Psychedelic 5-HT2A receptor agonism alters neurovascular coupling and differentially affects neuronal and hemodynamic measures of brain function

Abstract

Human neuroimaging studies report that psychedelics induce serotonin-2A receptor (5-HT_2AR)-dependent changes in functional brain reorganization, presumably reflecting neuromodulation. However, these studies often overlook the potent vasoactive effects of serotonin. Here, we identified psilocybin-induced alterations in hemodynamic response functions during human fMRI, suggesting potential disruptions in neurovascular coupling (NVC). We then used wide-field optical imaging in awake Thy1-jRGECO1a mice to determine whether psychedelic-induced changes in hemodynamics arise from neuronal, vascular, or neurovascular effects. Exposure to the psychedelic 2,5-Dimethoxy-4-iodoamphetamine (DOI) differentially altered coupling between cortical excitatory neuronal versus hemodynamic activity, both during whisker stimulation and in the resting-state. Furthermore, DOI resulted in discordant changes between neuronal- versus hemodynamic-based assessments of functional connectivity. A selective 5-HT_2AR antagonist (MDL100907) reversed many of the effects of DOI. Our results demonstrate a dissociation between DOI-induced neuronal and hemodynamic signals, indicating a need to consider neurovascular effects of psychedelics when interpreting blood-based measures of brain function.

1). Introduction:

Clinical trials have provided compelling evidence that psychedelics offer rapid and substantial improvement in mood disorders^{1, 2} and substance use disorders³. These remarkable findings have ignited growing interest in elucidating underlying therapeutic mechanisms. Psychedelics, such as psilocybin and lysergic acid diethylamide, elicit profound changes in perception, mood, and cognition⁴, collectively known as the enigmatic “psychedelic experience”. Understanding the neurobiological basis of these effects is a crucial step towards advancing the therapeutic potential of these agents.

Recent high-profile human neuroimaging studies employing functional magnetic resonance imaging (fMRI) have shed light onto the effects of psychedelics on functional brain organization. For example, psychedelics have been shown to enhance brain network integration and increased global resting-state functional connectivity (RSFC)^{5, 6}. These discoveries have led to a burgeoning body of research employing fMRI to examine the acute and persistent network-level effects of psychedelics. fMRI indirectly indexes neural activity through changes in blood-flow and metabolism that are reflected in modulations of blood-oxygenation-level-dependent (BOLD) signaling. This linkage is commonly referred to as neurovascular coupling (NVC)⁷. Through NVC, changes in RSFC are frequently interpreted as reflecting altered interregional neuronal synchrony. However, it remains unclear whether psychedelic-induced changes in RSFC are a result of neuronal activity, alterations in vascular tone, NVC, or some combination thereof.

The etymology of serotonin (‘serum’+’tone’) reflects its discovery as a substance in blood serum that strongly modulates vascular tone⁸. It was later found to exert potent neuromodulatory effects in the brain, influencing, for example, mood, sleep, and memory⁹. Psychedelic-induced hallucinations depend on activity of 5-HT receptors, primarily at the 5-HT_2A receptor (5-HT_2AR)¹⁰ and also elicit dose-dependent vasoactive effects^{11, 12}. Furthermore, within the central nervous system, 5-HT_2ARs are expressed not only in neurons but also astrocytes and other cells involved in NVC^{13, 14}. These factors raise key questions when interpreting altered brain activity from psychedelics as measured by fMRI.

To this end, we first reanalyzed published human fMRI data^{15, 16} and observed that hallucinogenic, 5-HT_2AR agonism via psilocybin modified task-evoked hemodynamic responses, prompting the question of whether these effects were neuronal, vascular, or neurovascular in origin. Therefore, we employed wide-field optical imaging (WFOI) to capture both neuronal and hemodynamic activity in awake mice¹⁷ acutely exposed to the hallucinogenic, 5-HT_2AR agonist, 2,5-Dimethoxy-4-iodoamphetamine (DOI). Our findings suggest that hallucinogenic, 5-HT_2AR agonism substantially alters NVC in a regional- and frequency-specific manner, thereby affecting hemodynamic estimates of underlying neuronal activity. Notably, many effects were dose-dependent and reversed by the selective 5-HT_2AR antagonist, MDL100907 (MDL). These observations suggest caution when interpreting the effects of classic psychedelics on blood-based measures of brain function.

2). Results

2.1). Psilocybin changes evoked hemodynamic activity during human fMRI

Task-evoked responses¹⁵ were evaluated in subjects who received psilocybin, methylphenidate (to control for subjective effects), or no compound while performing a simple auditory-visual matching task (Extended Data Fig. 1). A double gamma function was used to model hemodynamic response functions (HRFs) in left/right visual cortex and auditory cortex, and left hand and language regions, and characterized by three-parameters: peak-value, dispersion, and time-to-peak. Psilocybin decreased time-to-peak in all regions except the right visual region. Psilocybin also resulted in decreased dispersion within the left/right visual and auditory regions and decreased peak amplitude in left/right visual regions.

These findings suggest a potentially altered relation between evoked neuronal and hemodynamic activity under acute psychedelic exposure, highlighting the potential for psychedelics to disrupt NVC.

2.2. WFOI during hallucinogenic 5-HT_2AR agonism

WFOI was performed in awake mice expressing the red-shifted genetically encoded calcium indicator, jRGECO1a, under a Thy1 promotor. This approach enabled simultaneous assessment of cortical excitatory neuronal activity and hemodynamics following hallucinogenic doses of DOI during task-based and resting-state conditions (Fig. 1). The selective 5-HT_2AR antagonist, MDL, was co-administered with DOI to investigate the role of the 5-HT_2AR on observed changes. Sub-hallucinogenic doses of DOI and the non-hallucinogenic 5-HT_2AR agonist, Lisuride, were used to assess the effects specific to hallucinogenic 5-HT_2AR activation. Hallucinogenic versus non-hallucinogenic doses were confirmed by head twitch responses (HTR; see methods; Fig. S1). For all WFOI, raw jRGECO1a fluorescence was corrected for absorption artifacts due to hemoglobin (Fig. S2). Potential changes in movement during WFOI were evaluated via motion tracking of facial/body movements and pupillometry. Movement during WFOI was not altered by any drug compound (Extended Data Fig. 2) consistent with prior reports¹⁸.

Figure 1. Measuring the effect of 5HT_2A receptor agonism on cortical neuronal and hemodynamic activity.

a) Schematic of WFOI system. WFOI facilitates concurrent imaging of calcium and hemodynamic activity over the cortex. A custom light engine consisting of 532nm and 625nm LEDs illuminated the skull. Fluorescence emission and diffuse reflectance were collected by a lens, split by a 580nm dichroic, and sampled by two sCMOS cameras. A 500nm long pass filter in front of CMOS1 passed 530nm reflectance while the dichroic and a 590 nm long pass filter in front of CMOS2 passed jRGECO1a emission and 625nm reflectance. A second 700nm long pass filter in front of CMOS2 blocked near infrared light used for movement and pupil monitoring. b) Imaging paradigm. Awake whisker stimulation was interwoven with awake resting-state imaging ($n=8$). Ten seconds of air puffs (40PSI, 100ms pulses delivered at 3Hz) were followed by 200s of recovery. The last 180s were considered “resting-state” epochs. This paradigm was repeated 9 times for each imaging session. Stimulus-evoked and resting-state segments were analyzed for both calcium and hemodynamic signals. Neurovascular coupling was assessed under stimulus and resting-state conditions. c) Experimental timeline. Compound randomization was performed prior to imaging, and the imager was blinded to compound identity. Prior to imaging, mice were acclimated to WFOI for 7days (45min/day). Mice were imaged for ~30min to establish pre-injection metrics, injected with a compound, and after a period of 30min, imaged again for ~30min. One week later, the procedure was repeated with a different compound.

Changes in calcium fluorescence, primarily reflecting action potential–dependent activation of voltage-gated calcium channels¹⁹, were used as a proxy for neuronal activity. However, activation of Gq-coupled receptors like 5-HT_2AR can also trigger intracellular calcium release²⁰, potentially leading to decoupling between calcium signaling and neuronal spiking activity. To evaluate the impact of these effects on WFOI measures, we conducted ex vivo electrophysiology experiments on brain slices from Thy1-jRGECO1a mice with bath-applied DOI. These experiments revealed a slow (~200 second rise time) increase in calcium fluorescence likely mediated by Gq-coupled intracellular calcium release triggered by DOI. However, temporal filtering in the WFOI analysis effectively excluded this action potential-independent signal from our in vivo data (Extended Data Fig. 3). Furthermore, DOI did not alter electrically-evoked calcium activity, indicating that bulk changes in jRGECO1a fluorescence reliably report spike-related calcium flux rather than intracellular calcium signaling following 5-HT_2AR agonism (Extended Data Fig. 4). Therefore, all WFOI findings are assumed to reflect ensemble-level, action-potential-dependent neuronal activity.

2.3). DOI dissociates evoked neuronal and hemodynamic activity

2.3.1). Response topographies and dynamics

Cortical responses to whisker stimulation were assessed pixel-wise using Cohen’s d effect size (Fig 2a). Before compound injection, stimulation of the right whiskers elicited stereotypical neuronal and hemodynamic responses within left primary somatosensory barrel (S1b) cortex and motor barrel cortex (Figs. 2a, b). DOI differentially modified calcium versus hemodynamic response topographies (Fig 2a, S3). For example, maps of evoked calcium activity decreased in the retrosplenial and somatomotor regions, while hemodynamic activity decreased in the motor cortex but increased in the somatosensory and auditory regions (black arrows, Fig. 2a). Within S1b (black contour, Fig. 2a), DOI reduced the calcium response effect but did not alter the hemodynamic response effect or area.

Figure 2. Hallucinogenic 5-HT_2A receptor agonism differentially alters stimulus-evoked calcium and hemodynamic activity.

a) Response effect maps. Cortical responses to whisker stimulation are reported prior to saline injection and after compound injection (Saline, DOI (4mg/kg), MDL (0.1mg/kg), DOI+MDL, $n=8$). Effect maps were calculated as Cohen’s D effect size between pre- and intra-stimulus periods. The black contour represents the region-of-interest (ROI) extracted from the average of all pre-injection maps (Fig. S3). DOI decreased calcium response effect sizes within the ROI (−25.7% [−32.4%, −6.5%]; pre- vs. post-injection: p=0.032). In contrast, hemodynamic response effects were not differentially altered by any compound (Kruskal-Wallis: p=0.643). DOI differentially affected calcium vs. hemodynamic response topographies (black arrows); DOI resulted in reduced calcium responses in retrosplenial and posterior cingulate regions but decreased hemodynamic responses in somatomotor regions and increased in contralateral auditory cortex. Consequently, DOI decreased the spatial correlation between calcium and hemodynamic response topographies ($\Delta r$ = −0.45 [−0.63, −0.28]; Kruskal-Wallis: p=0.009; saline vs. DOI post-hoc, two-sided Wilcoxon’s sign-rank test, p=0.009). b) Evoked response dynamics. Time courses within the ROI were averaged and visualized as medians with shaded error representing the 25^th and 75^th percentiles. Both calcium and hemodynamic response dynamics were normalized by their average intra-stimulus magnitude before injection to facilitate visual comparison within and across compounds. Therefore, a post injection value of 100% indicates a 2x increase in the intra-stimulus mean compared to the pre-injection mean. Evoked temporal responses were characterized over the following time windows after stimulus onset: peak response (0.2–2.2s for calcium and 2.0–4.0s for hemoglobin), transient response (0–5s for both contrasts), and sustained response (5–10s for both contrasts). To account for changes in pre- and intra-stimulus variability, Cohen’s D effect size was also computed. DOI increased the peak (26.0% [24.2%, 36.2%]; pre- vs. post-injection: p=0.031) and sustained calcium response (40.0% [3.4%, 120.7%]; pre- vs. post-injection: p=0.031), but not the transient response and effect size, suggesting an increase in peri-stimulus variability. In contrast, hemodynamic sustained responses decreased after both DOI (−70.3% [−132.3%, −39.5%]; pre- vs. post-injection: p=0.031) and DOI+MDL (−93.6 [−138.5%, 56.6%]; pre- vs. post-injection: p=0.031). Significance was assessed using Kruskal-Wallis test and evaluated post-hoc via two-sided Wilcoxon’s sign-rank tests corrected for multiple comparisons using Bonferroni correction. *p<0.05.

Time traces within S1b (black contour, Fig. 2a) were averaged to investigate whether DOI alters the temporal dynamics of evoked activity (Fig. 2b). DOI increased both the calcium peak and sustained responses but did not alter the temporal effect size, suggesting increased pre- and/or intra-stimulus variability. In contrast, the sustained hemodynamic response decreased following both DOI and DOI+MDL, with a significant reduction in effect size observed after DOI. Following DOI, hemodynamic responses exhibited a larger (i.e., more negative) post-stimulus undershoot resulting from decreased oxygenation and increased deoxygenation (Fig. S3b–c; Supplemental Section D1).

2.3.2). DOI impacts stimulus-evoked NVC

The differential effects of DOI on stimulus-evoked neuronal and hemodynamic activity suggests that DOI alters NVC. NVC was modelled as a causal, linear system between evoked calcium and hemodynamic activity and estimated using weighted least squares deconvolution (Fig. 3a; model efficacy evaluated in Fig. S4a, b; Supplemental Section D2). DOI altered stimulus-evoked NVC, supporting the hypothesis that the observed changes in human HRFs under acute psychedelic exposure are likely driven by altered NVC. These effects were typified by narrower HRFs (decreased FWHM), increased transduction of frequencies above 0.5 Hz (Fig. 3, Fig. S5a) and a larger post-stimulus undershoot compared to pre-injection conditions (Fig. S5b, c). Notably, many DOI-induced changes in evoked activity and NVC were reversed when DOI was co-administered with MDL (Figs. 2–3 and Figs. S3a and S5a).

