomer (Kwik-Sil, World Precision Instruments, Sarasota, FL, USA) was applied to enclose the fiber implantation site, and dental cement (SB, Sun-Medical Co., Shiga, Japan) was then applied to thinly cover the area around the fiber cannula to fix it onto the skull.

Histology

The mice were deeply anesthetized with isoflurane, and then transcardially perfused with saline and 4% formaldehyde (PFA) at a rate of 3 ml/min. The brain was carefully extracted, post-fixed in 4% PFA overnight at 4 °C, and cryoprotected in a 30% sucrose solution for 3 days. Then, the 40-μm-thick coronal sections were serially cut using a cryostat (CM1950, Leica Biosystems) and transferred to 0.1 M PBS. The sections containing the S1FL or S1 barrel field were incubated in 10% donkey serum (00-8120, Invitrogen) for 1 h at room temperature. Next, the sections were incubated overnight at 4 °C with primary antibodies against the following proteins: S100β (1:800, S2532, Sigma), and SST (1:1200, T4103, BMA Biomedicals). After washing with PBS three times, the sections were incubated with the appropriate fluorescence-conjugated secondary antibody (1:400) for 1 h at room temperature, and then washed in PBS three times. Fluorescence images were acquired with a TCS SP8 confocal microscope (Leica Microsystems) and a ×20 or 40× objective lens. The images were analyzed with ImageJ and Imaris (Bitplane, RRID:SCR_007370) software.

Forepaw, whisker air puff or optogenetic stimulation

Sensory or optogenetic stimulus was controlled by a pulse stimulator (Master 9, World Precision Instrument). For forepaw stimulation, electrical stimulation was performed on the right palm via two needles inserted under the skin with 0.5 mA current and 5 Hz107, generated by a current generator (ISO-Flex, AMPI). For whisker air puff stimulation, the air was delivered by picospritzer III (Parker Hannifin, USA), triggered by Master 9 at 5 Hz, 100 ms and 3.5 psi. To activate light-gated ion channel ChR2, photo stimulation was done with a 473 nm blue laser (MBL-III-473, Changchun New Industries Optoelectronics Tech), delivered at 1, 5, 20 and 40 Hz with 200, 40, 10 and 5 ms pulse width, respectively. The laser power at the fiber tip (200 μm diameter for optical studies and 105 μm diameter for fMRI studies) was 3.0 mW, measured by a power meter (PM100D, Thorlabs, United States).

Intracortical drug infusion

Before OIS experiments, a 0.2 mm in diameter craniotomy was made on the thinned-skull mouse. The dura was pierced by a micro needle. Then, 0.5 µL of NOS inhibitor (L-NAME:100 mM; Sigma-Aldrich, product number: N5751), selective somatostatin receptor antagonist (CYN 154806: 20 µM; Abcam, product number: ab154806), selective GABAB receptor blocker (SCH 50911: 10 mM, Tocris, product number: 0984) or saline was intracortically infused by a micro-injection pump (Harvard Apparatus, United States) via a glass pipette at a rate of 50 nL/min.

Chemogenetic inhibition of SST neurons or astrocytes with CNO injection

The DREADD actuator CNO was diluted with DMSO (1 mg/100 µl). CNO-DMSO solution was mixed with saline with a ratio (1:19)108. CNO was applied via a single IP bolus injection (CNO dose: 5 mg/kg).

Anesthesia, wakefulness, and animal maintenance

For awake optical experiments, all mice were habituated to head-fixation on a disk treadmill for about 15 min for acclimation before imaging. For ketamine/xylazine condition, all mice were initially anesthetized with 5% isoflurane in oxygen and air and then given an induction dose with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) intraperitoneally107,109. During experiments, supplementary anesthetic doses (25 and 1.25 mg/kg of ketamine and xylazine, respectively) were injected every 30–40 min via an IP line107,110.

For layer fMRI studies with chemogenetic inhibition of SST neurons, an IV bolus of 0.05 mg/kg (diluted to 0.05 mg/mL) of dexmedetomidine was given, followed by the discontinuation of isoflurane. Ten minutes after IV bolus, IV infusion of 0.05 mg/kg/h was started. Isoflurane was sustained at 0.3–0.5%110.

