Functional dissection of neural circuitry using a genetic reporter for fMRI

Abstract

The complex connectivity of the mammalian brain underlies its function, but understanding how interconnected brain regions interact in neural processing remains a formidable challenge. Here we address this problem by introducing a genetic probe that permits selective functional imaging of distributed neural populations defined by viral labeling techniques. The probe is an engineered enzyme that transduces cytosolic calcium dynamics of probe-expressing cells into localized hemodynamic responses that can be specifically visualized by functional magnetic resonance imaging. Using a viral vector that undergoes retrograde transport, we apply the probe to characterize a brain-wide network of presynaptic inputs to the striatum activated in a deep brain stimulation paradigm in rats. The results reveal engagement of surprisingly diverse projection sources and inform an integrated model of striatal function relevant to reward behavior and therapeutic neurostimulation approaches. Our work thus establishes a strategy for mechanistic analysis of multiregional neural systems in the mammalian brain.

Editor summary:

This study introduces genetically encoded imaging probes that convert intracellular calcium signaling into hemodynamic fMRI responses. The authors show how the probes can be used to map information flow in reward-related brain circuitry in rats.

INTRODUCTION

Optical measurements of signaling dynamics in well-defined neural circuit components are possible by combining invasive optical imaging methods with fluorescent probes like the genetically encoded calcium sensor GCaMP,¹ but these approaches only operate over limited volumes. To permit measurements of circuit-level processing on a brain-wide scale, we sought to construct an analog of GCaMP that could be detected by noninvasive imaging methods like functional magnetic resonance imaging (fMRI),^{2, 3} functional photoacoustic tomography,⁴ or functional ultrasound.⁵ These methods usually operate by monitoring blood flow changes evoked nonspecifically by neural population activity.⁶ We reasoned that a molecular probe designed to couple the intracellular signaling activity of genetically targeted cells to hemodynamic changes could form the basis of cell-specific functional neuroimaging, provided that output from the probe could be adequately distinguished from endogenous hemodynamic signatures (Fig. 1a). We refer to this concept as hemogenetic imaging, because it employs a genetic probe to produce a hemodynamic readout.

Figure 1 |. Principle of hemogenetic imaging with NOSTICs.

a, Hemogenetic imaging involves detection of localized changes in blood flow or volume (green arrows) induced by genetically encoded activity reporters (cyan) analogous to fluorescent probes like GCaMP. Hemogenetic reporters signal to nearby blood vessels (cyan wireless symbol), inducing hemodynamic image contrast that can be detected by fMRI and related methods. Endogenous drivers of activity-dependent signals (dark blue) are suppressed, silenced, or subtracted. b, NOSTIC-1 and NOSTIC-1P reporters are calcium-dependent chimeric enzymes that combine the catalytic domain of iNOS with calmodulin-binding (CaM) and reductase domains of nNOS; NOSTIC-1P also contains the synapse-targeting nNOS PDZ domain. c, Measurement of nitrite formation following NO release from cells expressing nNOS, iNOS, NOSTIC-1 (N-1), and NOSTIC-1P (N-1P) in absence (left) or presence (right) of stimulation using the Ca²⁺ ionophore A23187. Calcium-dependent induction of nNOS, N-1, and N-1P is significant with two-sided t-test p ≤ 0.0018 (n = 3). Error bars denote SEM of three measurements. d, Activity measurements titrated with the nNOS-selective inhibitor L-NAME (left) or iNOS-selective inhibitor 1400W (right). Horizontal dashed lines denote half-maximal activity levels, at which IC₅₀ values of the inhibitors apply. Error bars denote SEM of three measurements each. e, Sensory-driven fMRI signals in rat cortex visualized with or without NOS inhibition using 50 mg/kg L-NAME (top, n = 4) or 25 mg/kg 1400W (bottom, n = 3). Images display percent signal change (%SC, color bar) superimposed on an anatomical scan. Box plots denote median (center line), first quartiles (boxes), and full range (whiskers) of fMRI responses observed in S1 forelimb region, over animals. The L-NAME effect is significant with two-sided t-test p = 0.04.

We chose a molecular basis for hemogenetic reporters by considering contributions to normal neurovascular coupling in mammalian brains. Although many signaling pathways have been implicated,⁷ a particular strong effect is mediated by neuronal nitric oxide synthase (nNOS). This enzyme catalyzes local formation of the cell-permeable gaseous vasodilator nitric oxide (NO) during neural signaling. Optogenetic stimulation of nNOS-positive neurons in the brain produces robust hemodynamic responses that scale with stimulus strength,⁸ and pharmacological nNOS inhibition substantially reduces fMRI responses to sensory stimuli.^{9, 10} Neuronal signaling activates nNOS via its association with calcium-bound calmodulin, the same protein domain that actuates fluorescent sensors like GCaMP.¹¹ These considerations led us to design hemogenetic probes termed “nitric oxide synthases for targeting image contrast” (NOSTICs) by engineering an nNOS isoform that retains calcium sensitivity but that can be selectively manipulated by drugs that do not perturb intrinsic hemodynamic responses.

In this work, we sought to establish the principle of hemogenetic imaging with NOSTIC probes and begin to explore applications in neurobiology. Our results comprise protein engineering of the NOSTIC reporters, in vitro and in vivo validation studies, and functional analysis of neural circuitry in rodent brains using the approach. We demonstrate that hemogenetic fMRI provides brain-wide information about neural information flow that can be combined with complementary data to yield insights into brain mechanisms.

RESULTS

Construction and validation of hemogenetic reporters

First-generation NOSTICs, called NOSTIC-1 and NOSTIC-1P, were created by replacing the catalytic domain of nNOS with corresponding residues from inducible nitric oxide synthase (iNOS), an immune-associated NOS variant that displays distinctive sensitivity to inhibitors (Fig. 1b and Supplementary Table 1).¹² NOSTIC-1P further contains the PDZ domain of nNOS, which directs association with synaptic proteins,¹³ and which NOSTIC-1 lacks. These probes therefore combine NOS activity with the possibility of cellular or subcellular targeting.

Calcium dependence and pharmacological sensitivity of the NOSTIC probes were examined in cell culture. Both NOSTIC-1 and NOSTIC-1P express robustly in HEK293 cells, with primarily cytosolic expression consistent with previous reports¹⁴ (Supplementary Fig. 1). To evoke cytosolic calcium elevations similar to those observed in a variety of signal transduction mechanisms, we applied the widely used calcium ionophore A23187. Stimulation with this drug evokes comparable levels of NO production by both NOSTICs and nNOS (Fig. 1c). Calcium-dependent NO increases for nNOS, NOSTIC-1, and NOSOTIC-1P are all significant with t-test p ≤ 0.0018 (n = 3). The nNOS-selective inhibitor L-nitroarginine methyl ester (L-NAME) blocks this effect with a concentration producing 50% inhibition (IC₅₀) of 20 μM for nNOS, but more than 50 μM for the NOSTICs (Fig. 1d). In contrast, the iNOS-specific blocker 1400W inhibits NOSTICs with IC₅₀ values of about 25 μM, with over tenfold lesser effect on nNOS. Stimulation and pharmacological blockade of NOSTIC-dependent NO production could also be observed in a neuronal cell line (Extended Data Fig. 1), indicating that NOSTICs are suitable for brain applications and functional in multiple cell types. Importantly, drugs that respectively inhibit nNOS and NOSTICs produce sharply different effects on endogenous fMRI responses, as tested using a sensory stimulation paradigm in rats (Fig. 1e, Extended Data Fig. 2, Supplementary Fig. 2). While L-NAME injection reduces the cortical response to electrical forepaw stimulation by 2.42 ± 0.53% (significant with paired t-test p = 0.04, n = 4), 1400W produces virtually no change (0.15 ± 0.07% increase, paired t-test p = 0.29, n = 3). These results thus demonstrate a basis for drug-dependent selective imaging of NOSTIC reporters by hemodynamic neuroimaging.

To validate the probes in vivo, we examined the ability of NOSTIC-expressing cells to induce activity-dependent hemodynamic imaging signals following xenotransplantation into rat brains. Only NOSTIC-1 was used for these studies, because of its greater calcium-dependent NO production compared with NOSTIC-1P in vitro. HEK293 cells stably expressing NOSTIC-1 were implanted into the striatum and imaged using a 9.4 T magnetic resonance imaging (MRI) scanner during stimulation with 10 μM A23187 (Fig. 2a and Supplementary Fig. 3). Calcium ionophore treatment induces robust blood oxygen level-dependent (BOLD) fMRI responses in the neighborhood of the cells (Fig. 2b). In contrast, control experiments performed in the presence of the NOSTIC inhibitor 1400W or using cells that express green fluorescent protein (GFP) in place of NOSTIC-1 result in negligible fMRI responses. The A23187-induced NOSTIC signal persists for ~10 min, likely reflecting slow washout of the ionophore (Fig. 2c), and results in a mean signal increase of 6.2 ± 1.4% (Fig. 2d) that is significantly greater than both controls (t-test p ≤ 0.001, n = 5). The full-width at half-height of the NOSTIC response averages 1.4 ± 0.1 mm, similar to the mean xenograft diameter of 1.8 ± 0.5 mm (Extended Data Fig. 3). This implies that the point-spread function width of NOSTIC-dependent fMRI signals is comparable to or less than the value of ~1 mm reported for intrinsic BOLD fMRI responses at 9.4 T.¹⁵ These experiments show that stimulus-induced NO production by implanted cells is sufficient for induction of local BOLD responses and demonstrate that NOSTICs can selectively report the activation of probe-expressing cells in fMRI.