Figure 3. Hallucinogenic 5-HT_2A receptor agonism alters stimulus-evoked neurovascular coupling.

a) Stimulus-evoked hemodynamic response functions (HRFs). Neurovascular coupling was modeled as a linear, causal system between calcium (input) and hemodynamic activity (output), with hemodynamic response functions estimated using weighted least squares deconvolution ($n=8$ mice). A 3 second delay was introduced between calcium and hemodynamics to assess acausal relations. HRFs were parametrized by their peak amplitude (peak, green), time-to-peak (TTP, cyan), and full-width-at-half-maximum (FWHM, purple) for each mouse, and compound (Fig. S5). DOI (4mg/kg) significantly decreased the FWHM of the HRF (−31.4% [−36.0%, −27.8%]; pre- vs. post-injection: p=0.031; Kruskal-Wallis: p=0.010; saline vs. DOI p=0.028; DOI vs. MDL (0.1mg/kg): p=0.006; purple arrow). *p<0.01. The y-axis has units of $\mu$Mol/$\Delta$F/F(%). Significance was assessed using Kruskal-Wallis test and evaluated post-hoc via two-sided Wilcoxon’s sign-rank tests corrected for multiple comparisons using Bonferroni correction. b) Stimulus-evoked transfer functions. Transfer functions were calculated as the power spectral density estimate of the HRF. A trend of increased transduction in frequencies above 0.5 Hz was observed for DOI conditions only (delta band increase +367% [57%, 472%]; pre- vs. post-injection: p=0.063). Data are presented as medians and 25^th and 75^th percentiles across mice.

2.4). DOI alters resting-state neuronal and hemodynamic activity

Psychedelics have been shown to markedly alter both regional neuronal synchrony^21–23 and brain-wide hemodynamic activity^{5, 23}. However, the relation between large-scale changes in spontaneous neuronal and hemodynamic activity induced by psychedelics remains unclear. Prior to compound injection, power spectral density estimates (PSDE) of global neuronal and hemodynamic activity exhibited a 1/frequency-like profile (Fig. 4a, b), in accordance with previous work²⁴. Following saline injection, neuronal activity exhibited broad band reductions (0.16–5.12 Hz; Extended Data Fig. 5a). While DOI decreased some low-frequency calcium activity (0.16–0.64 Hz), it also increased activity in higher frequencies overlapping with the delta band (0.32–5.12 Hz). Notably, hemodynamics under DOI also exhibited increased power in the lower range of the delta band (0.32–0.64 Hz; Extended Data Fig. 5b). Many of these changes were mitigated when DOI was co-administered with MDL (Fig. 4 and Extended Data Fig. 5).

Figure 4. Hallucinogenic 5-HT_2A receptor agonism differentially alters regional resting-state neuronal and hemodynamic activity.

Power spectral density estimates (PSDE) were computed for each cortical region and averaged to create a Global PSDE. a) Global PSDEs of calcium activity. Data are presented as mean +/− std across mice ($n=8$). Plots are color-coded by frequency range: infraslow activity (ISA), 0.01–0.08Hz, intermediate (inter): 0.08–0.5Hz, and delta 0.5–4.0Hz (quantification in Extended Data Fig. 5a–b). After DOI injection (4mg/kg), power decreased between 0.08Hz and 0.64Hz (0.08–0.32Hz: pre- vs. post- injection p=0.009; 0.16–0.64Hz: pre- vs. post-injection: p=0.020), but increased above 0.32Hz (0.32–1.28Hz: p=0.0006; 0.64–2.56Hz: p=0.0018; 1.28–5.15Hz: p=0.0017; saline vs. DOI post-hoc, 0.32–1.28Hz: p=0.0005; 0.64–2.56Hz: p=0.0007; 1.28–5.15Hz: p=0.0013). DOI co-administered with MDL (0.1mg/kg) largely preserved pre-injection spectral characteristics (increases observed between 0.02–0.08Hz, pre- vs. post- injection: p=0.018; 0.04–0.16Hz, pre- vs. post- injection: p=0.008). b) Global PSDEs of hemodynamic activity. Following saline injection hemodynamic power increased in lower frequencies (0.02–0.08Hz: vs. post- injection: p=0.045; 0.04Hz-0.16Hz: vs. post- injection: p=0.044). DOI increased over lower delta band frequencies (0.32–1.28Hz: vs. post- injection: p=0.002; ANOVA: p<0.0001; post-hoc DOI vs saline: p<0.0001; 0.64–2.56Hz: vs. post- injection: p=0.006; ANOVA: p<0.0001; post-hoc DOI vs saline: p<0.0001). Significance in a-b was determined via ANOVA and evaluated post-hoc via two-tailed t-test corrected for multiple comparisons using Bonferroni correction. c) Regional PSDE changes. Data are displayed as fractional changes from pre-injection. DOI increased calcium power (Left) in the anterior-to-posterior and lateral-to-medial directions: DOI increased ISA power in cingulate, frontal, and secondary motor networks (164% [156%,186%]) but only slightly increased ISA power in somatosensory and visual networks (26% [17%, 57%] and 13% [10%, 17%], respectively). This observation was attenuated under DOI+MDL (e.g., cingulate and frontal networks (124% [94%, 157%])). Intermediate activity exhibited similar spatial changes as ISA but markedly smaller in magnitude. For instance, DOI and DOI+MDL minimally increased cingulate (68% and 69%), frontal (33% and 42%), and secondary motor networks (14% and 38%). DOI increased delta band power in medial regions such as cingulate (197%) and secondary motor regions (153%). DOI+MDL largely reversed these changes (maximum of 49%). Hemodynamic activity (Right) reveals differential alterations compared to calcium. Under DOI, ISA appreciably increased over somatosensory cortex (212% [127%, 237%]), an effect attenuated in the DOI+MDL condition (162% [154%, 178%]). Under both DOI and DOI+MDL, changes in intermediate activity reflected those over ISA but approximately 3x smaller. Notably, increased delta band activity was observed under DOI in hemodynamic activity over somatosensory and portions of motor and visual regions, with a 314% increase in primary somatosensory barrel cortex. Full-band calcium and hemodynamic PSDEs for each region and compound are displayed in Extended Data Fig. 5.

Given the observed changes in global activity, we next examined whether these effects were region-specific (Figs. 4c top row, Extended Data Fig. 5c). PSDEs for each cortical region, as defined by the Allen Mouse Brain Common Coordinate Framework, were integrated over three frequency ranges (Infraslow activity, ISA: 0.01–0.08 Hz, Intermediate: 0.08–0.50 Hz, and delta: 0.5–4.0 Hz, Fig. 4c top row). Saline and MDL had minimal effects on calcium band-limited power and slightly larger effects on hemodynamics.

DOI led to discordant changes in regional neuronal and hemodynamic activity. For example, DOI increased infraslow calcium fluctuations in the cingulate, frontal, and secondary motor networks, whereas it increased hemodynamic activity in primary and secondary somatosensory regions (Fig. 4c, top). Across the intermediate band, DOI-induced changes in calcium power often opposed hemodynamic changes in several regions. For example, calcium activity in the visual, parietal, and somatosensory regions decreased, while hemodynamic activity in these areas remained unchanged or even increased. Additionally, calcium activity in the cingulate cortex increased, while hemodynamic activity decreased (Fig. 4c, middle). Finally, the DOI-induced changes in delta band calcium activity were largest in the cingulate, secondary motor, and other medial regions. In contrast, increased delta band hemodynamics were primarily observed in the somatosensory cortex (Fig. 4c, bottom). Co-administration of DOI and MDL largely reversed the regional effects observed with DOI alone (Fig. 4c).

2.5). DOI dramatically alters resting-state neurovascular coupling

Following our observations of compound- and region-specific changes in neuronal and hemodynamic activity, we investigated whether 5-HT_2AR agonism differentially affects global NVC (Figs. 5a–b and Extended Data Fig. 6a) or specifically alters regional NVC (Figs. 5c and Extended Data Fig. 6b).

Figure 5. Hallucinogenic 5HT_2A receptor agonism alters neurovascular coupling at local and global scales.

a) Resting-state global hemodynamic response functions. Pre-injection resting-state hemodynamic response functions (HRFs) resembled those observed during whisker stimulation ($n=8$). After DOI injection (4mg/kg), HRFs deviated from the canonical form, exhibiting a peak before time zero (black arrow), indicating a retrograde (vascular-to-neural) relation. Co-administration of MDL (0.1mg/kg) with DOI reversed these effects. HRFs were parameterized by the peak value, time to peak (TTP), and full-width-at-half maximum (FWHM). DOI increased the TTP (ANOVA: p=0.020; saline vs. DOI: p=0.019), reduced model fit (saline: from 0.80±0.04 to 0.76±0.05; DOI from 0.76±0.05 to 0.70±0.05; ANOVA: p=0.040; saline vs. DOI: p=0.048), and decreased the FWHM (purple arrow, −42.7%±71.1%, pre- vs. post-injection: p=0.015; ANOVA: p=0.020; saline vs. DOI: p=0.018). The y-axis has units of $\Delta \mu$Mol/$\Delta$F/F(%). b) Resting-state global transfer functions. Transfer functions (TFs) were computed as the power of the HRF. DOI induced a quasi-oscillatory feature in lower delta frequencies (321.9%±275.2%: p=0.018; ANOVA: p=0.004; saline vs. DOI: p=0.004; DOI vs. DOI+MDL: p=0.028). TFs have units of ($\Delta \mu$Mol/$\Delta$F/F(%))²/Hz) and displayed as dB. For panels a and b, data are presented as mean +/− std across mice and significance was determined via ANOVA and evaluated post-hoc via two-tailed t-test corrected for multiple comparisons using Bonferroni correction. c) Regional neurovascular coupling. DOI induced topographical changes in NVC (Peak, TTP, FHWM, Delta Transduction) of opposite polarity in higher order brain regions (e.g., frontal, and cingulate) vs. lateral somatosensory and parietal regions. Coupling amplitude decreased (lower peak values) in frontal/cingulate regions (−62%), while more modest increases occurred in somatosensory regions (e.g., barrel cortex: 32%). DOI delayed NVC (increased TTP) in frontal (51%) and anterior-lateral parts of somatomotor regions (32%). Strikingly, DOI substantially decreased FWHM of the HRF over most of the cortex with the largest changes occurring in cingulate (−142%) and frontal (−183%) regions. DOI substantially increased delta band transduction over most of the cortex (cortical-wide: 237% [79%, 350%], 793% in primary somatosensory barrel cortex), except in cingulate regions (reduced by 56%). DOI administered with MDL predominantly reversed these effects.

Prior to compound administration, global HRFs under resting-state conditions closely resembled stimulus-evoked HRFs (Fig. 5a) and reflect those reported previously^24–26. However, following DOI, the HRF changed significantly, showing a prominent peak before time zero and reduced model fit (i.e., correlation between measured and predicted hemodynamics; Fig. S4b), suggesting a deviation from the model’s assumptions of causality and linearity. This emergent acausal peak manifested as increased delta band transduction in the transfer function (TF; Fig. 5b). Moreover, HRF features after time zero (neuronal-to-hemodynamic relation) were also significantly affected by DOI, including increased time-to-peak and decreased full-width-at-half-maximum (Extended Data Fig. 6a).

To determine where these NVC changes were occurring, regional HRFs and TFs were parameterized and mapped onto the cortex (Fig. 5C, Extended Data Fig. 6b). Consistent with previous reports of regional variations in NVC²⁷, pre-injection HRFs varied across the cortex (Figs. 4c and Extended Data Fig. 6b, left). Critically, after DOI injection, a global deviation from the canonical HRF—characterized by a dominant quasi-acausal peak—was present ubiquitously across the cortex (Extended Data Fig. 6b). As a result, HRF widths decreased over most of the cortex, with the largest decrease occurring within the cingulate, frontal, and motor regions (Fig. 5c, FWHM). Delta band transduction increased within visual, parietal and somatosensory cortex and modestly decreased in cingulate (Fig. 5c). DOI often induced opposing changes in NVC between higher-order versus lower-order brain regions (Fig. 5c). For example, DOI increased HRF delays (larger TTP) in secondary somatosensory and primary motor areas, while shortening delays in the cingulate cortex (Fig. 5c, TTP). Further, HRF peaks decreased in motor and higher-order regions (frontal, cingulate, retrosplenial) but increased in somatosensory and parietal regions (Fig. 5c, Peak). Importantly, the acausal feature (i.e., a peak preceding time zero) was observed not only in HRFs but also in lagged cross-covariance and phase relations analyses (Extended Data Fig. 7 and Fig. S6, Supplemental Section N1).

We next investigated the degree to which the observed changes in resting-state NVC depended on hallucinogenic versus non-hallucinogenic 5-HT_2AR agonism. Sub-hallucinogenic doses of DOI (Extended Data Fig. 8a–c) did not impact global (Extended Data Fig. 8b) or local (Extended Data Fig. 8c) NVC. Separately, we investigated the non-hallucinogenic, 5-HT2_AR agonist, Lisuride, selected for its ability to target signaling pathways similar to classical psychedelics that may influence vascular tone, without inducing hallucinations at a dose known to alter locomotion²⁸ (Extended Data Fig. 8a). Unlike hallucinogenic doses of DOI, Lisuride did not alter the shape of the HRF (Extended Data Fig. 8e) or NVC topographies (Extended Data Fig. 8e). Instead, Lisuride affected cortical calcium and hemodynamic activity similarly to saline, producing broad-band reductions in ISA/intermediate calcium activity and more modest effects on hemodynamic signaling (Extended Data Fig. 8f).

2.6). Discordant RSFC reported by calcium vs. hemodynamic activity

Human fMRI studies have reported dramatic network-level alterations in functional brain organization following acute psychedelic exposure^{5, 6, 23, 29}. These alterations in RSFC, particularly in regions linked with higher-order cognitive processes^{5, 30–32}, are often attributed to neuronal origins (e.g., ‘neural correlates’^{30, 31, 33, 34}), and serve as biomarkers of the subjective effects of psychedelics³³. However, considering our observations of marked and region-dependent alterations in NVC following psychedelic exposure, it remains unclear whether changes in hemodynamic-based RSFC reflect reorganization of neuronal network connectivity.