The mouse’s body temperature was maintained at 37 °C by an electrical (or hot-water) heating pad for optical imaging and electrophysiology (or MRI experiments). A mixture of oxygen and air (ratio 3:7) was delivered continuously by a small animal ventilator (SAR-1000, CWE, Ardmore, USA or TOPO, Kent Scientific Corporation, Torrington, CT, USA) via a nose cone at a rate of 1 L/min. Temperature and respiration rates were continuously monitored throughout the experiments with an animal monitoring system (Physiosuite, Kent Scientific Corporation, USA; Model 1030, Small Animal Instruments, Stony Brook, USA) and recorded with a data acquisition system (Acknowledge, Biopac Systems).

Electrophysiology under anesthesia

Under anesthesia, a craniotomy of 2 mm in diameter was made over the left S1FL area identified using 530-nm CBV-like OIS images responding to right forepaw stimulation. A 16-channel opto-electrode with 50 μm spacing between channels (A1x16-5 mm-50-177-OA16LP, NeuroNexus, Ann Arbor, MI, USA) was slowly inserted to a depth of 1000 μm. Data were sampled at 30 kHz using a neural data acquisition system (Cerebus™, Blackrock Microsystems, Salt Lake City, Utah USA). Five trials were repeated for each stimulation condition, and averaged for each mouse.

Multi-wavelength optical imaging: intrinsic imaging, calcium imaging, and fluorescence imaging

Optical imaging was performed by a home-built multi-wavelength system43,44. The system used two TTL-controlled light-emitting diodes centered at 530 nm (CBV-like) and 625 nm (BOLD-like) (M530L4 and M625L4 coupled with LEDD1B driver, Thorlabs Inc. Newton, USA) or 470 nm (M470L4 for calcium and SST1.0 excitation) to alternatively illuminate the dorsal cortex through the thinned skull, which was controlled via a data acquisition card (Arduino board). A high-resolution sCMOS digital camera (Andor Zyla 4.2, Oxford Instruments) acquired high-speed images through a 485 nm longpass filter (DM485, Olympus) to prevent the saturation of the camera by 473 nm photo-stimulus. CBV-like 530-nm and BOLD-like 625-nm (or fluorescent 470-nm) optical images were obtained at 10 or 20 frames per second over a 13 × 13 mm2 field of view (FOV) with a 2048 × 2048 matrix size. Photo-stimulus was interleaved between optical frames, and a 10-ms delay between laser pulse and camera frame was used to minimize the effect of 473-nm stimulation laser directly to the CMOS camera. For each experimental condition, at least 10 trials were averaged to enhance the OIS sensitivity.

BOLD and CBV-weighted fMRI

All MRI experiments were performed on a horizontal bore 15.2 T/11 cm Bruker BioSpec MR system (Billerica, MA, USA) with an actively shielded 6-cm diameter gradient operating with a maximum strength of 100 gauss/cm and a rise time of 110 μs. A customized single-loop elliptic 10 × 12 mm2 surface coil positioned on top of the mouse’s head was used. The magnetic field homogeneity was initially shimmed globally using a field map method and then was optimized using the MAPSHIM protocol with an ellipsoid shim volume covering the cerebrum (ParaVision 6, Bruker BioSpin).

The T2-weighted anatomic image was obtained with the rapid acquisition with refocused echoes (RARE) sequence with repetition time (TR)/echo time (TE) = 5300/32 ms. For BOLD fMRI, single-shot gradient echo echo-planar imaging (EPI) was used with the following parameters: FOV = 15 × 7 mm2, matrix size = 96 × 48, in-phase resolution = 156 × 156 μm2, slice thickness of 500 μm, number of slices = 20, bandwidth = 300 kHz, TR/TE = 1000/11 ms, and flip angle (FA) = 30°. For high-resolution CBV-weighted fMRI, a single 0.5-mm-thick slice containing S1FL responding to forepaw stimulation was selected from scout whole-brain BOLD fMRI experiments. With three out-of-volume saturations, CBV-weighted laminar fMRI data were acquired using conventional gradient echo-based imaging (FLASH) with TR/TE = 31.25/3 ms, FA = 15 °, FOV = 9 × 4.5 mm2, matrix size = 72 × 36, partial Fourier = 1.12 along the phase-encoding direction, in-plane resolution of 125 × 125 μm2 and temporal resolution of 1 s, and following the injection of MION (45 mg/kg; Feraheme, AMAG Pharmaceuticals, Waltham, USA) to enhance laminar specificity85.