Figure 2 |. NOSTIC-1 reports activation of genetically-modified cells in vivo.

a, Timeline of intracranial implantation and imaging of NOSTIC-expressing or control cells stimulated by infusion of A23187 in the absence or presence of the NOSTIC-1 inhibitor 1400W. b, BOLD MRI contrast changes induced by A23187-mediated stimulation: (top) NOSTIC-1-expressing cells without 1400W, (middle) NOSTIC-1 cells with 1400W, (bottom) GFP-expressing cells without 1400W. Each panel shows three coronal slices within 1 mm of the cell implantation site with a percent signal change (%SC) map overlaid on a representative anatomical scan. c, Mean time courses of MRI signals in ROIs centered around the cell implantation sites in b for animals implanted with NOSTIC-1-expressing cells in the absence (cyan) and presence (dark blue) of 1400W, and control GFP cells without 1400W (dark gray). Stimulation period denoted by light magenta box; shading denotes SEM of n = 6 (NOSTIC-1 −1400W), n = 4 (+1400W), or n = 4 (GFP). d, Average signal change amplitudes corresponding to the data in c. Box plots denote median (center line), first quartiles (boxes), and full range (whiskers) of response amplitudes across animals. The NOSTIC-1 −1400W response is statistically significant with two-sided t-test p = 0.006 (n = 6), while the control responses are not significant with two-sided t-test p ≥ 0.11 (n = 4).

Functional imaging of targeted neural circuitry with NOSTIC

We next applied NOSTIC-1 to analyze neural circuit function in the striatum, a major hub for integration of sensory, motor, and affective input from remote brain areas, and a target of deep brain stimulation (DBS) therapies in the clinic. Understanding which connections provide active inputs during behaviorally relevant tasks or stimuli is essential to formulating circuit-level models of striatal function. Anatomical connections of the striatum have been studied by applying viral vectors that undergo axonal transport and express fluorescent proteins detectable in postmortem microscopy;^16–19 functional assessment of the projections should therefore become possible by using similar vectors to drive expression of NOSTIC-1, followed by functional imaging. We used this approach to measure striatal input during electrical stimulation of the lateral hypothalamus (LH),²⁰ a widely used addiction model in rodents and an exploratory DBS treatment for refractory obesity and depression in patients.^{21, 22} A specific unanswered question in this context is whether striatal activation due to LH stimulation arises primarily from excitation of striatal afferents passing through this region,²³ or whether alternative circuit-level mechanisms might help explain striatal responses and associated behavioral effects.

To label direct striatal afferents with NOSTIC-1, we constructed a herpes simplex virus (HSV) vector encoding NOSTIC-1 and mCherry (HSV-hEF1α-NOSTIC-IRES-mCherry). Similar vectors have been shown to produce persistent expression following retrograde transport from sites of injection in the rodent brain,²⁴ implying that NOSTICs delivered this way should report the activity of neural populations that provide direct presynaptic input to striatal cells from throughout the brain. After confirming the ability of HSV-hEF1α-NOSTIC-1-IRES-mCherry to induce intracellular calcium-dependent NO production in cultured cells (Extended Data Fig. 4), we injected the virus into the anterior central caudate putamen (CPu) of six rats implanted with electrodes targeted to the medial forebrain bundle region of LH (Fig. 3a). Histological results confirmed that virally driven probe expression does not strongly correlate with inflammatory markers (Supplementary Fig. 4). Three weeks after viral transduction, animals were sedated with medetomidine and imaged by BOLD fMRI during alternating blocks of LH stimulation and rest, first before and then after systemic treatment with the NOSTIC-1 inhibitor 1400W. These data permit identification of NOSTIC-specific signals based on the difference between fMRI responses observed in the absence vs. presence of the drug (Fig. 3a), reflecting presynaptic functional input to CPu during LH stimulation.

Figure 3 |. Hemogenetic functional imaging of striatal circuitry.

a, Approach for hemogenetic analysis of striatal circuit function: An HSV vector delivers NOSTIC-1 to neurons presynaptic to a striatal injection site (cyan). Functional imaging during stimulation via an LH-targeted electrode (magenta) performed in absence (top) and presence (middle) of inhibition with 1400W, which suppresses NOSTIC-mediated fMRI signals but not endogenous nonspecific fMRI responses (purple). NOSTIC-specific responses to DBS are obtained by subtraction (bottom). b, Average fMRI responses to LH stimulation in the −1400W condition, among six animals infected with NOSTIC-encoding HSV. Significant responses with F-test p ≤ 0.01 shown. c, Mean of fMRI response amplitude differences between −1400W and +1400W conditions observed for animals infected with NOSTIC-1 (n = 6) or mCherry-only (n = 5) viruses; select ROIs labeled. d, Average time courses of fMRI signals observed in CPu under each condition. Pink box denotes stimulation period. Shading denotes SEM over six animals. e, Difference signal amplitudes for NOSTIC-1 (left, n = 6) or mCherry control animals (right, n = 5) in eight ROIs labeled in c (also see Supplementary Fig. 1). Box plots denote median (center line), first quartiles (boxes), and full range (whiskers) over multiple animals. f, Absolute fMRI response amplitudes illustrating consistent 1400W-dependent responses in CPu, ECx, MCx, and SN, all significant with two-sided paired t-test p ≤ 0.05 (n = 6).

A broad distribution of brain areas is activated both before and after 1400W treatment (Fig. 3b, Extended Data Fig. 5, Supplementary Fig. 2), but difference maps denoting fMRI amplitudes specifically attributable to NOSTIC-expressing cells reveal a limited set of peaks (Fig. 3c). These maps are calculated by subtracting response time courses computed in unbiased fashion for each voxel before vs. after 1400W treatment, but a more rigid fMRI analysis that assumes a fixed hemodynamic response for all voxels also produces similar results (Supplementary Fig. 5). In both analyses, 1400W-dependent difference peaks are largely absent in data from rats transduced with the control vector HSV-hEF1α-mCherry in place of the NOSTIC-1 vector (Fig. 3c, Extended Data Fig. 6, Supplementary Figs. 5). Difference signals observed in the maps of Fig. 3c reflect 1400W-dependent scaling of the time courses of underlying fMRI signals (Fig. 3d and Extended Data Fig. 7), which in some brain regions display a characteristic extended profile²⁵ shown previously to parallel striatal electrophysiological recordings.²⁶ Similarity of time courses observed in the absence vs. presence of 1400W indicates that NOSTIC-dependent hemogenetic signals are temporally similar to endogenous BOLD responses, suggesting a shared mechanistic basis for intrinsic and NOSTIC-mediated fMRI signals.

Difference signals from each animal were further analyzed over eight anatomically defined regions of interest (ROIs) ipsilateral to the LH stimulation site, focusing on areas in which peak effects of Fig. 3c are observed. Average fMRI difference amplitudes in several of these ROIs are substantially higher in NOSTIC animals than in rats transduced with the control mCherry-expressing HSV (Fig. 3e), providing a basis for statistically robust analysis of differences between the two conditions. No significant ROI-level difference signals are observed in the controls (paired t-test p ≥ 0.08, n = 5), but 1400W-dependent fMRI signal differences from the NOSTIC animals are statistically significant with paired t-test p ≤ 0.05 (n = 6) in CPu (2.2 ± 0.5% signal difference), entorhinal cortex (ECx, 2.6 ± 0.8%), motor cortex (MCx, 2.9 ± 1.1%), and substantia nigra (SN, 1.7 ± 0.6%) (Fig. 3f and Extended Data Fig. 8). When fMRI response amplitude differences are expressed as percent change in pre-1400W signal (Extended Data Fig. 9), NOSTIC results in CPu, ECx, MCx and SN are also especially pronounced, with 1400W-induced reductions of 38–51% and paired t-test p ≤ 0.008 (n = 6) in each case. Group-averaged 1400W-dependent fMRI amplitude differences are significantly greater in NOSTIC animals than in controls when compared across all ROIs examined (paired t-test p = 0.046, n = 8 ROIs) or over CPu, ECx, MCx, and SN in particular (paired t-test p = 0.0065, n = 4 ROIs), providing further evidence for specificity of the results. NOSTIC-based hemogenetic imaging results thus identify multiple brain regions that provide consistent presynaptic contributions to the striatum during LH stimulation, including cortical sources with striatal afferents that do not traverse the DBS target site.¹⁶

Confirmation of hemogenetic results using other techniques

To assess alignment of the hemogenetic fMRI data with established anatomical and functional measurement techniques, we began by comparing NOSTIC imaging results to structural connectivity patterns revealed by retrograde viral tracing. Unbiased brain-wide histological analysis using SHIELD-based tissue clearing²⁷ in animals infected by CPu injections of an mCherry-encoding HSV shows prominent fluorescence corresponding to cells in CPu, ECx, MCx, medial thalamus (MT), and SN (Fig. 4a); additional features are observable in individual tissue slices (Fig. 4b). As expected, the brain regions most heavily labeled by striatal injection of the HSV vector thus include the sources of striatal input most reliably identified by NOSTIC-based fMRI studies.