To align with human neuroimaging analyses, RSFC based on infraslow activity (ISA) was evaluated in parts of the mouse default-mode network (i.e., retrosplenial and cingulate cortex, homologous structures to the posterior and anterior cingulate cortex in humans³⁵) as well as in frontal and secondary motor cortex (M2, Fig. 6). When evaluating RSFC of cingulate cortex (Fig. 6b, Table 1) calcium-based RSFC reported enhanced intra-network RSFC, increased inter-network RSFC between cingulate and frontal cortex, and reduced connectivity between cingulate and parts of retrosplenial cortex (Fig. 6b, Table 1). No changes in cingulate RSFC were reported when assessed with hemodynamics (Fig. 6b, Table 1). DOI-induced changes in calcium-based RSFC within frontal cortex (Fig. 6c, Table 1) largely mirrored those observed within cingulate cortex; however, hemodynamic-based RSFC measures reported larger anticorrelations between frontal cortex and parts of retrosplenial and parietal cortex (Fig. 6c, Table 1). While more modest changes intra-and inter-network RSFC strength were observed in M2 cortex after DOI, disparities between calcium- and hemodynamic-based connectivity assessments remained (Fig. 6d, Table 1).

Figure 6. Calcium and hemodynamic activity report differing accounts of RSFC changes arising from hallucinogenic 5-HT_2AR agonism.

RSFC was evaluated in regions exhibiting the largest NVC changes (Frontal, Cingulate and Retrosplenial) and secondary Motor (M2) over infraslow (ISA, 0.02–0.08Hz) and delta band (0.5–4.0Hz) activity ($n=8$). a) Retrosplenial RSFC: DOI (4mg/kg) decreased ISA calcium RSFC within retrosplenial and between retrosplenial and lateral visual (left hemisphere: $\Delta r=-0.22\pm 0.14$, right hemisphere: $\Delta r=-0.31\pm 0.19$). DOI enhanced ISA HbT RSFC between retrosplenial and frontal/cingulate ($\Delta r=0.29\pm 0.12$). MDL (0.1mg/kg) did not fully reverse the effects of DOI (ISA calcium RSFC between retrosplenial and frontal/cingulate $\Delta r=0.25\pm 0.20$; between retrosplenial and medial somatosensory $\Delta r=0.25\pm 0.16$, and between retrosplenial and lateral visual cortex $\Delta r=-0.28\pm 0.20$. ISA HbT RSFC between retrosplenial and frontal/cingulate $\Delta r=0.26\pm 0.16$). DOI increased delta calcium anticorrelations between retrosplenial and M2/posterior somatosensory ($\Delta r=-0.11\pm 0.02$). b) Cingulate RSFC: DOI enhanced ISA calcium anticorrelations between cingulate and posterior-lateral retrosplenial ($\Delta r=-0.41\pm 0.17$) and strengthened ISA calcium connectivity between cingulate/frontal and anterior-medial retrosplenial regions ($\Delta r=0.40\pm 0.24$). These DOI-induced ISA calcium RSFC changes were not reflected in ISA HbT RSFC. DOI increased anticorrelated delta calcium activity between cingulate and retrosplenial ($\Delta r=-0.14\pm 0.07$) and increased delta calcium connectivity between cingulate and frontal/medial somatosensory ($\Delta r=0.15\pm 0.04$). MDL reversed all DOI effects. c) Frontal RSFC: Saline modestly decreased ISA calcium connectivity between frontal and somatosensory (left, $\Delta r=-0.13\pm 0.02$; right. $\Delta r=-0.12\pm 0.02$). DOI enhanced inter-network ISA calcium RSFC between frontal and cingulate ($\Delta r=0.31\pm 0.15$) and ISA HbT connectivity between frontal and posterior retrosplenial ($\Delta r=0.18\pm 0.10$). DOI increased delta calcium RSFC between frontal and secondary motor/posterior somatosensory ($\Delta r=0.18\pm 0.03$) and increased anticorrelations between frontal and retrosplenial ($\Delta r=-0.11\pm 0.05$). MDL attenuated DOI effects, excluding ISA calcium RSFC between frontal and retrosplenial ($\Delta r=0.20\pm 0.13$) and ISA HbT RSFC between frontal and somatosensory ($\Delta r=-0.16\pm 0.06$). d) M2 RSFC: DOI increased negative ISA calcium RSFC between M2 and lateral somatosensory ($\Delta r=-0.18\pm 0.12$), and increased connectivity between M2 and anterior cingulate/posterior retrosplenial ($\Delta r=-0.15\pm 0.07$). This effect was not observed in ISA HbT RSFC. Delta calcium RSFC under DOI was increased within sensorimotor regions (left hemisphere: $\Delta r=0.09\pm 0.03$, right hemisphere: $\Delta r=0.09\pm 0.05$) with increased anticorrelations between M2 and retrosplenial/visual ($\Delta r=0.09\pm 0.04$). DOI+MDL reversed many DOI effects, excluding decreases in ISA calcium and ISA HbT connectivity between M2 and somatosensory cortex. MDL alone slightly decreased connectivity between cingulate and right somatosensory regions ($\Delta r=-0.09\pm 0.13$). Range of ISA RSFC maps and delta RSFC maps are |r|<0.5 and |r|<0.3, respectively. RSFC is further evaluated in Extended Fig. 9. Statistical differences (black contours) in seed-based RSFC were evaluated on a cluster-wise basis.

Table 1).

Spatial similarity between RSFC changes estimated by calcium and hemodynamic signaling.

	Frontal	Cingulate	Retrosplenial	M2
Saline	y=0.72x+0.01; r=0.61	y=0.06x+0.03; r=0.05	y=0.63x+0.01; r=0.49	y=0.78x−0.02; r=0.58
DOI	y=0.18x+0.00; r=0.29	y=0.04x+0.05; r=0.05	y=0.04x+0.02; r=0.05	y=0.21x−0.02; r=0.25
DOI+MDL	y=0.24x+0.00; r=0.37	y=0.37x+0.02; r=0.45	y=0.54x+0.02; r=0.56	y=0.50x+0.00; r=0.41
MDL	y=0.34x−0.01; r=0.43	y=0.17x+0.01; r=0.18	Y=0.42x+0.02; r=0.42	y=0.61x−0.01; r=0.59

Differences (i.e., post- minus pre-injection) in ISA (i.e., 0.01–0.08Hz) RSFC topography were computed for each mouse ($n=8$), compound, and region (Fig. 6). Spatial similarity (Pearson $r$ spatial correlation) and linear regression (slope and intercept) evaluated the linear, topographical correspondence between calcium- and hemodynamic-based RSFC. DOI significantly dissociated (i.e., decreased correlation) neuronal vs. hemodynamic reports of RSFC in retrosplenial (Kruskal-Wallis, p=0.028) and secondary motor (Kruskal-Wallis, p=0.038) regions. There was also a trend of dissociation in the frontal region. Interestingly, MDL alone decreased RSFC correspondence more than DOI+MDL in the cingulate and retrosplenial region, but increased correspondence more than MDL in frontal and secondary motor regions. Significance was determined using Kruskal-Wallis tests and evaluated post-hoc using two-sided Wilcoxon sign-rank test and corrected for multiple comparisons using the Bonferroni method.

Another advantage of calcium-based measures of RSFC is the ability to capture faster neuronal dynamics, which exhibit diverse features compared to infraslow hemodynamics²⁴. Prior to injection, delta band (0.5–4.0 Hz) calcium activity reveals distinct functional topography compared to ISA RSFC, in line with prior work²⁴ (Fig. 6a–d, right). As observed with ISA RSFC, DOI modified delta band RSFC, with a widespread decrease in retrosplenial connectivity with the rest of the cortex (Fig. 6a–d). Moreover, cingulate connectivity increased with frontal and medial motor/anterior lateral somatosensory regions and parietal regions (Fig. 6b). Interestingly, the observed changes in delta RSFC for frontal and retrosplenial regions were topographically similar, but of opposite sign, excluding decreased RSFC with retrosplenial regions (Fig. 6c compared to Fig. 6b). Lastly, M2 increased connectivity with most of the somatomotor regions and decreased in retrosplenial and visual regions (Fig. 6d).

These pronounced differential alterations in inter- versus intra-network connectivity were further examined through community detection analysis (Extended Data Fig. 9; Supplemental Section N2). DOI significantly decreased ISA calcium silhouette scores, indicating weaker intra-network connections and stronger inter-network connections across most of the cortex. The most substantial decrease was observed in the retrosplenial region and along the posterior boundary of the somatosensory cortex. These differences were not observed in hemodynamic scores. Scores derived from delta band calcium activity were generally lower compared to those based on ISA, with the largest DOI-induced increases occurring in the retrosplenial cortex. Notably, the most significant changes in delta band silhouette scores were found at parcel boundaries, suggesting a shift in network borders. These differences are largely reversed when DOI is co-administered with MDL.

Taken together, these results suggest that, while baseline RSFC assessed through calcium and hemodynamic signals are often topographically similar, administration of DOI leads to significant disparities between calcium- and hemodynamic-based measures of RSFC.

3.). Discussion

3.1). Summary of present findings

A reanalysis of previously published human fMRI data demonstrated evidence that hallucinogenic, 5-HT_2AR agonism via psilocybin can modify task-evoked hemodynamic responses. To assess whether this phenomenon could potentially be attributed to a neuronal, vascular, and/or neurovascular effect, we employed WFOI in awake mice undergoing acute exposure to the hallucinogenic, 5-HT_2A receptor agonist DOI. DOI differentially altered cortical excitatory neuronal versus hemodynamic activity during task (whisker stimulation) and in the resting-state, prompting direct assessment of NVC. During task, DOI narrowed stimulus-evoked HRFs and enhanced transduction of higher frequency (>0.5 Hz) neuronal activity into subsequent hemodynamics. During rest, DOI dramatically altered the HRF shape (i.e., emergent acausal peak) in a region dependent manner, a phenomenon especially prominent in higher-order brain regions (e.g., retrosplenial cortex). Further, calcium and hemodynamic activity reported differing accounts of RSFC changes under DOI. Co-administration of DOI with the 5-HT_2AR antagonist, MDL, reversed many of the effects observed under DOI alone. This work builds on a body of work demonstrating that psychedelics exhibit neurovascular effects across species^{16, 36, 37}.

3.2). Local and global biomarkers of psychedelics

Classic psychedelics are 5-HT_2AR agonists with hallucinogenic properties¹⁰. This receptor subtype is predominantly located in cortical regions associated with “higher” cognitive functions^{13, 38} (e.g., default mode network, DMN)^{13, 35, 39}, regions thought to be key mediators of the psychedelic experience^{29, 40}. Within these key brain regions in animal models, electrophysiological studies have demonstrated regionally specific modulations of neuronal excitability following psychedelic administration^{21, 23, 41, 42}. For instance, administration of DOI increased pyramidal neuronal firing rate and decreased low-frequency power in local-field potentials within rat prefrontal cortex⁴³. Although these results are informative, they are susceptible to regional expression of receptor subtypes and may not capture emergent brain-wide phenomenology.

Recent resting-state fMRI studies in humans have indirectly examined the brain-wide neuronal effects of psychedelic agents on network organization^{5, 6, 30–32}. Notwithstanding some preprocessing- and experimentally- dependent inconsistencies⁴⁴, there is a broad consensus that psychedelics reduce RSFC within association networks, particularly within the DMN^{16, 29, 40}, and increase functional connectivity between association networks and other regions of the brain^{5, 23, 30}. Notably, the DMN – a constellation of brain regions consistently impacted by psychedelics– exhibits both decreased intra-network connectivity and increased inter-network connectivity^{5, 29, 40, 45}. More globally, psychedelics also result in dissolution and fragmentation of RSFC patterns, which have been quantitated using entropic measures^{6, 16, 32}. This reorganization is also reported as reduced modularity (i.e., decreased intra-network connectivity with enhanced inter-network connectivity)^{5, 46}. Together, these findings are often interpreted as psychedelics modulating neural activity and communication. However, interpreting purely hemodynamic measures as direct reflections of neural processes is inadequate without considering the neurovascular effects of psychedelics.

3.3). DOI Disrupts Regional NVC and RSFC

Our observations of DOI-induced increases in delta band calcium activity are concordant with prior EEG studies in humans examining the effects of other psychedelics on neuronal excitability^{21–23, 41, 42}. Importantly, increased delta band activity following DOI was also observed in hemodynamic signals. Enhanced delta band transduction and decreased HRF width was particularly pronounced in lateral portions of the mouse cortex, which could explain the observed increase in delta band hemodynamic activity in this region. This topographical pattern, which reflects regional differences in NVC, may correspond to key observations in human fMRI reporting the loss of functional segregation between primary sensory cortical areas (e.g., somatosensory and visual) and the default mode network^{29, 40, 45}.

Our findings suggest a nuanced effect of psychedelics on RSFC measured by neuronal versus hemodynamic activity. Moreover, these differences obscure subtle changes in RSFC structure induced by psychedelics. For example, neuronal based measures demonstrate increased intra-network RSFC within cingulate cortex and decreased inter-network RSFC between cingulate and retrosplenial regions. These effects were not observed in hemodynamic reports of RSFC. Additionally, we do not observe brain-wide (or DMN-wide) decreases in neuronal network “modularity”. Instead, DOI enhanced silhouette scores—akin to increased modularity—in the anterior regions of the retrosplenial cortex and in the posterior parts of the cingulate cortex (both components of the mouse DMN), with more widespread decreases observed in the posterior and lateral retrosplenial cortex. These changes were also not reflected in hemodynamic measures. The severing of connectivity between these regions is especially relevant given common observations of psychedelics on inter- and intra-network connectivity in humans^{45, 47, 48}.