Data analysis and statistical analysis

Neural data analysis

For spike detection, electrophysiological recording data were analyzed with custom-written MATLAB code. The raw signals were separated into local field potential (LFP) using a lowpass filter (<300 Hz) and MUA using a bandpass filter (300–6000 Hz). For spike detection, the spikes were counted if the peak exceeded three times the standard deviation (SD) of each channel. MUA responses were calculated in 1-s bins.

Separation of ∆MUA During ON and OFF Periods: For 20 Hz (40 Hz) stimulation, despite using a 10 ms pulse (5 ms), corresponding to a 20% duty cycle, a neural rebound was observed immediately after stimulus offset. To account for this, an additional 5 ms of post-stimulus data (2.5 ms for 40 Hz) was included in the ON period, while the remaining 35 ms (12.5 ms for 40 Hz) was assigned to the OFF period. For 1 Hz and 5 Hz stimulation, the ON period was defined as 200 ms and 40 ms, respectively, while the OFF period was set at 800 ms and 160 ms. MUA data from both the pulse and inter-pulse periods were then summed for each 1-s interval.

Optical imaging data analyses

All repeated trials were averaged for each experimental condition. Two-wavelength intrinsic OIS responses were converted to absolute changes of HbT with the modified Beer–Lambert Law43,111,112. For experiments using single-wavelength intrinsic OIS acquisition, relative changes in HbT were estimated from 530 nm OIS signals.

For calcium and SST1.0 imaging, two-wavelength imaging experiments were performed for the correction of hemodynamic contribution to calcium fluorescent signals. The fluorescent signal was corrected for hemodynamic responses using green reflectance only (single-wavelength method) as described by Y. Ma's study113. The correction was done by simply dividing the fluorescent ratio by the green reflectance ratio as (ΔF/F + 1)/(ΔR/R + 1)−1. However, during simultaneous optogenetic stimulation, a boxcar-like artifact—characterized by a reduction in fluorescence signal during the stimulation period—was observed in both AAV-injected and non-injected mice (Supplementary Fig. 3b), even after hemodynamic correction. This artifact is likely due to crosstalk between the blue laser and eYFP in SST-ChR2 mice. To account for this, fluorescent (e.g., astrocytic Ca2+) time courses evoked by optogenetic stimulation were additionally corrected by subtracting the boxcar-like artifact obtained from the control group without AAV injection.

To quantify optical responses, a 1-mm circular ROI was selected around the optogenetic stimulation fiber or in regions with strong responses to forepaw or whisker stimulation. The same ROI was consistently used to extract time courses of responses in each animal.

Determination of hemodynamic response function

The hemodynamic response function (HRF) was modeled as the difference of two gamma density functions114. The HRF equation is expressed as gamma1/max(gamma1)–dip×gamma2/max(gamma2), where ‘dip’ represents the coefficient of the second gamma density (gamma2). For simplicity, this study assumes dip = 0. The first gamma function (gamma1) is characterized by two parameters: the time to peak and FWHM.

fMRI data analyses

All fMRI data were preprocessed and analyzed using readily available software packages: Analysis of Functional NeuroImages115, FMRIB Software Library116, Advanced Normalization Tools117, and Matlab (Mathworks, Natick, MA, USA). As processing steps were reported previously42,49,118,119,120, each functional image volume was preprocessed with slice timing correction and motion correction by aligning every image to the first image of each volume. To generate group-averaged datasets, preprocessed data in the same session were averaged, and linear detrending was performed to reduce signal drifts. Then, spatial smoothing was performed using a Gaussian kernel with an FWHM of 0.3 mm. Spatial normalization was conducted using the following procedure. First, multislice functional EPI images were co-registered to the anatomical T2*-weighted images from the same subject using an affine transformation. Second, the T2*-weighted images of all subjects were normalized and averaged while applying linear- and nonlinear transformations to generate a mouse brain template. The mouse brain template was finally co-registered to the Allen Mouse Brain Atlas. Third, all EPI images co-registered in the first step were normalized to the Allen Mouse Brain Atlas (50 × 50 × 500 μm3) using the co-registration parameters obtained in the second step. Group-averaged activation maps were made by a GLM analysis using a stimulation paradigm convolved with a gamma-variate function. Activation voxels were determined using a threshold of the t value (P < 0.05, family-wise error corrected with the Bonferroni post-hoc test).