Figure 4 |. Alternative measures of striatal input.

a, Unbiased fluorescence mapping in a cleared brain hemisphere imaged after HSV-mCherry delivery to rat CPu. Areas with substantial mCherry expression are labeled, along with the olfactory bulb (OB), for reference. Dorsal (D), medial (M), and caudal (C) directions denoted. Scale bar = 2 mm. b, HSV-driven mCherry expression in specific slices, with anatomical ROIs labeled on the atlas overlay (white) and bregma coordinates in yellow. Scale bars = 1 mm. c, Visualization of c-Fos protein (green) and HSV-directed mCherry expression (red) in animals infected with NOSTIC-encoding HSV and stimulated as in the fMRI experiments. Arrowheads indicate co-stained cells. Scale bar = 100 μm. d, Data from two regions without significant NOSTIC responses, showing c-Fos staining only in CCx (top) and non-overlapping c-Fos and mCherry staining in MT (bottom). Scale bar = 100 μm. e, Quantification of mCherry (red), c-Fos (green), and overlapping (yellow) staining densities in six ROIs (bar graphs), also showing relative frequencies at top (pie charts); n = 4 animals in each case. f, Comparison of mCherry and c-Fos double-stained cell densities vs. NOSTIC amplitudes in six ROIs. Error bars denote SEM of n = 4 (horizontal) or n = 6 (vertical). g, Schematic of photometry experiments, showing striatal HSV-GCaMP6s injection site (green), LH electrode (magenta), and fiber placement (gray) in a sagittal section at top, with histological visualization of GCaMP6s-expressing SN neurons at bottom. Scale bar = 200 μm. h, Representative photometry traces indicating relative change in fluorescence (ΔF/F) from striatally projecting SN cells in a GCaMP6s-expressing animal (top) and a GFP-expressing control (bottom). Pink box denotes LH stimulation period. Shading = SEM over 40 stimulus presentations. i, Photometry response amplitudes recorded from three test and control animals, as in panel h (difference significant with t test p = 0.028).

To complement this purely anatomical validation with a functional readout, we combined HSV-mCherry retrograde tracing from CPu with measurements of activity-dependent c-Fos protein expression induced by the same stimulus protocol used in the fMRI experiments. We predicted that striatal input regions identified in the NOSTIC experiments of Fig. 3 would show the greatest density of c-Fos-positive cells also labeled by HSV-mCherry. Microscopy images confirm that doubly stained cells are indeed much more readily observed in CPu, ECx, MCx, and SN (Fig. 4c) than in two brain areas, cingulate cortex (CCx) and MT, where NOSTIC signals are not statistically significant (Fig. 4d). Results of this analysis are quantified in Fig. 4e and indicate densities of c-Fos-positive HSV-labeled cells from 5.3 to 25 cells/mm² in CPu, ECx, MCx, and SN, but only 0.066 and 1.2 cells/mm² in CCx and MT, respectively. Across animals, c-Fos/mCherry-labeled cell densities in CPu and SN in particular are significantly higher than values in CCx and MT (t-test p ≤ 0.016, n = 4). Comparison of double-labeled cell densities with NOSTIC signals in the six regions analyzed (Fig. 4f) reveals positive, though not statistically significant correlations of R = 0.3 when all six brain areas are included or 0.8 when SN is omitted; treatment of SN as an outlier may be justified by the fact that this small structure is more susceptible to partial volume effects in the fMRI analysis. In any case, the broad agreement between c-Fos histochemistry and NOSTIC results we observed provides further experimental support for the hemogenetic results of Fig. 3 and suggests that larger NOSTIC-mediated fMRI amplitudes on average reflect greater densities of stimulus-activated neurons in each brain region.

To support the NOSTIC-based fMRI results using a more dynamic activity measure, we used a retrogradely transported HSV vector to express the genetically encoded fluorescent calcium sensor GCaMP6s in place of the NOSTIC-1 probe; the virus was again introduced into CPu and fiber photometry was applied to measure from striatally projecting SN neurons during LH-targeted DBS (Fig. 4g). Photometry recordings reveal stimulus-evoked calcium transients that are absent when a GFP-encoding virus is used in place of HSV-GCaMP6s (Fig. 4h and Supplementary Fig. 6). Mean fluorescence change amplitudes of 2.2% and 0.1% are observed in GCaMP6s and GFP conditions, respectively (Fig. 4i). This difference is significant with t-test p = 0.028 (n = 3), and confirms the presence of an LH stimulus-activated neural population that projects from SN to CPu. Although we did not perform photometry on further striatal input regions identified by the hemogenetic imaging approach, the SN-to-CPu result is particularly important because of longstanding uncertainty about whether striatal input evoked by LH stimulation travels directly from the DBS site or indirectly, via the midbrain.²³ NOSTIC, photometry, and c-Fos results all indicate that the indirect route is at least partially involved.

Comparing hemogenetic imaging with complementary measures

Although NOSTIC-based hemogenetic imaging in its current form requires two measurements, the resulting two datasets provide information not only about probe-expressing neural populations, but also about intrinsic population activity. By comparing NOSTIC-mediated responses to intrinsic nonspecific fMRI signals recorded from the same animals in the presence of 1400W, the response profiles of neural populations with differing synaptic relationships to the striatum can be distinguished (Fig. 5a). NOSTIC and nonspecific fMRI signals are weakly negatively correlated, with R = −0.56 (p = 0.15, n = 8 ROIs). Several regions—CCx, insular cortex (ICx), MT, and S1—show statistically significant nonspecific fMRI responses to LH stimulation (t-test p ≤ 0.045, n = 6) but not significant NOSTIC responses. Although NOSTIC signals in ECx and MT are significantly different from each other (p = 0.022), the nonspecific imaging signals in these regions do not differ; conversely, although intrinsic fMRI signals differ significantly between CCx vs. SN, CPu vs. MT, and MCx vs. MT (p ≤ 0.044), NOSTIC signals among these regions could not be statistically discriminated. Ratios of mean NOSTIC signals to intrinsic fMRI signals are highest in CPu (1.1 ± 0.4, significant with t-test p = 0.047) and MCx (1.3 ± 0.6), suggesting that a particularly high proportion of stimulus-responsive cells in these two areas provide direct input to striatal HSV infection sites. The lowest relative NOSTIC signals are observed in ICx (0.22 ± 0.19) and MT (0.24 ± 0.15), suggesting that stimulus responses in these regions are elicited by striatal output or by pathways that are not directly related to striatal activity. These results collectively show that intrinsic and NOSTIC-mediated fMRI measures convey distinct information about brain responses to LH stimulation.

Figure 5 |. Multimodal analysis of striatal input/output relationships.

a, Comparison of NOSTIC signal amplitudes vs. nonspecific fMRI responses to LH stimulation (both n = 6) recorded in the presence of 1400W (top). Ratios of NOSTIC to nonspecific signals shown at bottom. b, Comparison of NOSTIC signals (n = 6) vs. Z-transformed resting state correlation coefficients (n = 4) at top, with ratios shown at bottom. c, Comparison of NOSTIC signals (n = 6) vs. retrogradely traced mCherry-labeled cell densities (n = 4) at top, with ratios shown at bottom. Error bars in panels a-c = SEM computed by jackknife resampling of group data; see text for additional statistics. d, Summary of hemogenetic imaging, nonspecific fMRI, resting state connectivity, and anatomical connectivity amplitudes across eight ROIs (left); more saturated colors denote statistically significant NOSTIC responses (cyan), mean nonspecific fMRI responses > 3% (indigo), resting state correlations to CPu with mean Z values > 0.8 (gray), and retrograde labeling of CPu-projecting regions containing > 5 cells/mm² (red). e, Sagittal brain diagram schematizing striatal circuitry consistent with the combined results. Circles denote DBS-evoked activation levels; lines denote functional connectivity defined by hemogenetic imaging (cyan), resting state analysis (gray, non-directional), or histology (red).

We next considered whether hemogenetic measurements of circuit-level functional connectivity are similar to traditional “resting state” functional connectivity assessments derived from cross-correlation of regional fMRI time courses in the absence of tasks or stimuli.²⁸ To explore this question, we acquired resting state fMRI data from four additional animals and computed Z-transformed correlation coefficients relating CPu time courses to each ROI (Fig. 5b and Extended Data Fig. 10). All eight ROIs display statistically significant nonzero Z-scores (t-test p ≤ 0.001, n = 4), but the relative strengths of each connection differ from those observed using the hemogenetic technique (R = −0.017, p = 0.97, n = 8 ROIs). While ECx and SN both exhibit significant NOSTIC signals (t-test p ≤ 0.028, n = 6), these two regions also display the lowest resting-state functional connectivity to CPu of the eight regions examined, with the SN Z-score significantly below those from all other regions except ECx (t-test p ≤ 0.018, n = 4). ECx and SN display the highest ratios of average NOSTIC signal to resting-state Z, with values significantly nonzero in ECx in particular (t-test p = 0.032, n = 6). The results therefore dissociate NOSTIC-based measurements from conventional resting-state connectivity and show that resting state analysis does not emphasize monosynaptic relationships engaged during LH-targeted stimulation.