Local changes in NVC can significantly impact the estimated RSFC structure inferred from hemodynamics. For instance, coherent neuronal activity exhibited within cingulate and M2 would suggest a degree of “functional connectivity” between these regions. Alterations in NVC in the cingulate cortex (e.g., our observed decreases in TTP and FWHM) will alter the phase and spectral content of hemodynamic signaling in this region, leading to apparent incoherence between cingulate and M2 when assessed through hemodynamics. This phenomenon extends to all regions “functionally connected” to cingulate cortex, demonstrating how local NVC perturbations can drive global changes in hemodynamic estimates of RSFC. Distinguishing neuronal RSFC changes from those driven by non-neuronal factors proves challenging when multiple regions exhibit nonuniform, region-dependent NVC alterations.

3.4). Neurovascular effects of serotonin and its receptors

Throughout the brain, the distribution of 5-HT_2ARs shapes neural, vascular, and neurovascular responses. These receptors are predominantly expressed in pyramidal cells and, to a lesser extent, in specific subsets of interneurons^{39, 49} and astrocytes¹³ – all cell types that participate in NVC^50–52. Activation of 5-HT_2ARs in these diverse cell populations can produce vasoactive effects that lead to complex, region-specific changes in hemodynamics. Moreover, hemodynamic responses can vary depending on the cell populations activated, sometimes resulting in opposing or biphasic effects^53–57. Separately, the widespread distribution of 5-HT_2ARs –and their differential activations via psychedelics – facilitates modulation of excitatory-inhibitory balances across the cortex^38,58–60 influencing both local and global brain perfusion^{7, 58–62}

The broad action of psychedelics across multiple brain regions and receptors^{43, 63–65} indicates that these ligands are not exclusively selective for 5-HT_2ARs but also bind to other 5-HTRs and Class A G protein-coupled receptors. This promiscuity complicates interpretations of how psychedelics influence the vasculature. For instance, activation of 5-HT_2ARs induces vasoconstriction via the PLC pathway involved in Gq-coupling, whereas activation of 5-HT_2B, 5-HT₇, and 5-HT₁ subtypes, typically promotes vasodilation⁶⁶. To that end, selective pharmacological antagonists are valuable tools for dissecting receptors-specific contributions to neurovascular dynamics. Comparing the shared and unique receptor profiles and physiological effects of compounds such as Lisuride, MDL+DOI, and DOI may help identify which receptors underlie the distinct neural and vascular responses observed. Additionally, contributions from non-serotonergic receptors should also be considered^67–69.

Together, these factors underscore the complexity of serotonergic modulation of neurovascular function. A detailed understanding of how psychedelics engage receptor- and cell type-specific mechanisms will be critical for interpreting their effects on brain-wide hemodynamics.

3.5). Hallucinogenic vs. non-hallucinogenic psychedelic effects

We did not observe altered NVC from sub-hallucinogenic doses of DOI or Lisuride. These observations might be explained by partial agonism of the 5-HT_2AR. DOI is a full 5-HT_2AR agonist that strongly activates the Gq-coupled cascade⁷⁰. This agonism leads to the phosphorylation of signaling proteins (e.g., PLC, PKC, ERK/MAPK, and CaMKII) and increased production of secondary messengers (e.g., inositol phosphate signaling molecules (IP3) and diacylglycerols) which influence synaptic activity and neuronal excitability. Importantly, increased production of IP3 in astrocytes during neuronal activity triggers a calcium signal that activates pathways leading nitric oxide (NO) release and vasodilation⁷¹. In contrast, Lisuride is a partial 5-HT_2AR agonist with weaker influence on these pathways^{70, 72}. Therefore, Lisuride and sub-hallucinogenic doses of DOI likely cause less phosphorylation of signaling proteins and reduced production of secondary messengers that might influence blood flow⁷². The more robust activation of Gq-mediated signaling pathways by full agonist, hallucinogenic psychedelics might explain their greater impact on vascular and neural activity.

3.6). A potential retrograde calcium-to-hemodynamic relation

DOI induced an apparent zero-lag notch in the HRFs, resulting in a peak at negative lags. These observations imply a relation in which hemodynamic activity precedes calcium activity. Importantly, the prominence of this effect varied regionally (e.g., largest in secondary motor and retrosplenial regions). Furthermore, this feature was present in calculations of lagged cross-covariance between unfiltered calcium and hemodynamic signals and appears in the frequency domain as a broad peak around ~0.8 Hz in the same regions. Before compound injection, the calcium-to-hemodynamic phase relation exhibited a positive slope over frequencies between zero and ~1 Hz, indicating that calcium strictly leads hemodynamic activity over these frequencies. However, DOI induced negative phase slopes (indicating hemodynamics lead calcium) in specific cortical regions (e.g., frontal, and cingulate) within the same frequency band. These observations could arise from DOI-induced alterations in the calcium-hemodynamic relation (a neurobiological account) or misinterpretation of the optical data (artifact). However, the fac that this notch was unevenly distributed across the cortex, appeared only under hallucinogenic doses of DOI, and was absent in stimulus-evoked HRFs subjected to the same analysis argues against it being an artifact.

An alternative account of this feature includes two DOI-dependent neurobiological phenomena: neuro-vascular coupling, the process by which neural activity regulates blood flow, and vasculo-neural coupling (VNC), where changes in blood flow influence neural activity^73–76. Many mechanisms that induce functional hyperemia often impact neuronal activity⁷⁴. For example, freely diffusing NO causes relaxation of smooth muscle⁷⁷ and modulates neuronal activity⁷⁸. Blood volume changes during functional hyperemia, can also lead to morphological changes in surrounding neurons, activating mechano-sensitive ion channels (e.g., amiloride-sensitive Na+ channels and stretch-activated cation channels, both of which are abundant in the mouse neocortex⁷⁹). Additionally, brain temperature, which is directly governed by blood flow⁸⁰, can affect neuronal activity. Furthermore, vaso-astrocytic interactions can indirectly impact neuronal activity through diffusible messengers (e.g., astrocytic-endothelial cell contact⁸¹ modulating NO release triggered by increased hemodynamic activity⁸²) and mechanical interactions (e.g., stretch-activated ion channels in astrocytes⁸³). Considered collectively, there is ample evidence that psychedelics might give rise to VNC both directly and indirectly.

3.7). Limitations

It is important to note the changes in NVC we observe are specifically between Thy1-expressing cortical neurons (primarily expressed in pyramidal cells of cortical layers 2/3 and 5) and the vasculature. Several studies have also demonstrated strong spatiotemporal coupling between Thy1-based ensemble activity and subsequent hemodynamic activity under both task and resting-state conditions^{26, 84, 85}. However, activity in subpopulations of excitatory and inhibitory neurons can be sufficient to alter cerebral blood flow^{7, 57, 61, 62, 86–88}. Future research could also incorporate electrophysiological measurements to determine the local contributions of excitatory and inhibitory neurons (i.e., comparing excitatory activity to total activity). Alternatively, a more global characterization of neurovascular relationships could be achieved by using mice expressing genetically encoded calcium or voltage indicators in other neurons or glia.

Serotonin can modulate systemic physiology¹⁴ (e.g., respiratory rate, cardiac rate, blood pressure), which can influence neurovascular relations^{27, 76, 89–94} and directly affect hemodynamic RSFC estimates^{27, 76, 89–97}. Furthermore, psychedelics have been shown to affect systemic physiology in a dose-dependent manner^{31, 98}. Importantly, methylphenidate, the non-psychedelic control condition used in the reanalyzed human fMRI study¹⁶, also produces significant physiological effects without altering stimulus-evoked HRFs. These crucial control data suggest that the potent effects of psychedelics on NVC may not be solely due to systemic physiological changes.

Although our NVC model accounts for many features of the observed data, it assumes causality, linearity, and time-invariance. Collectively, these assumptions may oversimplify the complexities of neurovascular relations. The most salient example of this oversimplification is the HRF measured after DOI, which contains a primary (positive lag) and secondary (negative lag) peak, challenging the assumption of causality. Therefore, more comprehensive models of coupling between neuronal and hemodynamic activity could account for intermediate steps^{24, 99, 100} or more complex retrograde phenomena (e.g., VNC)^73–76. Expanding NVC models to account for acausal, recursive, or nonlinear effects^{97, 101} could further characterize aspects of NVC alterations.

Several observations under DOI conditions were not fully reversed by the 5-HT_2AR antagonist, MDL, which raises questions of dose and of receptor-dependent psychedelic effects outside of 5-HT_2AR binding^{43, 63–65}. MDL is one of the most selective 5-HT_2AR antagonists currently available. Although it retains some affinity for 5-HT_2CRs, sigma, ${\alpha }_{1\text{A}}$, and ${\alpha }_{2\text{B}}$ receptors, there is limited in vivo evidence suggesting it acts strongly at these receptors at the dose used in this study¹⁰². Similarly, Lisuride has a different affinity profile for these, and other 5-HT receptors compared to DOI but did not alter NVC or induce HTRs in mice. Investigating a broader range of compounds and receptor subtypes would provide a more comprehensive understanding of the neurophysiology underlying psychedelic effects.

Emerging evidence also suggests sexual dichotomy in the pharmacokinetic and behavioral profile of multiple psychedelics¹⁰³. Although this study included both sexes, it was not adequately powered to evaluate sex-specific differences in outcome. Future studies will be critical to evaluate the degree to which different psychedelics affect sex-specific alterations in neurophysiology.

Lastly, saline often altered neurophysiology (excluding NVC), predominantly in a frequency-dependent manner. This phenomenon was also observed in the independent dataset examining Lisuride. These changes may be attributed to the mouse experiencing increased stress after injection due to handling and needle insertion. Such effects can manifest as decreased delta band power^{104, 105}. To that end, assessment of other compound conditions are interpreted with respect to these and other inherent “stress-inducing” effects.

3.8). Conclusions and next steps

Our findings highlight the need for more direct measures of neuronal activity and NVC to address questions about acute psychedelic intoxication. In human studies, this could be achieved with simultaneous EEG-fMRI. Our WFOI sampling rate resulted in sensitivity to hemodynamic activity in frequencies extending up to ~5 Hz. The largest hemodynamic effect of DOI occurred at frequencies above 0.2 Hz. The extent to which this phenomenon is relevant to human fMRI studies is not clear as BOLD-fMRI signals are markedly attenuated at frequencies above ~0.2 Hz in awake humans and are immeasurable (or aliased) with most fMRI acquisition speeds. To address this limitation, optical techniques such as diffuse optical tomography could be used to measure cortical hemodynamics with higher temporal resolution¹⁰⁶. More broadly, psychedelic research would benefit from a more cautious stance towards the conventional neurocentric interpretation of BOLD-fMRI. This shift encourages a broader, more holistic perspective concerning the genesis of blood-based signals^{50–52, 76, 89–91}. Such an approach could improve our understanding of the effects of psychedelics on brain function.

Methods

1). Hemodynamic response functions in task-based human fMRI data

We assessed changes in the hemodynamic response functions (HRFs) of previously published psychedelic human fMRI task data (Extended Data Fig. 1)¹⁶. Briefly, subjects were imaged under no drug, methylphenidate (40mg), and psilocybin (25mg) while performing a simple auditory-visual matching task. Generalized linear models (GLMs) were computed at pre-specified regions of interest (ROIs) which were selected from the Gordon-Laumann parcellation¹⁰⁷ (left/right calcarine sulcus (V1), left/right auditory cortex (A1), left language area (Wernicke-like area), left hand knob, left angular gyrus, and right angular gyrus (default mode). HRFs were fit using a double Gamma function, which can be described using thee parameters (peak value, dispersion, and time-to-peak)¹⁰⁸. MANOVAs were used to assess the differential effect of compound (or no drug) and participant on each parameter. A post-hoc t-test was conducted on regions where the MANOVA indicated a significant effect of compound.

2). Mice and animal preparation

All animal studies were approved by the Washington University School of Medicine Animal Studies Committee under guidelines and regulations consistent with the Guide for the Care and Use of Laboratory Animals, Public Health Service Policy on Humane Care and Use of Laboratory Animals, and the Animal Welfare Act and Animal Welfare Regulations. Animal reporting is according to ARRIVE guidelines.

A total of 40 mice were used for this study aged between 3.5 – 7 months of age. Twenty three mice expressing jRGECO1a driven by the thymus cell antigen 1 (Thy1) promotor (JAX Strain: Tg(Thy1-jRGECO1a) GP8.20Dkim/J; stock: 030525)^109–111 were used to visualize wide-field cortical calcium fluctuations, primarily from excitatory neurons of layers 2/3 and 5, in relation to hemodynamic activity^{111, 112} during hallucinogenic and non-hallucinogenic 5-HT_2AR agonism. Eight mice (4 male, 4 female) were imaged under acute exposure to hallucinogenic doses of DOI (4 mg/kg), MDL (0.10 mg/kg), DOI+MDL, and saline. Six mice (3 male, 3 female) were imaged under acute exposure to either saline or non-hallucinogenic doses of DOI (0.04mg/kg) and 9 mice (4 male, 5 female) were imaged under acute exposure to either saline or the non-hallucinogenic, 5-HT_2AR agonist Lisuride. Three mice expressing the non-activity-dependent fluorophore YFP, Thy1-YFP (JAX Strain: B6.Cg-Tg(Thy1-YFP)16Jrs/J; stock: 003709) were used to assess hemodynamic contamination of fluorescent signals. Four wild-type mice (JAX Strain: C57BL/6J; stock: 000664) were used to assess head-twitch responses. One of these mice experienced a fatality; however, collected data was included in the analysis. Finally, two wild-type mice and eight Thy1-jRGECO1a mice were used for slice experiments. Mice were group housed, given food and water ad libitum, and maintained on a 12:12 light/dark cycle at approximately 21 Celsius and 40% humidity.

Prior to imaging, a cranial window was secured to the intact skull of each mouse using dental cement, following scalp retraction under 2% isoflurane anesthesia¹¹³ (Fig. 1b). One hour before surgery, mice were given buprenorphine-SR (1.0 mg/kg subcutaneous) and recovered from surgery for three days prior to any handling or experimentation. Once recovered, and prior to any imaging experiments, all mice were behaviorally acclimated to the awake imaging apparatus for 45 minutes per day for 7 days to reduces stress^{17, 114}.