For the BOLD quantitative analyses, ROIs were defined based on the Allen Brain Atlas: the left primary somatosensory area (L-S1FL), left secondary somatosensory area (L-S2), right primary somatosensory and motor area (R-S1FL & M1), left thalamus area (L-Tha) and left dorsal striatum area (L-Stria). BOLD time courses were calculated as percent changes relative to the baseline period and extracted from defined ROIs for each subject.

Laminar fMRI analysis (cortical flattening of CBV-weighted fMRI data)

To spatially normalize individual animal data from a laminar fMRI study to common space, a study-specific template was constructed with CBV-weighted fMRI images, following a previously used pipeline119. First, the brain area was semi-automatically extracted from the temporal average of all fMRI images for each animal, and intensity nonuniformity was corrected using bias field estimation. Next, individual fMRI images were aligned to a representative animal image through 2D linear registration and averaged to generate the initial brain template. A study-specific template was then constructed by averaging all images after performing both linear and nonlinear registration to the initial brain template. Then, fMRI preprocessing steps were performed, including motion correction, linear signal detrending, trial averaging and spatial normalization to a study-specific template. Activation maps were obtained, and time courses were obtained from the identical ROI in each animal, defined from the common active voxels before and after CNO injection (black contour ROI shown in Fig. 8c).

To measure laminar-specific fMRI profiles, cortical areas, including the S1FL, were flattened by radially projecting 24 lines perpendicular to the cortical edges. The cortical depth profiles were resampled to double using bicubic interpolation, leading to a nominal resolution of 62.5 µm. Laminar boundaries for each cortical area, belonging to S1FL (yellow ROI, shown in Fig. 8d), were defined as a cortical thickness distribution from the Allen mouse brain atlas, and correlated well with histological staining (Supplementary Fig. 7b). Depth of S1FL was separated into 18 voxel-lines, starting from 62.5 µm (1st voxel-line) to 1187.5 µm (18th voxel-line). Our study defined layer 2/3 from 2nd to 6th voxel-lines (125 to 437.5 µm), layer 4 from 7th–9th (437.5 to 625 µm), layer 5 from 10th–14th (625 to 937.5 µm), and layer 6 from 15th–18th (937.5 to 1187.5 µm) (Supplementary Fig. 7b). Time courses were extracted from voxel-lines within S1FL and averaged within each corresponding layer. Finally, the fMRI signal changes within the same cortical depth were calculated in the same manner described above. To enhance the detectability of laminar responses, four 20-s stimulus blocks (40–20 stim–60–20 stim–60–20 stim–60–20 stim–60 s) were averaged. Group-averaged percent change maps were calculated from the percent change maps of individual animals.

Statistical analyses

In each experimental condition, all repeated measurements were averaged. Time courses and bar graphs in this study are presented as mean values and standard errors of the mean across animals. Statistical tests for measurements within the same animal were conducted using a paired sample two-tailed t test or a one-way repeated-measures ANOVA, followed by Bonferroni or Tukey post-hoc tests to assess the statistical significance of multiple comparisons. To compare measurements across groups, a two-sample two-tailed t test or a one-way ANOVA was performed. A significance level of p < 0.05 was considered statistically significant.

Reporting summary

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

Data availability

All source data are provided with this paper. The fMRI data generated in Figs. 2 and 8 and Supplementary Fig. 7f have been deposited in the Zenodo database under accession code (https://doi.org/10.5281/zenodo.15687724). The raw optical files are stored on servers at the Center for Neuroscience Imaging Research owing to their large size. These raw images can be provided by the corresponding author upon request. Source data are provided with this paper.

name：

Verification code：

My Account

ISO9001：2015

Certificate No.:J22Q28025760R0S

mail: info@enlibio.com mail: sale@enlibio.com Tel:+86-027-87610298

mail: alex@enlibio.com mail: order@enlibio.com Tel:+86-027-87610297

Add:B105,Building C1,R&D Building,Zone B,C,D,Wuhan National Bio-Industry Base,No.666 Gaoxin Avenue,East Lake New Technology Development Zone, Wuhan, China