The HSV-mCherry labeling data of Fig. 4, which indicate presynaptic anatomical connections to the CPu injection site, provide a third basis for comparison with our hemogenetic functional connectivity results. Quantification of mCherry fluorescence levels across regions of interest in four animals each shows that the density of HSV-labeled cells in each ROI does not directly scale with the amplitudes of NOSTIC-dependent activity signals in the experiments of Fig. 3 (Fig. 5c). Mean NOSTIC and mCherry labeling signals are uncorrelated (R = 0.37, p = 0.36, n = 8 ROIs), and the ratio of labeled cell density to NOSTIC signal varies considerably among regions. SN shows the greatest viral labeling density of 92 ± 26 cells/mm², significant above all ROIs except CPu and MCx (t-test p ≤ 0.027, n = 4), but displays lower NOSTIC responses than several other areas. Although average labeled cell densities above 5 cells/mm² were observed in all regions with statistically significant NOSTIC signals, moderate retrograde viral labeling was also consistently observed in MT (6.7 ± 1.9 cells/mm²; t-test p = 0.039, n = 4), a thalamic region in which LH stimulus-elicited NOSTIC signals were not pronounced. Our data thus demonstrate that connectivity strength judged using viral tract tracing does not predict functional responses measured using the hemogenetic imaging technique.

The combined brain-wide dataset consisting of circuit-specific hemogenetic readouts, nonspecific fMRI, resting state analysis, and anatomical tracing of striatal afferents (Fig. 5d) yields an integrated view of neural information flow in and around the rodent striatum (Fig. 5e). The NOSTIC-1 results—confirmed by c-Fos histochemistry and calcium photometry—identify ECx, MCx, and SN as distal input sources that shape striatal activity during stimulation, complementing local processing within CPu itself. Concurrent activation of cortical and midbrain afferents during LH-targeted DBS has not previously been documented, but might be important for behaviorally reinforcing properties of the stimulation.^{29, 30} Meanwhile, CCx, ICx, MT, and S1 are all strongly activated by LH stimulation and also correlated to CPu in the resting state, but these regions do not display statistically significant presynaptic input to CPu according to the NOSTIC-based fMRI measurements. A simple hypothesis to explain these results is that CCx, ICx, MT, and S1 are excited by striatal outputs, but we cannot rule out the possibility that alternative conduction pathways are involved; further application of hemogenetic imaging could resolve this issue. Medial thalamic input to striatum is notably not observed, despite notable anatomical and resting state connectivity between MT and CPu and activation of MT by LH stimulation. These input/output relationships might collectively generalize to other experimental contexts, particularly those related to reward processing and motor control.

DISCUSSION

The results we present here introduce and apply a sensitive strategy for brain-wide functional imaging of genetically targeted cells and neural projections. NOSTIC probes transduce intracellular calcium signaling into NO production, which in turn triggers hemodynamic responses that can be measured using fMRI and other neuroimaging modalities. Our studies demonstrate efficacy of the probes and their suitability for hemogenetic fMRI of neural populations labeled using a viral vector. ROI-level results are supported by established measures including activity-dependent c-Fos expression mapping and fluorescence recordings. The differential measurement strategy used in the current work provides readouts of both NOSTIC-specific and intrinsic stimulus-driven hemodynamic activity. By combining these measures with resting state fMRI and postmortem histological analysis, we define distinct relationships between function and anatomy in striatal circuitry during reward-related DBS.

NOSTICThe NOSTIC probes work by tapping into physiology that naturally relates neural activity to changes in blood flow. An advantage of this approach is that it exploits intrinsic biological amplification mechanisms to detect the function of even relatively sparse cell populations. That said, hemogenetic imaging will need to be further benchmarked against conventional measurements of neural activity in order to determine its sensitivity characteristics more thoroughly. It is likely that comparisons of intracellular calcium dynamics with NOSTIC signals over a range of activity levels and labeled cell densities will expose nonlinearities that are characteristic of neurovascular coupling.³¹ Such studies might also indicate whether perturbations are caused by the NOSTICs themselves, due to effects of excess NO production on neurophysiology or vascular biology.³² Side-effects could be mitigated if needed by inducing probe expression only when needed, or by inhibiting it prior to imaging experiments. An additional source of complexity could arise from uncertainty as to how exactly NO causes vasodilation; both direct vasodilatory and indirect effects of neuronal NO on the vasculature have been proposed, and astrocytes and endothelial cells may help mediate the action of neurogenic NO on vessels.^{33, 34} Such intricacies might particularly influence the performance of hemogenetic imaging in disease conditions where vascular status is altered or inflammatory NO production is a factor.³⁵ Comparison of NOSTIC-dependent results with control conditions and additional tests to rule out confounds will continue to be important as bases for rigorous experimentation using the hemogenetic approach.

Despite current limitations of the new technology, our analysis of striatal functional connectivity using the NOSTIC-1 activity reporter offers broader importance in several respects. First, this study demonstrates the complexity of responses to DBS. Although the LH stimulation we used is targeted at fibers that directly connect midbrain nuclei to CPu, our results demonstrate engagement of neural pathways that extend far beyond these structures. The activation of distal CPu afferents from cortex we observe could not be anticipated based on anatomical considerations alone, and yet might be critical for behavioral outcomes of the stimulation paradigm. Our data thus imply that both direct effects and broader circuit-level consequences of targeted stimulation are important in determining how DBS affects the brain. Results from the DBS paradigm are also relevant to striatal function more generally, although the stimulus we used may induce unphysiological patterns of signal propagation or neurochemical release that differentiate our experiments from more naturalistic scenarios.³⁶

Second, our results indicate limitations of conventional approaches for understanding brain-wide connectivity. Anatomical tracing fails to predict which presynaptic inputs to the striatum are most relevant to the stimulus paradigm we explored, and conventional fMRI assessments during stimulation and resting states likewise fail to highlight circuit-specific functional relationships identified by the hemogenetic imaging approach. NOSTIC-based imaging thus provides information that is difficult to obtain using other neural readouts, including the combination of fMRI with cellularly targeted stimuli such as optogenetic manipulations.^{37, 38} Although information about the activity of specific neuronal projections can be obtained using optical tagging³⁹ or genetically targeted fluorescent readouts such as the photometry we present in Fig. 4g–i, these approaches are technically challenging and difficult to apply over many brain regions in mammals. Meanwhile, methods such as the combined anatomical tracing and c-Fos histology we used for validation, as well as some forms of manganese-enhanced MRI,⁴⁰ can also combine sensitivity to anatomical and functional brain parameters, but they are not well-suited for monitoring activity dynamically on behaviorally relevant time scales.

Finally, the molecular neuroimaging strategy we used could be extended to many other experimental contexts. In the future, hemogenetic techniques can be applied in sedated or awake animals of many species, using virtually any wide-field imaging modality, and in conjunction with spatial or genetic targeting methods. Although we used an artificial stimulus to induce neural activity in our experiments, hemogenetic imaging may be compatible with more ecologically relevant stimuli or behavioral settings as well. As an alternative to hemodynamic imaging readouts, detection of NOSTIC probes could potentially be performed using molecular imaging strategies capable of directly measuring biogenic NO production in vivo.^41–43 The invasive viral injection methods employed here could be avoided if noninvasive viral delivery^44–46 or other transgenesis techniques prove feasible. Further efforts are required to evaluate probe performance across cell types, but our current results already demonstrate NOSTIC functionality in neural and non-neural cells and over multiple brain regions containing disparate cell populations; they even suggest the possibility of targeting the probes with subcellular precision using protein moieties such as the PDZ domain of NOSTIC-1P.

Targeted applications of NOSTICs might be particularly powerful for analyzing the basis of functional selectivity in high level sensory processing, or for characterizing the roles of distinct cell types and projections in neural mechanisms of behavior. Further protein engineering could produce probes that operate with improved signal-to-noise ratio and under conditions where endogenous hemodynamics are suppressed.⁴⁷ This would allow identification of NOSTIC-dependent signals without the subtractive imaging approach of Fig. 3, which in turn would improve the reliability of hemogenetic signal detection, particularly at single-animal and single-voxel levels. Protein engineering might also yield NOSTICs suitable for multiplexed measurements from more than one cell population in the same animal. Variants of our initial probes could be used to study processes such as neuroplasticity, neurochemical signaling, and neurovascular coupling or the biology of NO itself. Hemogenetic technology thus offers a general strategy for mechanistic dissection of neural function on a brain-wide scale.

METHODS

Animal subjects.

Adult male Sprague-Dawley rats (200–250 g, 7–8 weeks old) were purchased from Charles River Laboratories (Wilmington, MA). After arrival, animals were housed and maintained on a 12 hr light/dark cycle and permitted ad libitum access to food and water. All procedures were carried out in strict compliance with National Institutes of Health guidelines, under oversight of the Committee on Animal Care at the Massachusetts Institute of Technology as part of protocol number 0721-059-24.

Statistics and reproducibility.

Experiments were performed on multiple samples or animals to ensure reproducibility and permit group-level comparisons. Relevant statistics are reported throughout the text. Data distributions were assumed to be normal but were not formally tested. All statistical results were evaluated with two-sided significance criteria. No statistical method was used to predetermine sample size. Data from major technical failures (e.g. damaged tissue, poor image quality) were excluded from analysis. Experiments were not randomized. Randomized allocation was not performed, and investigators were not blinded to allocation during experiments and outcome assessment.