To determine the appropriate number of subjects and experimental design, an initial pilot study was conducted in Thy1-jRGECO1a (4 females, 5 males) using saline and Lisuride^{115, 116} under awake, resting-state conditions. Mice were imaged for 30 minutes before compound injection and for 90 minutes immediately following the injection. Because we observed the largest significant effects of Lisuride on cortical activity (see Extended Data Fig. 8f) 30 to 60 minutes post-injection, all subsequent WFOI occurs before and 30 to 60 minutes after injection of any compound. These results are in line with prior work demonstrating the largest effects of DOI occur 30 minutes after injection¹¹⁷.

Doses were determined from prior literature and through evaluation of behavior, namely the head twitch response (below)^118–122. From these pilot studies the following dosages were used for experimentation: DOI (4 mg/kg), MDL (0.10 mg/kg), DOI+MDL, and saline. To assess the effect of sub-hallucinogenic doses and non-hallucinogenic 5-HT_2AR agonists on NVC, low-dose DOI (0.04 mg/kg, as determined HTR) and Lisuride (1.0 mg/kg) were also used. To control for volumetric effects, all mice were injected with the same volume of liquid per mass (20 mL/kg). All compounds were injected into the intraperitoneal (i.p.) space and prepared on the day of experimentation.

3). Head twitch response and dose response

The head-twitch response (HTR) is a widely used behavioral assay in which rodents exhibit rapid side-to-side head movements after administration of serotonergic hallucinogens and other 5-HT_2AR agonists. HTRs strongly correlate with the duration of action and strength of these compounds^118–122. Extensive research has demonstrated that DOI induces powerful psychedelic effects and induces HTRs^{119, 121, 123, 124}. A custom magnetometer was built for automating HTR characterization based on open-source designs¹¹⁸. Briefly, a Plexiglas chamber (4” H × 4” W × 4” L) was constructed and wrapped with 18 AWG wire to cover ~90% of the box height. The solenoid output was connected to an amplifier (Phono preamp; PP444, Pyle) and recorded via a data acquisition card (USB-6001, National Instruments). Mice were ear tagged with a small magnetic disk (3mm diameter, 0.5mm thickness, ~58 mg) glued to the tag, so that when the mouse was placed in the chamber, any head twitches would induce an electromotive force in the coil. Raw HTR data were denoised using a discrete wavelet transform (wavedec, MATLAB) with a symmetric wavelet of order 6 (‘sym6’ MATLAB). Soft thresholding was performed using Stein’s unbiased risk estimation for each level’s detail coefficients before reconstructing the signal. Once denoised, data were band-passed filter from 70–110 Hz using a 20^th order infinite impulse response filter and rectified. Peaks associated with HTRs were identified (‘findpeaks’, MATLAB) as those having a minimum peak height of 200 times the median absolute deviation and separated by more than 3 seconds from any other identified peak¹¹⁸.

Pilot studies in 4 mice and 4 conditions (saline, DOI: 0.04 mg/kg, DOI: 0.4 mg/kg, and DOI: 4mg/kg) were used to validate our magnetometer and automated detection algorithm by comparing against results from a manual HTR scorer blinded to condition (${\text{R}}^2$=0.950 between automated and manual scoring of HTRs; Fig. S1a). These data were also used to assess the dose response of DOI (Fig. S1b). A second cohort of mice (2M, 2F) was used to assess HTRs following randomized injection of each compound (Saline, DOI, MDL, DOI+MDL; same dose and concentration as WFOI experiments; Fig. S1c). Lastly, a third cohort of mice (2M, 2F) was used to assess HTRs following Saline, Lisuride (1.0 mg/kg), and a sub-hallucinogenic (low dose) of DOI (0.04 mg/kg, as determined by compound response curve) and underwent the same protocol as cohort 2. For all cohorts, HTR recordings began 30 minutes after injection of compound. For cohort 1 (HTR dose response), recordings were 10 minutes long, and for cohorts 2 (Saline, DOI, MDL, DOI+MDL) and 3 (low dose DOI and the non-hallucinogenic, 5-HT_2AR agonist, Lisuride), recordings were 30 minutes long (Extended Data Fig 8a). For all cohorts, tests of compound effects on individual mice were separated by at least 3 days to avoid potential changes in tolerance and drug administration was randomized and experimenters were blinded.

4). Ex vivo electrophysiological and imaging recording

The slice preparation procedures were modified from our previous study¹²⁵. Adult Thy1-jRGECO1a (n = 4) and C57BL/6J (n = 2) mice were anaesthetized via i.p. injection of a cocktail containing ketamine, xylazine and acepromazine (69.57 mg/ml; 4.35mg/ml; 0.87mg/ml; i.p. 182mg/kg, respectively). Mice were then perfused with cold slicing-aCSF consisting of 184 mM sucrose, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 10 mM MgSO₄, 20 mM HEPES, 30 mM NaHCO₃, 25 mM glucose, 0.5 mM CaCl₂, 5 mM sodium ascorbate and 3 mM sodium pyruvate, oxygenated with 95% O₂ and 5% CO₂. The pH was 7.3–7.4 and osmolality adjusted to 315–320 mOsm with sucrose. The forebrain was dissected and embedded with 2% agarose in slice-aCSF and coronal slices were cut with 300 μm thickness using a vibratome (VF310-0Z, Precisionary Instruments, MA, USA). Slices were incubated at elevated temperature (32°C) for 30 minutes and then transferred to room temperature in holding-aCSF consisting of 92 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO4, 30 mM NaHCO₃, 20 mM HEPES, 25 mM glucose, 2 mM MgSO₄, 2 mM CaCl₂, 5 mM sodium ascorbate and 3 mM sodium pyruvate, oxygenated with 95% O2 and 5% CO2. pH was 7.3–7.4 and osmolality adjusted to 310–315 mOsm. This slice preparation procedures were modified from previous study¹²⁵. Both electrophysiological and imaging recordings took place in a recording chamber mounted on upright microscope (BX51WI, Olympus Optical Co., Ltd, Tokyo, Japan) as previously described¹²⁶ with continuous perfusion of warm (29–31°C) recording-aCSF consisting of 124 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 24 mM NaHCO₃, 5 mM HEPES, 12.5 mM glucose, 2 mM MgCl₂, 2 mM CaCl₂, oxygenated with 95% O2 and 5% CO2 and pH 7.3–7.4 with osmolality adjusted to 305–310 mOsm using sucrose. Cell-attached recordings were performed through borosilicate glass pipette (GC150F-10, Warner Instruments, Hamden, CT, USA) with a resistance around 6–9 MΩ when filled with recording-aCSF. Electrical stimulation (20 Hz with 200 μs) was generated by a stimulus isolator (A385, World Precision Instruments, FL, USA) and delivered through a bipolar tungsten electrode (FHC, Inc., Bowdoin, ME, USA). Stimulation intensity (60–90 & 100–125 μA for Layer 2/3 & local stimulation, respectively) was adjusted based on the input-output relationship. Electrophysiological signals were gathered using Multiclamp 700B amplifier (Molecular Devices, San Jose, CA, USA) and digitized at 10 kHz via Axon Digidata 1550B interface (Molecular Devices, CA, USA) coupled with Clampex software (Molecular Devices, CA, USA). A subset of cell-attached recordings were performed in C57BL/6J mice without calcium imaging due to mouse availability. Calcium imaging in Thy1-jRGECO1a recordings were collected by epifluorescence equipment mounted on microscope coupled with a highspeed camera (ORCA-Flash4.0LT, Hamamatsu Photonics, Shizuoka, Japan). To account for photobleaching, raw traces at each pixel were fit to a DC-shifted exponential, $\text{a}{\text{e}}^{-\text{bx}}+\text{c}$, using a least-squares approach, and the exponential was regressed. $\Delta \text{F}/{\text{F}}_0$ was computed such that ${\text{F}}_0$ was the average of the epoch before drug application. Signals were then filtered using a second order Butterworth filter (0.01Hz – 5 Hz), as performed during WFOI analysis. Global calcium fluctuations were computed as the average signal across the entire field-of-view.

5). Imaging Hardware

The experimental setup included four separate cameras (Fig. 1a), 2 for cortical neuroimaging (WFOI) and 2 for movement monitoring (1 for monitoring murine motion, and 1 for monitoring pupil dynamics).

5.1). Wide-field Optical Imaging (WFOI)

Awake cortical neuroimaging was performed using a modified version of a previously-published WFOI system²⁴. A custom light engine consisting of 470 nm (measured peak λ = 457 nm, LCS-0470-50-22, Mightex Systems), 530 nm (measured peak λ = 525 nm, LCS-0525-60-22, Mightex Systems), and 625 nm (measured peak λ = 637 nm, M625L3, Thorlabs) light-emitting diodes (LED) illuminated the skull (Fig. 1a). Diffuse reflected light for optical intrinsic signal imaging and fluorescence emission were collected by a lens (focal length = 75 mm, NMV-75M1, Navitar), split by a 580 nm dichroic (FF580-FDi02-t3-25×36, Semrock) and sampled by two scientific complementary metal–oxide–semiconductor (CMOS) cameras with USB3 connectivity (Zyla 5.5, Andor). A 500 nm long pass filter (FF01-500/LP-25, Semrock) in front of Camera 1 (CMOS1) passed 530 nm reflectance. The 580 nm dichroic and a 593 nm long pass filter (FF01-593/LP-25, Semrock) in front of Camera 2 (CMOS2) blocked 530 nm jRGECO1a excitation light and passed jRGECO1a emission and 625 nm reflectance. A separate 700 nm short pass filter (Thorlabs, FESH0700) placed behind 593 nm long pass filter and in front of CMOS2 blocked near-infrared illumination used for movement monitoring (see below). Each WFOI camera sensor was cropped to 1024×1024 pixels, and 2×2 binning was performed to increase the acquisition rate and improve the signal-to-noise ratio (SNR) of each image. The frame rate of each camera was 60 Hz, with all contrasts imaged at 20 Hz. The LED and camera exposures were synchronized and triggered via a data acquisition card (PCI-6733, National Instruments) using MATLAB R2022 (MathWorks). LED spectra were measured using an Ocean Optics USB 2000+ spectrometer.

5.2). Motion and Pupil Tracking

Mouse motion and fluctuations in pupil diameter were each tracked by CMOS cameras (Thorlabs, DCC3260C) placed in front of the mouse (see below, Extended Data Fig. 2). The motion camera was placed approximately 6 cm away from the mouse’s snout and captured spontaneous facial and forepaw movements. The pupil camera was places approximately 3 cm away from the mouse head. The fields-of-view of each camera were as follows: Motion FOV: ~70 mm-by-70 mm, 480 pixels-by-480 pixels, 0.146 mm/pixel; Pupil FOV, ~8 mm-by- 8mm, 480 pixels-by-480 pixels, 0.017 mm/pixel). Both FOVs were illuminated by a 780 nm LED (Thorlabs, M780L3). The frame rate of each camera was 20 Hz and was time-locked to the 625 nm LED used for WFOI. Specular reflection in Motion/Pupil image sequences was minimized using crossed near-infrared, linear polarizers (Thorlabs, LPVISC100).

6). Data Acquisition

During imaging, mice were supported by a felt pouch and mounted to the WFOI system using an aluminum bracket attached to the cranial window^{17, 24, 57, 114, 127}. The awake neuroimaging paradigm consisted of interleaved epochs of “resting-state” (or “task-free”) and “stimulus-evoked” imaging, as performed with human fMRI¹²⁸ (Fig. 1b,). “Resting-state” is experimentally defined as the subject in the recording apparatus, isolated from extraneous stimuli and not subjected to any experimentally-imposed stimuli (e.g., overt sensory inputs)¹²⁹. While the animals are head fixed and secured in the felt pouch (which limits any head motion and large movements of the limbs) they are free to look around, shift their body position, whisk, groom, etc. Resting-state epochs were 180 seconds long. Stimulus-evoked responses were elicited by delivering puffs of air (40 PSI, 100 ms pulses at a frequency of 3 Hz) to the right whiskers for 10 seconds, followed by a 20-second recovery period. This resulted in a total stimulus block length of 30 seconds (Fig. 1b). Before and after injection, resting-state and stimulus-evoked epochs were each repeated nine times, resulting in 27 minutes of resting-state data and 4.5 minutes of task data, for a total of 31.5 minutes. Our initial pilot studies determined the largest effect in WFOI measures (e.g., changes in cortical power) occurred between 30 min and 60 min after injection of Lisuride. Therefore, after compound randomization, mice were imaged for another 31.5 minutes beginning 30 minutes after injection.

7). Motion and Pupil Tracking

Prior to any analysis, each pixel’s time trace was normalized by its temporal mean (via division) to account for illumination inhomogeneities. Motion image sequences were spatially decimated by a factor of two to expedite computation time. Optical flow (OF) was performed on all motion monitoring movies, as previously reported¹³⁰ (Extended Data Fig. 2a,i–iii). OF estimates were generated using the Lucas-Kanade method^{131, 132}. This algorithm produces a spatiotemporal series of vectors where, for each frame, a pixel’s vector represents the change in intensity magnitude and direction over the next few frames (n = 5; Extended Data Fig. 2a,iv). The root-sum square (RSS) of each frame’s vector magnitude field was computed to yield a single time-trace estimating movement (large values signify large changes in movement; Extended Data Fig. 2a,v).

Additionally, pupil dynamics are associated with arousal dynamics, which give rise to spontaneous behaviors such as whisking, grooming, and fidgeting^{97, 133–135}. Pupils were therefore monitored and tracked (via, an in-house, deep-learning segmentation algorithm) and compared across compounds (Extended Data Fig. 2d–g). Specifically, changes in pupil area were synchronized with the motion monitoring and WFOI cameras to determine whether compounds altered pupil dynamics during neuroimaging sessions. To extract pupil area from each image sequence, transfer learning from the pre-trained Segment Anything Model (SAM)¹³⁶ was employed utilizing the Vision Transformer model¹³⁷ (i.e., ‘facebook/sam-vit-huge’) in the SamProcessor framework. This model, initially trained on a large dataset, underwent fine-tuning to specialize in accurately delineating pupils for each frame. To optimize the model’s performance for the specific task of pupil segmentation and ensure effective knowledge transfer, all encoder layers were frozen. This strategic freezing of layers enabled the model to retrain its learned representations of general features while allowing the trainable parameters of the decoder to adapt specifically to our application (Extended Data Fig. 2c).

Adam optimization with an initial learning rate of 1E-3 and a scheduler that reduces the learning rate by 25% every 50 epochs was employed for training. Monai’s DiceFocalLoss (gamma value of 2.0) was utilized instead of the more conventional Dice or Dice cross-entropy losses due to the substantial class imbalance within the imaging frames (>80% of pixels were non-pupil). To mitigate overfitting, a dropout rate of 30% and a regularization constant of 1E-4 was applied. Training spanned 100 epochs with early stopping implemented to avoid overfitting; training stopped if validation loss at epoch n was less than the median validation loss between epochs $\text{n}-15$ to $\text{n}$.

For the first round of training, all pupil videos were clustered (k-means) with 80 clusters (cluster number determined via an elbow plot). The cluster centroids representing the largest clusters (the 10 clusters with the most data points) were manually segmented. At least 15 user-defined points were used to contour the pupil, which was then fit to an ellipse using a least-squares approach (Extended Data Fig. 2d). This process resulted in the labeling of 956 images. Data augmentation techniques were applied, including affine transformations with randomly selected parameters and random horizontal flips. Consequently, the final dataset comprised 1912 pupil images along with their respective labels. Of these, 80% were allocated for training and 20% for validation.

In the subsequent training phase, 15 pupil videos were randomly selected for frame-by-frame processing using the trained model. Poorly segmented frames often resulted in a large spatial shift in the likelihood distribution/map. (Extended Data Fig. 2e). Therefore, for each frame, the center of mass (COM) of the likelihood map was calculated and, for each video, the ten largest deviations of COM were identified. Frames before, at, and after these deviations were manually segmented (15 images per video). This process yielded approximately 450 additional labeled frames, which were also subjected to augmentation. These frames were appended into the training dataset, and the training process (i.e., starting with the SAM initial weights), along with outlier detection and refinement, was repeated once more. In this way, the training dataset was built and fine-tuned to model weaknesses before the final training and model implementation.

Once the model was adequately trained, every pupil video was analyzed at its native frame rate (20 Hz). The sequence of spatiotemporal likelihood maps was then median filtered using a 5-frame kernel. For each frame, contours (imcontour and roipoly, MATLAB 2022) were fit around the likelihood cloud and all pixels within this contour were counted, resulting in a time course of pupil area changes during imaging (example of pupil trace in Extended Data Fig. 2f). The temporal derivative of each pupil area time courses was evaluated for each mouse and compound condition to determine whether changes in pupil diameter were different pre- and post-injections. Because mice often closed their eyes during whisker stimulation, pupil analysis was only performed during resting-state epochs.

8). WFOI data pre-processing

8.1). Spatial normalization

Image sequences from each neuroimaging camera were spatially coregistered using a projective transform, which was constructed from four user-selected landmarks present within both camera’s FOV. WFOI image sequences were then coregistered to the Allen Mouse Brain Atlas. Briefly, two anatomical landmarks are identified and labeled on the cortex (1) the junction between the rostral rhinal sinus and the sagittal sinus and (2) the lamboid suture^{113, 138}. All pixels labeled as brain were used to create a brain mask for each mouse. All subsequent analyses are performed within this common space.

8.2). Spectroscopy and spatiotemporal filtering

For each WFOI camera, one second of dark frames were collected, averaged, and subtracted from image sequences¹¹³. Raw WFOI image sequences (512×512 pixels) were spatially binned (4×4) off-camera to further increase the SNR of each image, resulting in 128×128 pixel images. For each channel, slow trends in light levels were temporally detrended with a 5^th order polynomial fit^{17, 113, 138, 139}. Images were spatially smoothed using a Gaussian filter (5×5-pixel kernel, standard deviation of 1.2 pixels). Changes in wavelength-dependent diffuse reflectance were converted to changes in hemoglobin concentration using the modified Beer-Lambert law^{113, 138}. Specifically, for each pixel, ${\varphi }_{\lambda }\left(\text{t}\right)={\varphi }_{0,\lambda }\text{exp}(-\Delta {\mu }_{\text{a},\lambda }\left(\text{t}\right){\text{X}}_{\lambda })$, where ${\varphi }_{\lambda }\left(\text{t}\right)$ is the measured light intensity at each wavelength, ${\varphi }_{0,\lambda }$ is the average intensity over the imaging session, ${\mu }_{\text{a},\lambda }$ is the change in absorption coefficient due to changes in blood volume, and ${\text{X}}_{\lambda }$ is the differential pathlength that accounts for the volume of tissue sampled by photons at each wavelength. Isolating $\Delta {\mu }_{\text{a}}$ yields $\Delta {{\mu }_{\text{a}}}_{\lambda }\left(\text{t}\right)=-\text{ln}\left(\frac{{\varphi }_{\lambda }\left(\text{t}\right)}{{\varphi }_0}\right){/\text{X}}_{\lambda }$, where $\lambda$ indexes the channel used (e.g., LED centered at 530 nm). The system of equations $\Delta {\mu }_{\text{a},\lambda }={\text{E}}_{\lambda ,\text{i}}\Delta \left[\text{H}{\text{b}}_{\text{i}}\right]\left(\text{t}\right)$ was solved, such that $\text{E}$ is the wavelength-dependent extinction coefficient for the ${\text{i}}^{\text{th}}$ hemoglobin species (i.e., oxy- and deoxy-hemoglobin). Hemoglobin extinction coefficients were those tabulated by Scott Prahl¹⁴⁰. Differential pathlengths were computed assuming a semi-infinite geometry, initial concentration of hemoglobin ($\text{s}{\text{O}}_2=71\%$, ${\left[\text{HbT}\right]}_{\text{initial}}=76\text{micromol}$), and estimates of the reduced scattering coefficient¹⁴¹. Calculation of extinction coefficients and differential path lengths were weighted by LED spectra. All optical properties are reported in Table S1. All hemodynamic outcome measures were reported using total hemoglobin (HbT).

To accurately assess fluorescent emission in media with changing optical properties, such as varying chromophore concentrations, it is essential to correct for absorption artifacts. Accordingly, raw jRGECO1a fluorescence was corrected using estimated changes in $\Delta {\mu }_{\text{a}}$ at excitation and emission wavelengths^{25, 26, 142}. Corrected temporal fluctuations in fluorescence at each pixel were converted to percent change: $100*\frac{\Delta \text{F}\left(\text{t}\right)}{{\text{F}}_0}$, where $\text{F}\left(\text{t}\right)$ is the corrected fluorescence, ${\text{F}}_0$ is the temporal mean of $\text{F}$, and $\Delta \text{F}\left(\text{t}\right)=\text{F}\left(\text{t}\right)-{\text{F}}_0$. The efficacy of the hemodynamic correction was determined in mice expressing the yellow fluorescent protein (YFP) under a Thy1-promotor, which provides non-activity-dependent fluorescence (Fig. S2).

9). Signal pre-processing

Resting-state, spontaneous fluctuations in calcium and hemodynamic signaling (sampled at 20 Hz) were filtered to obtain frequencies between 0.02 – 0.08 Hz to assess infraslow activity, 0.08 – 0.5 Hz to assess intermediate activity, and 0.5 −4.0 Hz to assess delta band activity. Filtered signals were down sampled to 1.25 times the lowpass filter frequency cutoff (e.g., data filtered between 0.5–4Hz were resampled to 5 Hz). For stimulus evoked activity and NVC assessment (see below), data were filtered to retain frequencies between 0.01 Hz and 5 Hz and then down sampled to a sampling rate of 12 Hz. All filtering and down sampling was performed using a 5^th order Butterworth filter (butter, MATLAB 2022) and anti-aliasing interpolation (resample, MATLAB 2022). To minimize boundary effects from dividing the data into task and rest epochs, filtering was conducted prior to splicing. After filtering and down sampling, resting-state and stimulus-evoked activity were segmented into periods of rest and task using a Tukey window (with a Tukey parameter of 0.3, which attenuated 15% of the data at each end of the window).

10). Analysis of stimulus-evoked activity

To enhance the spatial specificity of evoked cortical topography, global signal regression (GSR) was performed¹⁴³. To facilitate averaging across task presentations, 5 seconds of activity occurring prior to stimulus onset was averaged and subtracted from each image in the block. Blocks were then averaged together for each mouse and group. Response maps of evoked calcium and hemodynamic activity were created by averaging images between 0.2–5.2 and 2.0–7.0 seconds after stimulus onset, respectively.

To account for potential changes in pre- and intra-stimulus variance, effect sizes of evoked activity (Fig. 2a) were quantified using Cohen’s D:

$$\text{d}=({\mu }_{\text{intra}}-{\mu }_{\text{pre}})/{\sigma }_{\text{pooled}},$$ (1)

where ${\mu }_{\text{pre}}$ and ${\mu }_{\text{intra}}$ are the mean of the baseline (5 seconds before stimulus onset) and the response period (0.2–5.2 seconds for calcium and 2.0–7.0 seconds for total hemoglobin), respectively, and

$${\sigma }_{\text{pooled}}=\sqrt{({\sigma }_{\text{pre}}^2+{\sigma }_{\text{intra}}^2)/2},$$ (2)

where ${\sigma }_{\text{pre}}$ and ${\sigma }_{\text{intra}}$ are the temporal standard deviations in activity and n is the number of sample points.

To create a high-SNR image representing evoked activity across all mice and groups, all 32 pre-injection effect maps (8 mice 4 compound conditions) were averaged together. From this map, a region of interest (ROI) was defined as all pixels within 25% of the maximum effect (Fig. S3a). To account for variability before and during stimulus presentations, we examined maximum response effect within the ROI. Similarly, the response area was computed using the number of pixels exceeding 75% of the maximum effect. The topographical correspondence between evoked calcium and hemodynamic activity was assessed via spatial similarities (Pearson correlation) between calcium and hemodynamic effect maps.

Response temporal dynamics for each mouse were evaluated by averaging all pixel time courses within the ROI. Only evoked activity exceeding 2 standard deviations of baseline (pre-stimulus) activity was included in subsequent analysis. Response dynamics are visualized in Figure 2b. Changes in evoked activity was compared across groups through quantification of the peak response (the sum between 0.2–2 seconds and 2.2–4.2 seconds, for calcium and HbT, respectively), transient response (the sum between 0–5 seconds), sustained response (the sum between 5–10 seconds), and the post-stimulus undershoot (the sum after stimulus-offset, 15–30 second; Fig. S3b,c). Additionally, the effect size between pre- (2-seconds before stimulus-onset) and post-stimulus (0.2–2 seconds and 2.2–4.2 seconds, for calcium and HbT, respectively) was computed as Cohen’s D (Fig. S3a).

11). Stimulus-based neurovascular coupling

Neurovascular coupling (NVC) was modeled as a causal, linear, time-shift-invariant process. For each pixel, let $\text{y}=\text{Xh}$ where $\text{y}$ is measured hemodynamic activity, $\text{h}$ is the hemodynamic response function (HRF), and $\text{X}$ is the convolution operator (each column is an $\text{m}$-shifted calcium signal). For task, x and y are the average response from all pixels within the ROI seen in Fig. 2. In expanded form, this equation becomes:

$$\left(\begin{array}{c}{\text{y}}_1\\ ⋮\\ {\text{y}}_{\text{n}}\end{array}\right)=\left(\begin{array}{ccc}{\text{x}}_1& \cdots & 0\\ ⋮& \ddots & ⋮\\ {\text{x}}_{\text{n}}& \cdots & {\text{x}}_{\text{n}-\text{m}}\end{array}\right)\left(\begin{array}{c}{\text{h}}_1\\ ⋮\\ {\text{h}}_{\text{m}}\end{array}\right),$$ (3)

where n is the number of sample points and m is the length of the convolution kernel. Weighted least squares deconvolution was employed to estimate the HRF:

$$\text{h}={\left({\text{X}}^{\text{T}}\text{WX}+\stackrel{-}{\lambda }\text{I}\right)}^{-1}{\text{X}}^{\text{T}}\text{Wy},$$ (4)

where $\text{W}$ is a weight matrix, $\stackrel{-}{\lambda }$ is a regularization constant, and $\text{I}$ is the identity matrix. We chose the weight matrix to be the singular value spectrum of $\text{X}$, namely $\text{W}$ (i.e., $\text{X}=\text{UW}{\text{V}}^{\text{T}}$). To standardize the regularization constant for all mice and conditions, $\stackrel{-}{\lambda }$ was defined as $\stackrel{-}{\lambda }=\lambda ·‖{\text{X}}^{\text{T}}\text{WX}‖$, where the fractional regularization constant, $\lambda$, was heuristically set to $\lambda ={10}^{-5}$.

Prior to deconvolution, an artificial 3 second delay was introduced between calcium and hemodynamic activity to examine temporal relations prior to time zero (stimulus-based NVC, Fig. 3; resting-state NVC, Fig. 5). To assess model fit, measured calcium activity was convolved with the HRF to produce predicted hemodynamic signals, which were then correlated with measured hemodynamics (Fig. S4). HRFs were parameterized by peak-value, time-to-peak, and full-width-at-half-maximum (Fig. 3a, left). Additionally, the oxy-, deoxy-, and total-hemoglobin post-stimulus HRF undershoot (summed activity 2-seconds after stimulus-onset) was compared across compounds (Fig. S5b,c). To assess the frequency-dependent properties of calcium-to-hemodynamic transduction, transfer functions (TF) were defined as the power spectral density estimate of the HRF. Transduction was quantified using integrated power in the ISA, intermediate, and Delta bands (defined above).