Construction of engineered NOS variants.

The plasmid pCWori containing human iNOS and rat nNOS constructs were a generous gift from the lab of Dr. Thomas Poulos at University of California, Irvine. The nNOS and iNOS were individually cloned into the pIRESpuro3 mammalian vector (Takara Bio USA, Mountain View, CA). An inverse polymerase chain reaction (PCR) was performed to create the backbone. Inserts were amplified from the pCWori backbone using overhangs ranging from 24–30 nucleotides on each end. A single-step, isothermal, ligation-free procedure was used to integrate the inserts. The entire length of the coding strand was verified using sequencing. A similar DNA assembly strategy was used to generate chimeric genes encoding NOSTIC-1 and NOSTIC-1P. Sequences of the chimeric proteins are provided in Table S1.

The following strategy was used to build genetic constructs for making stable cell lines: A lentiviral backbone containing the cytomegalovirus (CMV) promoter driving enhanced green fluorescent protein (GFP) with a blasticidin resistance cassette driven by a phosphoglycerate kinase (PGK) promoter (pLV6-eGFP-BLA) was used to generate pLV6-NOSTIC-1-BLA. The parent plasmid was treated with BamHI (NEB, Ipswich, MA) and EcoRI (NEB, Ipswich, MA) to make linearized vector. NOSTIC-1 was amplified using PCR with primers having 20–25 bp overhang. The purified PCR product and the linearized backbone were assembled using Gibson assembly (New England Biolabs, Ipswich, MA). DNA was transformed into Stable competent cells (New England Biolabs). The entire coding strand was verified by Sanger sequencing (Quintara Biosciences, Boston, MA).

Expression and activity of NOS clones in cell culture.

FreeStyle 293-F cells were obtained from Thermo Fisher (Waltham, MA) and maintained according to the supplier’s protocol. To test for expression and activity of NOS constructs expressed in these cells, high-purity DNA suitable for transfection was obtained using the Qiagen Plasmid Maxi Kit (Hilden, Germany). DNA was introduced into the cells using 293Fectin (Thermo Fisher) following the manufacturer’s protocols. 48 hours post-transfection, cells were pelleted and resuspended in Freestyle 293 Expression Medium (Thermo Fisher) containing 1 mM L-arginine and 1 mM CaCl₂ and seeded in six-well plates at two million cells per well. Cells were stimulated with 5 μM A23187 (MilliporeSigma, Burlington, MA). Catalytic activity of NOS constructs was determined by quantifying nitrite from the supernatant using the Griess test (Promega, Madison, WI). Assays were performed in multiwell plates and results read spectroscopically using a Molecular Devices (San Jose, CA) plate reader operated using the manufacturer-supplied SoftMax Pro software. For experiments to test for inhibitor sensitivity of different NOS constructs, N_ω-nitro-L-arginine methyl ester hydrochloride (L-NAME) and 1400W were purchased from MilliporeSigma and added to resuspension medium prior to stimulation of the cells.

293FT cells (Thermo Fisher, Waltham, MA) were used to generate stable lines and were maintained in accordance with manufacturer’s guidelines. Cells were co-transfected with either pLV6-NOSTIC-1-BLA or pLV6-eGFP along with plasmids, psPAX2 and pMD2.G (Addgene, Watertown, MA) using Lipofectamine 3000 (Thermo Fisher, Waltham, MA). 48 hours post-transfection the supernatant containing lentiviral particles was added to freshly seeded cells on a 10-cm tissue culture dish. 24-hours following the addition of the lentiviral supernatant, the media was aspirated and changed with fresh Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS) (Thermo Fisher) containing 10 μg/ml blasticidin. Media was changed every 48 hours until the dish was about 80% confluent.

Catalytic activity of the stably expressing NOSTIC cell line was tested using the Griess test after stimulation with A23187 as described above. To test for expression of NOSTIC, cells either stably expressing eGFP or NOSTIC-1 were seeded on poly-D-lysine coated cover slips (Thermo Fisher, Waltham, MA). When these cells were about 60% confluent, they were fixed with 4% paraformaldehyde (PFA) for 15 minutes, rinsed thrice with phosphate-buffered saline (PBS), permeabilized with buffer containing PBS and 3% TritonX (MilliporeSigma) for 15 minutes, and blocked with buffer containing PBS and 10% donkey serum (MilliporeSigma) for 60 minutes. The cells were then incubated with primary antibody (goat-anti-nNOS, Abcam, Cambridge, MA) at 1:200 dilution at 4 °C. Following overnight incubation, cells were rinsed thrice with PBS and incubated with secondary antibody (donkey anti-goat IgG, tagged with Alexa Fluor 488, Thermo Fisher) at 1:500 dilution for 1 hour. The cover slips were rinsed thrice and mounted on glass slides with ProLong Antifade Mountant (Thermo Fisher). These were then imaged on a confocal microscope (LSM 710, Carl Zeiss, Oberkochen, Germany), controlled with Zeiss ZEN software.

To examine NOSTIC-mediated NO production in central nervous system-derived cells, NOSTIC and control clones were expressed in CAD cells, a subclone of the catecholinergic rat Cath.a neuronal tumor line.⁴⁸ These cells were maintained in 1:1 DMEM:Ham’s F12 nutrient mixture plus 10% FBS (Thermo Fisher). To test for catalytic activity, NOSTIC-1 fused to mCherry (NOSTIC-mCherry) or mCherry alone (control) were transfected using Lipofectamine 3000. The next day, cells were incubated with 5 μM A23187 and 100 μM 1400W (where indicated) in reduced serum Opti-MEM medium (Thermo Fisher). Nitrite was quantified using the Griess test. To obtain micrographs, CAD cells were seeded on poly-D-lysine coverslips and transfected with NOSTIC-mCherry or mCherry. The following day, cells were incubated with 2 μM DAF (4-amino-5-methylamino-2′,7′-difluorescein, diaminofluorescein-FM) in Tyrode solution (VWR, Radnor, PA) and stimulated with 5 μM A23187 in the presence or absence of 100 μM 1400W, as indicated. The cells were fixed in 4% PFA in PBS and washed thrice before imaging.

Brain implantation of NOSTIC-expressing cells.

293FT cells stably expressing NOSTIC-1 or GFP were harvested by trypsinization, pelleted by centrifugation for 10 minutes at 300g, washed with cold PBS (pH 7.4), and resuspended at a density of 100,000 cells/μL in artificial cerebrospinal fluid (aCSF, 125 mM NaCl, 2.5 mM KCl, 1.4mM CaCl₂, 1 mM MgCl₂, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃).

Prior to intracranial cell implantation in rats, each animal was anesthetized with isoflurane (4% for induction, 2% for maintenance) and positioned in a stereotaxic instrument (Kopf Instruments, Tujunga, CA) with a water heating pad (Braintree Scientific, Braintree, MA) to keep the body temperature at 37 °C. The scalp was surgically retracted and two small holes were drilled into the skull above the target sites 0.5 mm posterior and ±3 mm lateral to bregma. 28G polyether ether ketone (PEEK) cannula guides (Plastics One, Roanoke, VA) designed to project 1 mm below the surface of the skull were lowered through the holes. The guides were affixed in place using SEcure light curing dental cement (Parkell, Edgewood, NY). A custom-fabricated plastic headpost was placed in front of the guide cannula and also secured with dental cement (Parkell).

To deliver cells, 33G metal internal cannulae (Plastics One) were lowered through the guide cannulae to a depth of 5.5 mm below the skull. Approximately 500,000 NOSTIC-expressing or control cells were infused via the cannulae at a flow rate of 0.2 μL/min for 25 min, with test and control sites alternating among experiments to negate potential biases. 10 minutes after the infusion the cannulae were slowly removed.

Magnetic resonance imaging of implanted cells in vivo.

After implantation of NOSTIC-expressing or control GFP-expressing cells, the metal cannulae were removed. Animals were intubated and ventilated with a small animal ventilator (Harvard Apparatus) which operated at 62 beats per minute with a 6 mL stroke volume, delivering oxygen and air as a 5:1 ratio mixture. Each animal was then transferred and fixed into a custom rat imaging cradle. A surface receive-only radiofrequency coil (Doty Scientific, Columbia, SC) was positioned over the head. 33G PEEK internal cannulae (Plastics One) were connected to PE-50 tubing (Plastics One) loaded with the calcium ionophore A23187 (10 μM), with or without the NOSTIC-1 inhibitor 1400W (100 μM), and was lowered through the guide cannulae to the same site for cells implantation, at a depth of 5.5 mm below the skull. Isoflurane was discontinued and the animal was sedated with medetomidine (IP bolus of 0.1 mg/kg, followed by 0.1 mg/kg/hr infusion) and paralyzed with pancuronium (IP bolus of 1 mg/kg, followed by 1 mg/kg/hr infusion). Each animal was then transferred to a 9.4 T Bruker (Billerica, MA) MRI scanner for imaging. Animals were warmed using a water-heating pad (Braintree). Heart rate and blood oxygenation saturation level were continuously monitored using an MRI-compatible infrared pulse oximeter (Nonin Medical, Plymouth, MN). Breathing rate and CO₂ level were continuously monitored using MRI-compatible SurgiVet CO₂ monitor (Smiths Medical, Dublin, OH).