12). Quantification of resting-state activity

Resting-state activity was quantified regionally by averaging all time courses across anatomical regions as defined in the Allen Mouse Brain Atlas¹⁴⁴ (Figs. 4c, Extended Data Fig. 5c; Lisuride, Extended Data Fig. 8f). Power spectral density estimates (PSDEs) for each region and epoch (9 resting-state epochs, 180 seconds each) were computed using a smooth spectral estimator (pwelch, MATLAB 2022). These estimates were then averaged across epochs to produce mouse-level PSDEs. Spectra were integrated over the ISA, intermediate, and delta bands (defined above). Global PSDEs for each mouse were obtained by averaging the PSDEs across all cortical regions. Finally, group-level cortical resting-state PSDEs were calculated by averaging across all mice (Figs. 4a,b, Extended Data Fig. 5a,b).

13). Resting-state neurovascular coupling

Resting-state NVC (Figs. 5, Extended Data Fig. 6; low-dose DOI, Extended Data Fig. 8b–c, and Lisuride, Extended Data Fig. 8d–f) was quantified regionally as defined above. Regional time courses were segmented into 30 second, non-overlapping, windows (Tukey window, lobe parameter of 0.3). HRFs were estimated for each 30 second epoch and the median across all epochs defined a regional HRF. Regional estimated HRFs were averaged across regions to obtain a cortical-wide (global) estimate of NVC. HRFs and TFs were parametrized in the same manner as stimulus-evoked NVC.

To better understand emergent features in resting-state NVC process (e.g., peaks at times prior to zero), cross-covariance, coherence, and phase-relation was also computed on the same resting-state epochs (Extended Data Figs. 8&9, see below) and regions.

14). Cross-covariance analysis

Regional cross-covariance functions between unfiltered calcium and HbT signals were computed to validate and assess emergent features (e.g., peaks prior to time zero; Extended Data Figs. 7, Fig. S6) present in neurovascular coupling estimates obtained using weighted deconvolution (see main text). From these functions, peak and time to peak were calculated for each region. The frequency-dependent content of cross-covariance functions were estimated by as the cross-spectral density, quantified as magnitude-squared coherence (MSC):

$$\text{MSC}=\frac{{\text{P}}_{\text{xy}}^2}{{\text{P}}_{\text{xx}}{\text{P}}_{\text{yy}}},$$ (5)

where ${\text{P}}_{\text{xy}},{\text{P}}_{\text{xx}}$, and ${\text{P}}_{\text{yy}}$ are the cross-spectrum power density and $\text{x}$- and $\text{y}$- PSDEs, respectively.

Phase information between calcium and hemodynamics were investigated through the angle between the real and imaginary parts of the cross-spectrum power density. Phase averaging was performed in the complex plain to avoid issues with phase wrapping. Owing to the features observed in the MSC plots (see Results), power was integrated over two frequency bands: between 0.5 Hz and 2 Hz and below 0.5 Hz.

15). Resting-state functional connectivity

Measures of resting-state functional connectivity (RSFC) were calculated using zero-lag correlation for all pixel pairs within the brain mask^{113, 114, 138, 139, 145}. Briefly, spontaneous activity was filtered to obtain frequencies over infraslow and delta frequency bands. Global signal regression was performed to reduce global coherence, resulting in increased spatial specificity of RSFC patterns^{143, 146, 147}. Whole cortex correlation matrices were reordered according to the mouse brain atlas. Regional RSFC maps for frontal, cingulate, retrosplenial, and secondary motor (M2) were computed from the whole cortex by averaging all seed-based RSFC maps in each region (Fig. 6). Psychedelic-induced changes in RSFC were computed as RSFC post-injection minus RSFC pre-injection. Spatial similarity (Pearson correlation) and linear relationships (linear regression) between calcium- and hemodynamic reports of RSFC differences were computed (Table 1) as:

$$\text{y}={\beta }_1\text{x}+{\beta }_0,$$ (6)

where $\text{x}$ and $\text{y}$ are vectorized RSFC difference matrices for calcium and HbT, respectively.

16). Inter- vs. Intra-network functional connectivity

Differential changes in inter- and intra-network connectivity is a common observation in studies examining the effects of psychedelics on systems-level brain function^{5, 6, 40, 47, 48}. Modularity, which measures the strength of connections within versus between network communities, is often used to quantify these changes. In the context of psychedelics, it is often interpreted as changes in network integration⁴⁶. However, while modularity is a commonly used measure to quantify these changes by assessing the strength of connections within versus between network communities, it is limited in topographical assessment. Furthermore, modularity obscures finer details of altered network organization, as it primarily focuses on community-level changes. In contrast, silhouette scores offer a more nuanced evaluation of network integration by measuring the similarity of individual nodes within clusters relative to their distance from other clusters.

To examine differences in RSFC structure before and after compound injection, RSFC matrices were computed for all pairwise comparisons within our field of view for each mouse before injection and averaged to yield a high-fidelity RSFC matrix (Extended Data Fig. 9a). Resting-state networks (RSNs) within this matrix were identified using Ward’s hierarchical clustering method^{148, 149}, which respects the intrinsic hierarchy within the data and has been previously applied to RSFC measures in humans^{148, 150, 151}. Each branch of the resulting dendrograms (Extended Data Fig. 9b) represents a potential cluster. Moving from the bottom of the dendrogram to the top of the tree (corresponding to increased variance when merging), the threshold for defining a cluster becomes less stringent. Merging branches indicate that clusters have combined at less stringent criterion thresholds. A branch that remains separate over a longer vertical distance represents more stable clusters across thresholds. A clustering threshold (black-dotted line, Extended Data Fig. 9b) was selected that aligns most closely with functional regions defined by the Allen Mouse Brain Atlas¹⁵² and previous data-driven clustering results¹²⁷. The resulting 13 clusters (Extended Data Fig. 9c) were used in subsequent analysis of RSFC changes through evaluation of silhouette scores.

Briefly, silhouette scores at each pixel quantify how similar a pixel is to its own cluster compared to others (Extended Data Fig. 9d,i), and like modularity¹⁵³, can be interpreted as a network measure related to cluster integration versus segregation. The algorithm computes a ratio comparing the similarity of a pixel’s RSFC map to those within its cluster (cohesion) to its dissimilarity from maps in the nearest neighboring cluster (separation). Silhouette scores range from −1 to +1, such that +1 indicates a point is well-clustered (i.e., integrated within its community and segregated from neighbors; strongly intra-connected, weakly inter-connected), 0 indicates that the point is on or near the boundary between two clusters (equal intra- and inter-connectivity), and −1 indicates a point is misclassified and likely should be assigned to a different cluster than its own (weakly intra-connected, strongly inter-connected). Cohesion and separation were computed relative to each functional cluster’s centroid, computed as the average RSFC map within a cluster (vectorized as ${\text{C}}_{\text{n}}$). Considering a pixel in the $\text{i}$-th pixel location (and vectorized RSFC Map, ${\text{r}}_{\text{i}}$) and the $\text{k}$-th cluster, cohesion and separation are defined as:

$${\text{a}}_{\text{i}}=\frac{1}{\left|{\text{C}}_{\text{n}}\right|}{\sum }_{\text{j}\in {\text{C}}_{\text{n}}\text{j}\ne \text{i}}{\text{r}}_{\text{i}}^{\text{T}}{\text{r}}_{\text{j}}$$ (7)

$${\text{b}}_{\text{i}}={\text{min}}_{{\text{C}}_{\text{n}}\ne {\text{C}}_{\text{m}}}\frac{1}{\left|{\text{C}}_{\text{m}}\right|}{\sum }_{\text{j}\in {\text{C}}_{\text{m}}}{\text{r}}_{\text{i}}^{\text{T}}{\text{r}}_{\text{j}}$$ (8)

Then, silhouette scores (${\text{s}}_{\text{i}}$) were computed at each pixel:

$${\text{s}}_{\text{i}}={(\text{b}}_{\text{i}}-{\text{a}}_{\text{i}})/\text{max}\left({\text{a}}_{\text{i}},{\text{b}}_{\text{i}}\right).$$ (9)

17). Statistics

The use of Gaussian statistics was determined using a Lilliefors test. Significant differences in movement and band-limited spectral content were determined using a Kruskal-Wallis one-way analysis of variance of the mean. For stimulus-evoked responses and associated HRFs and TFs, results are reported as medians and percentiles owing to the low number of stimulus presentations. Parameters extracted for evoked and resting-state activity and NVC were calculated as percent changes from pre-injection. Except where stated otherwise (owing to low n), significant effects of compounds were determined using a Kruskal-Wallis one-way analysis of variance. Post-hoc significance was assessed using the paired Wilcoxon sign-rank test. Significant effects of time (i.e., pre- versus post- injection comparisons) was assessed using the paired Wilcoxon signed-rank test. Multiple comparisons were corrected using the Bonferroni correction. When mean and standard deviation are used, effects of compounds were determined through one-way analysis of variance (ANOVA) and post-hoc t-tests. The effect of time (i.e., pre- versus post- injection) was determined to be significant using paired t-tests and corrected for multiple corrections using Bonferroni correction. All tests were two-tailed. For all RSFC measures, Pearson correlation coefficients were converted to Fisher z values prior to any mathematical operation and converted back to Pearson correlations for visualization. Statistical differences in seed-based RSFC were evaluated on a cluster-wise basis^{154, 155}, with a clustering threshold of 12 contiguous pixels.

Extended Data

Extended Data Figure 1. Psychedelics alter hemodynamic response functions (HRFs) in humans, as reported by BOLD-fMRI.

a) Previously published functional magnetic resonance imaging (fMRI) data were analyzed from humans undergoing a simple auditory-visual matching task during acute exposure to psilocybin, methylphenidate, or nothing. Generalized linear models were computed for task-evoked activity in left/right visual cortex, left hand, left/right auditory cortex, and left language. HRFs were fit using double Gamma basis functions and characterized by three-parameters: Peak Value (P), Dispersion (D), and Time to Peak (T). Data are presented as mean +/− SEM. b) Quantified changes in human hemodynamic response functions. Statistical differences for each parameter across conditions were assessed via 2-way ANOVA; significant effects were evaluated post-hoc via two-sided t-tests. p-values are indicated such that #<0.05, ##<0.01, ###<0.001. Data are presented as quantiles and all outliers (>99.3^rd percentile) were excluded for visualization purposes only. Top panel: Peak value decreased in left and right visual regions under psilocybin conditions (ANOVA, p=0.030 and p=0.020, respectively). Middle panel: Time-to-peak decreased in all regions except the right visual region (ANOVA, left visual, p=0.03 left hand, p=0.006; left auditory, p=0.002; right auditory, p=0.044; left language, p=0.007). Bottom panel: Dispersion decreased in the left and right visual regions, and left and right auditory regions (ANOVA, p=0.007, p=0.004, p=0.021, p=0.022, respectively).

Extended Data Figure 2. Mouse body movements and pupil dynamics do not differ across experimental conditions.

Two cameras were placed in front of the mouse to monitor body movement and pupil dynamics. Image sequences of mouse movement and pupil dynamics were time locked to WFOI. a) Motion monitoring using optical flow. (i) Optical flow (OF) estimates were generated using the Lucas-Kanade method (ii); OF yields a spatiotemporal series of vectors, with each pixel in each frame assigned a vector that reflects local spatiotemporal gradients. At each pixel, the vector magnitude provides a scalar measure of motion, with higher values corresponding to larger movements (iii). The temporal variance of vector magnitudes is shown, with yellow colors indicating larger variance, and demonstrates OF sensitivity to the motion of the felt pouch supporting the mouse (which moves when the mouse moves). (iv) The vector magnitude was computed at each pixel; higher values correspond to larger movement. (v) The root-sum square across all frames was computed to yield a single time trace of estimated movement. b) Distributions of OF-derived movement measures. No differences were observed between conditions ($n=8$ mice per condition, as assessed by differences between post- minus pre- injection; Kruskal-Wallis: p=0.262). c) Simplified schematic of the automated pupil segmentation algorithm. d) Example of a user-annotated pupil segmentation and the automated pupil segmentation using our algorithm. As demonstrated, user-guided segmentation matches well with automated segmentation. e) Automated outlier detection. Large spatial deviations in the likelihood distributions of the pupil mask’s center-of-mass are attributed to artifact due to the slow rate of pupil diameter fluctuations compared to the imaging frame rate. Outlier frames, along with the frames immediately before and after, were manually labeled and appended to the original training dataset to increase robustness of the algorithm. f) Example frames and time course of pupil area change during imaging. g) Changes in pupil area before and after compound injection. Distributions of pupil area changes did not differ before and after injection of any compound (saline, DOI (4mg/kg), MDL (0.1mg/kg), DOI+MDL, $n=8$ mice per condition).

Extended Data Figure 3. 5-HT_2AR-mediated action potential-independent increase in global calcium signal.

Ex vivo recordings of global calcium signals in acute brain slices upon DOI (10μM) and TTX (1μM) application ($n=6$ mice). a) Experimental arrangement. Global imaging and cell-attached recordings were conducted on prefrontal cortical region (e.g., cingulate, prelimbic and infralimbic cortex) and pyramidal cells within these regions, respectively. b) Subthreshold effects led by direct activation of 5-HT_2AR. Representative cell-attached recordings demonstrate lack of spontaneous activity of cortical pyramidal cells in acute brain slice preparation. Application of NMDA, but not DOI or DOI+TTX, resulted in burst firing in recorded cells (aCSF only: 3 cells, NMDA: 5 cells, DOI: 6 cells, DOI upon TTX: 4 cells). c) Tonic, TTX-resistant global calcium increases upon DOI application. Representative traces corrected for photobleaching show a small but steady increase of global calcium signal led by DOI application, which was not affected by TTX pretreatment. (1.55±0.96 vs. 0.87±0.78%, two-way, t-test: p=0.314). d) Elimination of 5-HT_2AR-mediated action potential-independent tonic calcium signal by temporal filtering. Representative filtered traces from the same recordings shown in panel c (0.01–5Hz) demonstrate complete removal of action potential-independent DOI-mediated calcium increase (aCSF: 0.00±0.01 vs. 0.01±0.03%, two-sided, Wilcoxon rank-sum test: p=0.343). Abbreviations: aca, anterior commissure (anterior part); AI, anterior insular cortex; Cg1, cingulate cortex (area 1); IL, infralimbic cortex; M1, primary motor cortex; M2, secondary motor cortex; PrL, prelimbic cortex.