All MRI images were obtained using a transmit-only 70 cm inner diameter linear volume coil (Bruker) and a 2 cm diameter receive-only surface coil (Doty) for excitation and detection, respectively. Scanner operation was controlled using the ParaVision 5.1 software (Bruker). Fieldmap shimming was performed to optimize B₀ homogeneity. High-resolution anatomical MRI images were acquired using a T₂-weighted rapid acquisition with refocused echoes (RARE) pulse sequence with RARE factor = 8, bandwidth (BW) = 200 kHz, effective echo time (TE) = 30 ms, repetition time (TR) = 5 s, in-plane field of view (FOV) = 2.56 × 1.28 cm, in-plane resolution of 100 μm × 100 μm, and slice thickness = 1 mm. Functional imaging scan series were acquired using a multi-echo echo-planar imaging (ME-EPI) pulse sequence with TE values = 10, 23, and 36 ms, flip angle (FA) = 90°, TR = 2 s, FOV = 2.56 × 1.28 cm, and in-plane resolution of 400 × 400 μm over five coronal slices with slice thickness = 1 mm.

MRI scans were continuously collected for 2 minutes baseline, 5 minutes during calcium stimulation or control treatment, and 10 minutes resting after the stimulation. Calcium stimulation was produced by infusion of a 10 μM solution of A23187 in aCSF, at a rate of 0.2 μL/min through the preimplanted plastic cannulae. Control treatment involved infusion of an identical solution that also contained 100 μM 1400W. T₂*-dependent contributions to the functional imaging data were extracted from the multi-echo image time series by examining data at three echo times, permitting voxels with strong TE dependence characteristic of BOLD contrast to be identified.⁴⁹ T₂*-weighted signal was defined as as an average of signals observed at each echo time, weighted by TE, while non-T₂*-dependent signal components, largely along the cell implantation tracks, were rejected.

The preprocessed data were then further analyzed in MATLAB (Mathworks, Natick, MA) for visualization and quantification of results. Response maps were computed as mean percent signal change during the expected peak response time point 1–100 s after A23187 stimulation onset minus the average baseline signal, 1–100 s before stimulus onset. Mean response amplitudes and time courses were evaluated over an ROI defined around each cell implantation site. These ROIs were defined by the edges of the implanted xenografts, as judged from the highresolution T₂-weighted anatomical scans, expanded by four voxels in each in-plane direction.

Preparation of herpes simplex viral vectors.

Herpes Simplex Virus-1 (HSV) vectors used in this study were produced by Rachael Neve at the Massachusetts General Hospital Gene Delivery Technology Core. The vectors are replication deficient and capable of retrograde transport from injection sites in rodent brain. NOSTIC-1 expression was directed by an HSV containing the NOSTIC-1 gene in tandem with the gene for mCherry, separated by an internal ribosome entry site (IRES), with expression directed by the human EF1α promoter (HSV-hEF1α -NOSTIC-1-IRES-mCherry). A control vector directed expression of mCherry only from the same promoter (HSV-hEF1α-mCherry).

Virally-delivered NOSTIC-1 was tested for catalytic activity in cell culture. Approximately 200,000 HEK 293FT (Thermo Fisher) cells were seeded on poly-D-lysine and laminin coated coverslips (Thermo Fisher) in a 24-well plate. The following day, 1 μL of 10⁹ pfu/mL HSV delivering NOSTIC-1 or mCherry was added to each well. 24–48 hours later, media was aspirated, washed, and the cells were stimulated with 5 μM A23187 in 400 μL FreeStyle 293 Expression medium (Thermo Fisher) per well, supplemented with 1 mM CaCl_2. Nitrite was quantified from supernatants using the Griess assay following overnight incubation.

Stereotaxic injection of viral vectors.

Animals were prepared for cranial surgery to expose the desired viral injection site at the anterior central caudate-putamen (CPu). Each animal was anesthetized with isoflurane (4% for induction, 2% for maintenance). A small hole was drilled into the skull above the target site, 0.5 mm posterior and 3 mm lateral to bregma. 35G metal internal canulae connected with Hamilton syringes preloaded with each viral vector (NOSTIC-1: 1 × 10⁹ pfu/mL, mCherry: 2 × 10⁹ pfu/mL) were then lowered into the brain and each vector was infused at a rate of 0.1 μL/min at two sites (3 μL per site), 6 mm and 5 mm below the skull surface. The canulae were slowly removed 10 minutes after the viral injection. Bone wax was applied to each injection site to seal it using a sterile cotton applicator tip. Skin incisions were closed by sterile wound clips and lidocaine gel (2%) was applied over the wound areas. Isoflurane was then discontinued and each rat was removed from the stereotaxic frame and placed in warmed cage on a heating pad to recover for 45 min. Slow-release buprenorphine (0.3 mg/kg, MIT Pharmacy) was administered subcutaneously to minimize pain and discomfort. Wound clips were removed 7–10 days after surgery and when the wounds had healed.

Implantation of stimulation electrodes.

Three weeks after viral injection, rats underwent a second surgery to implant electrodes suitable for deep brain stimulation (DBS) of the medial forebrain bundle region of the lateral hypothalamus (LH). Animals were prepared similarly for cranial surgery to expose the desired stimulation site. Bipolar stimulating electrodes were targeted to coordinates 3.4 mm posterior, 1.7 mm medial, and 8.6 mm ventral to bregma. In addition, a plastic headpost was placed to facilitate positioning of each animal during MRI scanning. After all the items were secured on the skull, dental cement was applied to the entire exposed skull surface area to hold them rigidly in place. After the surgery, animals were maintained under a heating pad and treated with 0.3 mg/kg slow-release buprenophine subcutaneously during a convalescence period. Rats were allowed 4–7 days for recovery prior to fMRI experiments.

Functional magnetic resonance imaging.

Preparation of animals for fMRI was similar to that described above for xenograft implantation experiments. After intubation via a tracheotomy and ventilation, each animal was transferred to the MRI cradle and fixed by the headpost. For imaging of rats during LH stimulation, each animal’s LH electrode was connected to a constantcurrent stimulus isolator (World Precision Instruments, Sarasota, FL) controlled from a laptop computer. For imaging of rats during somatosensory stimulation, each animal was implanted with subcutaneous electrodes in the forelimb, which were similarly connected to the stimulus isolator.

General MRI methods, including hardware and anatomical imaging, were similar to those described above for xenograft-implanted animals. Functional imaging of HSV-transfected rats included two minutes of baseline image acquisition followed by five cycles of electrical stimulation, each consisting of 16 s LH stimulation followed by 5 min of rest. Each 16 s stimulation block consisted of eight 1 s trains of 1 ms 0.17 mA current pulses delivered at a frequency of 60 Hz, separated by 1 s between trains. Functional imaging with forepaw stimulation consisted of six blocks of 3 mA 1 ms pulses repeated at 9 Hz for 10 s alternating with 40 s of rest for tests evaluating effects of 1400W and 20 s alternating with 80 s of rest for tests evaluating effects of L-NAME. In each case, the stimulation paradigm was synchronized with the image acquisition using a custom script executed in LabVIEW (National Instruments, Austin, TX). To minimize potential radiofrequency artifacts, the stimulator cable was filtered with a 60 MHz low-pass filter (Mini-Circuits, Brooklyn, NY) before entering the MRI enclosure.

A T₂*-weighted echo planar imaging (EPI) sequence was used for detection of stimulus-induced BOLD contrast, with BW = 200 kHz, a single TE of 16 ms, FA = 90, TR = 2s, FOV = 2.56 × 1.28 cm, in-plane resolution 400 × 400 μm, and 12 slices with slice thickness = 1 mm. In order to discern NOS inhibitor dependence of forepaw stimulus-induced fMRI signals, image series from each animal were acquired first in the absence and then in the presence of systemically delivered 1400W (25 mg/kg IV) or L-NAME (50 mg/kg IV). For hemogenetic experiments, identical procedures were applied using animals infected with NOSTIC-encoding and control viruses.

Analysis of functional imaging data.

Images were reconstructed using the ParaVision 5.1 software and then imported into the National Institute of Health AFNI software package⁵⁰ for further processing. High-resolution anatomical images of each animal were registered to a Waxholm coordinate space rat brain atlas.⁵¹ Functional imaging time series were preprocessed in steps that included slice timing correction, motion correction using a least-squares rigid-body volume registration algorithm, voxelwise intensity normalization, spatial smoothing with Gaussian spatial kernel of 0.5 mm full width at half maximum, and spatial resampling to double the image matrix size, all implemented in AFNI. Segmentation of brain from non-brain voxels was performed in MATLAB. Preprocessed time series were then co-registered onto the previously atlas-aligned anatomical images. Slices at the rostral and caudal extents of the imaging volume had poor image or registration quality and were eliminated from further analysis; slices exhibiting substantial electrode artifacts were likewise omitted. Regions of interest (ROIs) used for subsequent analyses were also defined with respect to the rat atlas and are displayed in Supplementary Fig. 2.