Extended Data Figure 4. 5-HT_2AR activation does not contribute to calcium fluctuations in electrically-evoked neural activity.

Ex vivo recordings of electrical stimulation-evoked calcium signal in acute brain slice upon DOI (10μM) application ($n=4$ mice). a-b) Experimental arrangement. Simultaneous cell-attached recordings and calcium imaging was conducted on layer V pyramidal cells in prefrontal cortical region (e.g., cingulate, prelimbic and infralimbic cortices) with a stimulation electrode placed in layer II/III. c) Direct activation of 5-HT_2AR does not affect the number of evoked action potentials. (top) Representative cell-attached recordings demonstrate evoked action potentials in both baseline and DOI administration. (bottom) Summarized plot showing the input-output relationship between stimulation intensity and the number of evoked action potentials. Note that all representative traces in c-e are from recordings with 50% of stimulation intensity. d) Activation of 5-HT_2AR does not affect the evoked calcium fluctuation in recorded cell. (top) Representative stimulus-evoked traces (solid: mean; pale: individual) demonstrate evoked calcium fluctuation in recorded cells shown in b. (bottom) Summarized plot showing the input-output relationship between stimulation intensity and evoked calcium fluctuation. e) Activation of 5-HT_2AR does not affect the evoked calcium fluctuation in recorded slices. (top) Representative peri-stimulus traces (solid: mean; pale: individual) demonstrate evoked calcium fluctuation in the whole FOV. (bottom) Summarized plot showing the input-output relationship between stimulation intensity and evoked calcium fluctuation. f) Activation of 5-HT_2AR does not affect the evoked calcium fluctuation under blockade of synaptic transmission. (upper left) Global calcium imaging recording was conducted on layer V pyramidal cells in prefrontal cortices with stimulation electrode locally placed in layer V under kynurenic acid (5mM) and picrotoxin (100μM). (lower left) Representative peri-stimulus traces (solid: mean; pale: individual) demonstrate evoked global calcium fluctuation. (right) No significant difference in evoked calcium fluctuation. Abbreviations: aca, anterior commissure (anterior part); AI, anterior insular cortex; AP, action potential; Cg1, cingulate cortex (area 1); IL, infralimbic cortex; M1, primary motor cortex; M2, secondary motor cortex; PrL, prelimbic cortex.

Extended Data Figure 5. Hallucinogenic 5-HT_2A receptor agonism alters resting-state activity.

Global (i.e., regionally averaged) PSDEs were integrated over double-octave bins and compared between compounds (Saline, DOI (4mg/kg), MDL (0.1mg/kg), DOI+MDL, $n=8$) for both calcium (a) and hemodynamic (b) activity. The lowest frequency bin was limited by the lower limit of 0.02Hz due to the window length used to segment resting-state from stimulus-evoked data. All measures were averaged over the cortex and comparisons were made between pre- and post-injection (post- minus pre-) and across compounds using two-sided t-tests (*<0.05, **<0.01, and ***<0.001) and one-way ANOVAs with post-hoc, two-sided, t-tests (#<0.05, ##<0.01, ###<0.001), respectively. Multiple comparisons were corrected for using the Bonferroni method. c) Fractional changes in broad-band activity attributed to hallucinogenic 5-HT_2A receptor agonism. Fractional changes in regional power spectra are displayed as the ratio of post-injection values divided by pre-injection values and regions are organized from anterior-to-posterior and from left-to-right. DOI differentially affected the spectral content of calcium (left) and hemodynamic (right) activity in a region-dependent manner. ISA: DOI increased calcium ISA power in frontal, cingulate, and motor cortices while hemodynamic ISA activity increased primarily in somatosensory regions. Intermediate: Calcium activity decreased in all cortical regions excluding frontal and cingulate. In contrast, hemodynamic activity decreased in frontal and cingulate cortex. Delta: Both calcium and hemodynamic activity exhibited increased delta band activity in nearly every cortical region examined with the largest increases occurring at ~0.8Hz hemodynamic activity. Across all frequencies examined, MDL largely reversed the effects of DOI. Saline and MDL minimally altered hemodynamics and calcium spectral content. Average fractional changes over the cortex are visualized on top of each column and plotted as mean +/− std across mice.

Extended Data Figure 6. Hallucinogenic 5-HT_2A receptor agonism alters resting-state neurovascular coupling.

a) Neurovascular coupling parameterization and quantification. Regional hemodynamic response functions (HRFs) were characterized by their peak value (peak), time to peak amplitude (TTP), and full width at half maximum (FWHM). These parameters were averaged across regions and compared across compounds (Saline, DOI (4mg/kg), MDL (0.1mg/kg), DOI+MDL, $n=8$). Predicted hemodynamic activity was calculated by convolving each region’s HRF with its corresponding measured calcium signal. Model fit was assessed via Pearson’s r between predicted and measured hemodynamic activity. Band-limited frequency transduction was calculated over three frequency bands: infraslow activity band (ISA, 0.03–0.08Hz), intermediate activity band (0.08–0.50Hz), and delta activity band (0.5–4.0Hz). The lowest frequency bin of the TF (i.e., 0.03Hz) was determined by the 30s long window used for estimating HRFs. All global measures were averaged over the cortex and comparisons were made between pre- and post-injection and across compounds (post- minus pre-) using two-tailed t-tests (*<0.05, **<0.01, and ***<0.001) and one-way ANOVAs evaluated post-hoc via two-tailed t-test (#<0.05, ##<0.01, ###<0.001), respectively. Multiple comparisons were corrected for using the Bonferroni method. b) Hallucinogenic 5-HT_2A receptor agonism alters region-specific NVC. Estimated HRFs were averaged across mice for each region and compound. Regions are organized from anterior-to-posterior and from left-to-right. Prior to compound injection, NVC differed (parameterized in Fig. 5c) across the cortex. For instance, somatosensory and parietal regions exhibited the strongest coupling (e.g., large peak value), while frontal and cingulate regions exhibited more modest coupling. DOI caused a zero-lag notch across the entire cortex (i.e., the acausal feature seen in Fig. 5a, black arrow), strongest motor, retrosplenial, and visual regions. Additionally, DOI increased coupling in the somatosensory and parietal regions. These effects were dampened when DOI was administered with MDL. Cortical-averaged HRFs are visualized above each column and plotted as mean +/− std across mice. The color axis has units of $\Delta \mu$Mol/$\Delta$F/F(%).

Extended Data Figure 7. Lagged cross-covariance and coherence between calcium and total hemoglobin recapitulate DOI-induced changes in neurovascular coupling.

a) Global hemodynamic response functions: The global HRF estimated using deconvolution following DOI injection (4mg/kg, $n=8$, top; Fig. 4a) and the global lagged cross-covariance function (CCF; bottom) between resting-state calcium and hemodynamics. Pre-injection CCFs display a simple, quasi-causal relation (calcium leads hemodynamics). DOI decreased coupling (peak value, −51% [−66%, −13%], pre- vs. post-injection: p=0.039; Kruskal-Wallis: p=0.004; saline vs. DOI: p=0.020; DOI vs. DOI+MDL: p=0.004; DOI vs. MDL (0.1mg/kg): p<0.001). DOI induced a zero-lag notch in the CCF (black arrow). Covariance has units of $\Delta \mu$Mol*$\Delta$F/F(%) and the HRF has units of $\Delta \mu$Mol/$\Delta$F/F(%). b) DOI-induced alterations in CCF over the cortex. Regional distribution of cross covariance after the injection of DOI reveals a negative-lag peak across the cortex. CCFs for all regions and compounds are displayed in Fig. S6a. c) Magnitude Squared Coherence (MSC) between calcium and hemodynamic activity. Before compound injection, MSC exhibited a large 0.2Hz peak, demonstrating the band-limited nature of NVC. After injection of DOI, a coherence peak ~0.8Hz (~0.5Hz half-bandwidth) emerged (0.5–2.0Hz: +170% [110% 230%]; pre- vs. post-injection: p=0.008; Kruskal-Wallis: p=0.004; saline vs. DOI: p=0.009; DOI vs. DOI+MDL: p=0.030; DOI vs. MDL: p=0.025). Moreover, the 0.2Hz peak present before injection largely diminished (<0.5Hz: −47% [−53%, 49%], pre- vs. post-injection: p=0.008). This phenomenon signifies a DOI-induced shift in the coherent frequencies contained in both neuronal and hemodynamic activity. DOI+MDL largely reversed the effects of DOI alone (Fig. S6). Coherence for all regions and compounds are displayed in Fig. S6b. d) Phase relation between calcium and hemodynamics. Prior to injection, global calcium activity led hemodynamic activity (positive slope at frequencies <~1Hz; left). After DOI injection, phase relations between calcium and hemodynamic activity flipped sign (e.g., from positive to negative slope in primary motor and primary somatosensory hindpaw; right), indicating that hemodynamics precede calcium over these frequencies. Fig. S6c reports phase relations for all regions and compounds. All data are presented mean +/− std across mice. Significance was determined via Kruskal-Wallis tests and post-hoc, two-sided Wilcoxon’s sign-rank corrected post-hoc using Bonferroni correction.

Extended Data Figure 8. Non-hallucinogenic doses of DOI and the non-hallucinogenic, 5-HT_2AR agonist, Lisuride do not alter NVC.

a) Head twitch responses (HTRs) recorded during from sub-hallucinogenic DOI (low-dose, 0.04mg/kg) and the non-hallucinogenic, psychedelic ligand, Lisuride (0.1mg/kg). No differences in HTRs were observed between groups (Kruskal-Wallis, p>0.99). HTRs were recorded for 30 minutes ($n=4$). Significance was determined via Kruskal-Wallis tests and evaluated post-hoc via two-sided Wilcoxon’s sign-rank test corrected for multiple comparisons using Bonferroni correction. a-f) WFOI imaging of non-hallucinogenic ligands. Two groups of Thy1-jRGECO1a mice were imaged under resting-state conditions for 30min before and 30min after injection. Cohort 1 (4M, 5F) received saline and Lisuride (0.1mg/kg); cohort 2 (2M, 2F) received saline and a sub-hallucinogenic dose of DOI (low-dose DOI, 0.04mg/kg. b) Hemodynamic response functions before and after injection of low-dose DOI. Low-dose DOI did not alter global estimates of NVC. c) Regional changes NVC following low-dose DOI. Full-width-at-half-maximum (FWHM), time-to-peak (TTP), and peak value are visualized across the cortex. Sub-hallucinogenic doses of DOI did not alter any regional HRF parameters compared to Saline. d) Hemodynamic response functions before and after Lisuride injection. No significant changes to global hemodynamic response functions were observed after the injection of Lisuride. e) Regional changes NVC following Lisuride. Parameterized hemodynamic response functions reveal no differences in regional NVC after injection of low-dose DOI. f) Global power spectral density estimates before and after injection of Lisuride and d) Integrated, band-limited power over double octave frequency bins. Spectra are reported as median and shaded 25^th and 75^th percentiles. Band-limited power is consistently decreased after injection of both saline and Lisuride, like results reported in Extended Fig. 5a,b. Lisuride modestly affected power over the 0.32–1.28Hz bin (p=0.009). Band-limited power was tested for significance via Kruskal-Wallis tests and evaluated post-hoc via two-tailed Wilcoxon’s sign-rank corrected for multiple comparisons using Bonferroni correction.

Extended Data Figure 9. Hallucinogenic 5-HT_2A receptor agonism alters inter- vs. intra-network connectivity.

a) Whole-cortex RSFC matrices. Average, pre-injection, whole-cortex RSFC matrices across mice ($n=8$) for infralow (ISA; 0.02–0.08Hz) calcium and HbT activity and delta (0.5–4Hz) calcium activity. Matrices represent all pixel-pair comparisons and are organized by network assignment. b) Hierarchical clustering of RSFC. RSFC matrices were hierarchically clustered using Ward’s method and visualized as a dendrogram. c) Resultant clusters at a specific hierarchical level. A distance threshold was selected to create a parcellation containing 13 clusters. These parcels were used to assess inter- and intra-regional relationships. d) Pre-injection Silhouette scores. Silhouette scores were computed at each pixel to assess the ratio between cohesion (intra-cluster distance metric) and separation (inter-cluster metric). Silhouette scores range from −1 to +1, such that +1 indicates a point is well-clustered and far from neighboring clusters (strongly intra-connected, weakly inter-connected), 0 indicates that the point is on, or near, the boundary between two clusters (equally intra- and inter-connected/equally likely to belong to any neighboring cluster), and −1 indicates a point is misclassified and likely should be assigned to a different cluster than its own (weakly intra-connected, strongly inter-connected). i) A graphical demonstration of cohesion (black cluster) and separation to the nearest cluster (red cluster). ii) Pre-injection silhouette scores for each species and frequency band. e) Changes in Silhouette scores after compound injection. DOI (4mg/kg) predominantly decreases ISA calcium silhouette scores everywhere (regions are less intra-connected and more inter-connected), an effect not fully reversed in under DOI+MDL. Importantly, ISA calcium scores in the cingulate and retrosplenial cortices—constellations of the mouse default mode network—showed opposite effects: cingulate scores marginally increased, while retrosplenial scores markedly decreased. This divergence was substantially reduced when scores were computed using infralow hemodynamic activity. Delta calcium activity under DOI suggests that the compound primarily alters network boundaries rather than modulating inter- or intra-network interactions

Data Availability Statement

The Allen Mouse Brain Atlas was downloaded from: https://alleninstitute.github.io/AllenSDK/reference_space.htmlProcessed

Processed WFOI data are available here: https://doi.org/10.5281/zenodo.15857641

Raw Head twitch response data are available here: https://doi.org/10.5281/zenodo.15857233

All other data supporting the findings of this study will be made available upon request.

Code Availability Statement

Matlab processing code is available on GitHub: https://github.com/BauerLabCodebase/WFOI-Textbook-Chapter.