For all functional imaging data, stimulus-dependent impulse response functions (IRFs) were estimated from T₂*-weighted EPI time courses using the 3dDeconvolve general linear modeling implementation in AFNI. This analysis computes IRFs that vary independently across positions in the brain, and thus avoids artifacts that can arise in general linear modeling approaches that employ a single spatially invariant hemodynamic response function. To compute voxel-specific IRFs, LH stimulus time blocks were used as event regressors. Six motion correction parameters from each animal were included as nuisance regressors, along with a linear baseline term. IRFs of 242 s duration (121 image frames) were obtained in units of percent signal change for each voxel.

For hemogenetic experiments, mean percent signal change maps were derived from IRFs computed at the group level, among HSV-NOSTIC (n = 6) or HSV-mCherry (n = 5) animals, separately in the absence and presence of 1400W. Response amplitudes were defined as the average IRF signal over 174 s (87 image frames) beginning at stimulus onset, minus the baseline signal defined by the mean of response at time points 0–40 s (20 frames) before stimulus onset. To correct for systematic changes in hemodynamic amplitudes between pre- and post-1400W conditions, fMRI responses obtained after 1400W treatment were renormalized so that the response amplitude along the midline (assumed to be NOSTIC-independent) matched that observed before 1400W treatment. Difference maps were then calculated by subtracting the mean response amplitude map with 1400W treatment from the corrected mean response amplitude map without 1400W treatment from the same set of animals. Results are presented for voxels that showed significant stimulus responses as defined by F-test results of p ≤ 0.01 from the group regression analyses in the pre-1400W condition.

A second set of 1400W-dependent difference maps was computed using a different general linear modeling approach that assumes a fixed stimulus-associated IRF (i.e. hemodynamic response) for each voxel in each animal. The CPu IRF computed from the analysis described above was used for this purpose. Input to the regression analysis consisted of group data from all animals imaged in both pre- and post-1400W experimental periods, normalized by midline responses as above, to account for signal drift. These data were subjected to a standard regression analysis that included regressors to account for intrinsic BOLD signal, NOSTIC-specific BOLD signal, and six motion correction parameters as nuisance regressors, along with a linear baseline term. The intrinsic BOLD regressor was computed by convolving all stimulus blocks with the CPu IRF, while the NOSTIC regressor was computed by convolving pre-1400W stimulus blocks only with the CPu IRF. Maps of regression coefficients (β values) for the NOSTIC-specific BOLD signal are presented in Supplementary Fig. 5.

ROI-specific response values were obtained by computing IRF amplitudes described above separately for every animal, thus accounting for inter-animal variability, and then averaging each animal’s IRFs over the corresponding anatomical regions defined in Supplementary Fig. 2. Similarly, ROI-specific time courses were obtained by averaging signal time courses over anatomically defined regions and then over animals. All voxels from each ROI were included in the calculations, independent of statistical thresholds. Statistical analysis of 1400W effects was performed using paired t-tests of differences in each ROI before vs. after drug treatment. Brain regions contralateral to the LH stimulation electrode were not analyzed in detail, although some 1400W-dependent differences and HSV-driven NOSTIC expression were observed in contralateral CPu and MCx, in particular. A dorsoventral region of the brain also displayed substantial 1400W-dependent modulation in one animal only and contributes to signal in the NOSTIC-1 difference map of Fig. 3c; this difference signal is not statistically significant (paired t-test p = 0.14, n = 6). Unless otherwise noted, error bars and error margins associated with ROI-specific results throughout the text denote SEM over multiple animals.

For forepaw stimulation experiments of Fig. 1 and Extended Data Fig. 2, mean response amplitudes were determined as the average IRF signal over 6 s (3 image frames) beginning after stimulus onset, minus the pre-stimulation baseline average. Difference maps of Fig. 1e were computed by subtracting the mean response map with 1400W or L-NAME treatment from the map without treatment from the same animals, and statistical evaluation of ROI-level effects was performed using paired t-tests of differences in the S1FL region before vs. after drug treatment. For resting state connectivity analysis presented in Fig. 5 and Extended Data Fig. 10, mean signal time courses for each ROI were calculated from EPI image series acquired from four naïve rats in the absence of stimulation. The Pearson correlation coefficients and the corresponding Fisher-transformed Z values were computed between the time courses of the viral injection subregion of CPu (bregma Ȓ0.5 mm) and the eight ROIs. Student’s t-tests were used to evaluate statistical significance of Z values and associated comparisons.

Photometry recordings.

Six rats were injected with LT-HSV-GCaMP6s or LT-HSV-GFP viral vectors into the CPu, using methods analogous to injection of LT-HSV-NOSTIC-1 for hemogenetic fMRI experiments. Three weeks after viral injection, the animals were anesthetized with isoflurane (4% for induction, 2% for maintenance), fixed into a stereotaxic frame, and implanted with an LH stimulation electrode. A multimodal optic fiber (200 μm diameter, Thorlabs, Newton, NJ) was subsequently inserted into a craniotomy above SN (4.8 mm posterior, 2.0–2.5 mm lateral and 7.0 mm ventral to bregma) and held in place by the stereotaxic arm.

Fluorescent signals were recorded at different depths ranging from 7.0–8.5 mm ventral to bregma by lowering the fiber in 0.5 mm steps for each recording during LH stimulation delivered at 60 Hz for 1 s blocks every 10 s at 0.15 mA current. A commercially available setup (FOM-II, NPI Electronic GmbH, Tamm, Germany) delivered light for excitation of the calcium indicator via a 650 mW LED with a nominal peak wavelength of 470 nm. LED power was controlled by an adjustable current source. The light beam was focused using a fixed-focus collimator onto one end of the optic fiber, which was connected to the system using an SMA connector. An aspheric lens was used to guide emitted fluorescence back from the same fiber, where it was focused onto a silicon photomultiplier photodetector with an active area of 3 × 3 mm² and a photon detection efficiency of 20 % at 490 nm. The recorded fluorescence signals were low-pass filtered at 50 Hz, amplified with a gain of 50 (LPBF-01GD, NPI Electronic), and digitized with a sampling frequency of 1 kHz using a multifunction data acquisition interface (Power1401, Cambridge Electronic Design, Cambridge, UK) and its corresponding acquisition software (Spike2). Fluorescence signals were analyzed in MATLAB. All traces represent relative changes in fluorescence intensity (ΔF/F) and were low-pass filtered at 10 Hz and smoothed using a running average of 0.5 s before being averaged over 40 stimulation trials each.

Brain clearing and imaging.

Three weeks after the HSV-mCherry virus injection, rats were transcardially perfused with 200 mL of ice-cold phosphate-buffered saline containing 0.02% sodium azide (PBSN), followed by 200 mL of ice-cold SHIELD perfusion solution at a flow rate of 60 mL/min.⁵² The brain was extracted and incubated in SHIELD perfusion solution at 4 °C on an orbital shaker for 2 days. The brain was then moved to 50 mL of SHIELD OFF solution and incubated on an orbital shaker for 3 days at 4°C. After SHIELD OFF incubation, the brain was incubated in 40 mL of SHIELD ON solution that had been pre-warmed at 37 °C for 20 minutes. SHIELD ON incubation occurred on an orbital shaker at 37 °C for 1 day. Following the completion of SHIELD processing, the brain was washed in PBSN for at least 1 day. SHIELD processing solutions were obtained from LifeCanvas Technologies (Cambridge, MA).

The brain was hemisected after washing. The meninges surrounding the brain were then stained using a solution of 1% toluidine blue in 70% ethanol, diluted 10-fold in PBSN. The brain was briefly immersed in the staining solution and then placed under a stereoscope equipped with a fiber-optical illuminator. A blunt-tipped paint brush was used to separate the dura layer from the surface of the brain. Hemispheres were then passively incubated for 48 hours at room-temperature in a tissue clearing buffer containing sodium dodecyl sulfate. After incubation, intact hemispheres were rapidly delipidated via stochastic electrotransport at 45 °C for 10–14 days.⁵³ Optical clearing proceeded until the core of each sample showed no signs of opacity. Samples were then rinsed in PBSN three times for a minimum of 12 hours each. Prior to volumetric labeling, samples were passively incubated for 48 hours at room temperature in eFLASH sample buffer.⁵⁴ Rapid 24 hour immunostaining of whole hemispheres with polyclonal CF640R rabbit anti-RFP (Biotium, Fremont, CA) was carried out using the eFLASH protocol.⁵⁴ The samples were again passively washed in PBSN three times for a minimum of 12 hours each before refractive index matching.

After clearing and labelling, the intact rat hemisphere was optically cleared using a Protosbased immersion medium. The tissue was incubated for two days at 37 °C until tissue was transparent. Once optically cleared, the tissue was mounted in an agarose block (1.5% wt/vol agarose powder dissolved in immersion medium) and volumetrically imaged with an axially-swept lightsheet microscope (LifeCanvas Technologies) using a custom 3.6x magnification, 0.2 numerical aperture detection objective with a uniform axial resolution of approximately 4 μm. This system is equipped with an Oxxius (Lannion, France) L4Cc laser combiner unit equipped with 488 nm, 561 nm, 642 nm, and 785nm diodes; anti-RFP images were acquired using the 642 nm unit. The 900 GB raw light sheet dataset was stitched into a single 92 GB Imaris (Bitplane, Zürich, Switzerland) dataset for visualization.

Conventional immunohistochemical analysis.

For immunohistochemistry analysis of activity-dependent gene induction in conjunction with other markers, rats injected with HSV-hEF1a-NOSTIC-1-IRES-mCherry were anesthetized via an intraperitoneal injection of a ketamine (90 mg/kg) and xylazine (10 mg/kg) mixture. Thirty minutes after the induction of anesthesia, the rats were subjected to a sequence of five trials of LH stimulation using the same stimulation parameters as in the fMRI experiments, with 1 min rests between trials. Rats were then transferred back to their own cage for 90 minutes to allow for c-Fos expression and accumulation before the animals were sacrificed by lethal intraperitoneal injection of a sodium pentobarbital (0.5 mg/g Fatal-Plus), in conjunction with transcardial perfusion with 4% paraformaldehyde (PFA).

Brains were harvested and fixed in PFA for an additional 48 hours at 4 °C. Extracted brains were then sectioned into 50 μm-thick free-floating coronal slices using a semi-automatic vibratome (VT1200, Leica Biosystems, Wetzlar, Germany). Slices were stored in PBS at 4 °C. Prior to staining, slices were washed thrice in PBS + 0.1% TritonX (PBST) and blocked for 1 hour with 10% v/v donkey serum (MilliporeSigma). Each slice was then incubated overnight at 4 °C with gentle shaking in PBST with 1% serum and primary antibody. The following primary antibodies were used: goat anti-mCherry (1:200, Lifespan Biosciences, Seattle, WA), rabbit anti-mCherry (1:200, Abcam), rabbit anti-c-Fos (1:500, Synaptic Systems, Goettingen, Germany), goat anti-IBA-1 (1:500, Abcam), rabbit anti-nitrotyrosine (1:200, MilliporeSigma).

The following day, tissue was washed thrice with PBST and then probed with secondary antibody (Thermo Fisher or Abcam). The following secondary antibodies were used: donkey anti-rabbit IgG tagged with Alexa Fluor 594, donkey anti-goat IgG tagged with Alexa Fluor 488, donkey anti-goat IgG tagged with Alexa Fluor 594, or donkey anti-rabbit IgG tagged with Alexa Fluor 488, all applied in PBST with 1% serum for one hour. Slices were then washed thrice in PBS, mounted on glass slides using Prolong Antifade Mountant with DAPI (Thermo Fisher) and imaged on a confocal microscope (LSM 710, Carl Zeiss, Oberkochen, Germany).

To perform quantitative analysis of immediate early gene activation, ImageJ (National Institutes of Health, Bethesda, MD) was used to quantify the total number of c-Fos- or mCherry-expressing cells by counting all cells exhibiting homogenous fluorescence at noticeably higher levels than the background fluorescence in the images, which was kept constant throughout the counting process for all images. Counting regions were defined by overlaying brain atlas sections⁵⁵ on the histological images based on landmarks for the relevant brain areas, and cells were distinguished by both color and distribution of fluorescence. Cells expressing both c-Fos and mCherry markers were identified by overlap of morphological features in the green and red channels of each image.

Code availability.

Processing scripts used in the data analysis are available from the corresponding author upon request.

Data availability.

Source data for Figures 1–5 and associated Extended Data Figures are provided with this submission.

Extended Data

Extended Data Fig. 1. NOSTIC functionality in a neuronal cell line.

a, NOSTIC-1 fused to mCherry (cyan), as well as mCherry alone (gray), were expressed in CAD cells and stimulated with A23187 in the absence (−) or presence (+) of 100 μM 1400W; nitrite formation following NO release from the stimulated cells was measured for each condition. The effect of 1400W inhibition is significant for NOSTIC-expressing cells (paired t-test p = 0.013, n = 4) but not for control cells (p = 0.59, n = 4). b, NO production by A23187-stimulated CAD cells was visualized by staining with the fluorescent NO probe diaminofluorescein (DAF). NOSTIC-expressing cells indicated by red mCherry staining show higher levels of green DAF staining in the absence (top) than in the presence (bottom) of NOSTIC inhibition with 1400W. Scale bar = 200 μm. This experiment was performed twice.

Extended Data Fig. 2. Time courses of fMRI responses to forepaw stimulation.

Mean fMRI signal time courses (% signal change) observed in the S1 forelimb region in response to stimulation of the contralateral paw (pink shaded region) in the presence vs. absence of L-NAME (left) or 1400W (right), corresponding to maps and mean response amplitudes reported in Fig. 1e. For arbitrary reasons, experiments testing effects of L-NAME and 1400W were performed with stimulation blocks of 20 s and 10 s, respectively, and differences in the response durations observed left vs. right arise from this fact.

Extended Data Fig. 3. Spatial extent of xenograft-induced hemogenetic responses.

a, Left: Anatomical MRI showing T₂-weighted (T₂w) contrast in the neighborhood of cell implantation for a represenative animal from the experiments of main text Fig. 2. Right: Average BOLD contrast map (color) induced by A23187 stimulation in units of percent signal change (%SC), overlayed on a T₂-weighted scan. b, Cross sections of the contrast patterns through the dotted line in panel a, with solid lines denoting means and shading denoting SEM across six animals’ BOLD (red) and T₂w (black) signal changes. Full width at half maximum is 1.4 ± 0.1 mm for the hemogenetic BOLD responses, indicating point-spread function comparable to or less than conventional hemodynamic fMRI.

Extended Data Fig. 4. Evaluation of HSV vectors in cell culture.

a, Western blotting was used to analyze cells transfected with GFP or NOSTIC-1 (N-1), along with cells infected with HSV encoding mCherry (mC) or NOSTIC-1-IRES-mCherry (N-1/mC). Expression and size of NOSTIC-1 was confirmed using blotting with an anti-FLAG tag (α-FLAG) antibody (top). Expression and size of mCherry in the HSV infected cells was assessed using an anti-mCherry antibody (α-mCherry, middle), and anti-glyceraldehyde 3-phosphate dehydrogenase (α-GAPDH) was used as a loading control (bottom). Relevant molecular weight (MW) markers labeled at left. Expected sizes: NOSTIC-1 140 kD, mCherry 29 kD, GAPDH 37 kD. b, Griess test measurement of nitrite production in the presence of cells transfected with GFP or NOSTIC-1 or virally transduced with mCherry or NOSTIC-1, in the presence or absence of A23187 stimulation, as indicated. Only stimulated cells expressing NOSTIC-1 show strong evidence of stimulus-dependent NO production, as indicated by the Griess test results.

Extended Data Fig. 5. LH stimulus responses in NOSTIC-expressing rats after 1400W treatment.

Average fMRI responses to LH stimulation in the +1400W condition, among 6 animals infected with NOSTIC-encoding HSV, analogous to −1400W condition data in main text Fig. 3b. Significant responses with F-test p ≤ 0.01 shown.

Extended Data Fig. 6. Time courses of responses to LH stimulation in multiple ROIs.

Average time courses of fMRI signals observed in eight ROIs in the presence and absence of 1400W, analogous to data of Fig. 3d. Pink boxes denotes stimulation period. Shading denotes SEM over six animals each.

Extended Data Fig. 7. LH stimulus responses in rats treated with control HSV vectors.

Average fMRI responses to LH stimulation in the −1400W (left) and +1400W (right) conditions, among 5 animals infected with mCherry-encoding HSV, corresponding to difference maps shown in main text Fig. 3C. Significant responses with F-test p ≤ 0.01 shown.

Extended Data Fig. 8. 1400W-dependent fMRI amplitudes observed in multiple ROIs.

Absolute fMRI response amplitudes observed in each of eight ROIs examined in NOSTIC and control mCherry HSV-treated animals, analogous to data presented in Fig. 3f. 1400W-dependent effects in CPu, ECx, MCx, and SN are all significant with paired t-test p ≤ 0.05 (n = 6).

Extended Data Fig. 9. Relative changes in fMRI amplitudes upon 1400W treatment.

ROI-averaged fMRI amplitudes from the experiments of Fig. 3 were used to compute relative signal differences as 100 × (pre – post)/pre, where pre and post are the fMRI amplitudes observed before and after 1400W application in each rat. Results are shown as box plots for NOSTIC-1-treated (left, n = 6) and control mCherry-treated (right, n = 5) animals. Box plots denote median (center line), first quartiles (boxes), and full range (whiskers) over animals. NOSTIC results are significant in CPu, ECx, MCx, and SN (t-test p ≤ 0.008); control results are significant in CCx (t-test p = 0.037).

Extended Data Fig. 10. Resting-state functional connectivity to CPu.

A map of Z-transformed correlation coefficients relating resting state fMRI signal in a striatal seed region to voxels across the rest of the brain.

Abbreviations:

BOLD

Blood Oxygen Level-Dependent

CPu

Caudate-Putamen

CCx

Cingulate Cortex

DBS

Deep Brain Stimulation

ECx

Entorhinal Cortex

fMRI

Functional Magnetic Resonance Imaging

GFP

Green Fluorescent Protein

HSV

Herpes Simplex Virus

iNOS

Inducible Nitric Oxide Synthase

ICx

Insular Cortex

LH

Lateral Hypothalamus

L-NAME

L-nitroarginine methyl ester

MRI

Magnetic Resonance Imaging

MT

Medial Thalamus

MCx

Motor Cortex

nNOS

Neuronal Nitric Oxide Synthase

NOSTIC

Nitric Oxide Synthase for Targeting Image Contrast

S1

Primary Somatosensory Cortex

SN

Substantia Nigra