Noninvasive focal gene transfer of chemogenetic proteins in the primate brain

Summary

We leverage our recent development of transcranial focused ultrasound (tFUS) for noninvasive, focal delivery of adeno-associated viruses (AAVs)—encoding excitatory designer receptors exclusively activated by designer drugs (DREADDs)—to marmoset frontal cortex. Using FDG positron emission tomography, we demonstrate significant increases in glucose metabolism at the site of viral delivery after administering the DREADD-specific agonist deschloroclozapine (DCZ), as compared to vehicle control. Focal neuronal DREADD expression was confirmed by immunohistochemistry at the site of opening. Through a comparison of awake resting-state functional connectivity and structural connectivity from the sites of delivery, we demonstrate that increases in glucose metabolism occur within both mono- and polysynaptically connected brain regions. Taken together, these results demonstrate the ability to focally deliver excitatory chemogenetics without the need for surgery, allowing for activation of long-range frontal cortex circuits of the primate brain.

Graphical abstract

Highlights

• •We demonstrate focal disruption of marmoset BBB using transcranial focused ultrasound
• •Systemically delivered AAV vectors encoding excitatory DREADDs cross the BBB
• •DREADD expression is constrained to region of BBB disruption
• •Noninvasively transduced DREADDs alter FDG uptake locally and in connected areas

Motivation

The development of chemogenetics, including designer receptors exclusively activated by designer drugs (DREADDs), has enabled long-lasting and reversible neuromodulation in the primate brain. Although systemically delivered synthetic ligands allow for noninvasive actuation of chemogenetics, intraparenchymal injection remains atop the available methods to precisely deliver chemogenetic payloads within the brain. When considering therapeutic injections, however, the requirement of transcranial surgery does not translate well to the clinic. Here, we sought to use transcranial focused ultrasound-mediated viral gene delivery to noninvasively transduce functional DREADDs within the primate brain—ameliorating the need for neurosurgical intervention to employ this foundational method of circuit modulation.

The most common method of viral delivery to the brain is intracranial injection, which translates poorly to the clinic. Corigliano et al. establish a platform for noninvasive gene therapy administration by utilizing transcranial focused ultrasound-mediated viral gene delivery to noninvasively express functional excitatory chemogenetics within marmoset frontal cortex.

Introduction

Viral-based gene therapies, which leverage the endemic infectivity of a virus to deliver genetic material to cells, hold great promise for the treatment of many otherwise intractable diseases and disorders (see Kohn et al.¹ for review). Several viral gene therapies have now received US Food and Drug Administration (FDA) approval¹ with many others in development.² Although neurological diseases/disorders are being targeted heavily for the development of new gene therapies,² these treatments will invariably require a method to bypass the blood-brain barrier (BBB)—a selectively restrictive structure regulating molecular access to the central nervous system (CNS).^3,4 Generalizable methods facilitating safe penetrance of therapeutics across the BBB, however, remain elusive.³ Currently, trephination and microinjection are atop the available methods for precise viral vector delivery to the CNS, but present notable health risks including infection and permanent tissue damage.⁵ While viral injection does yield robust transduction of neural tissue,⁶ these potential health complications prevent translation to the clinic at scale.⁴ Newly designed adeno-associated virus (AAV) variants possessing the ability to cross the BBB after intravenous (IV) injection^{7,8,9,10,11,12,13} are an attractive option for delivering genetic material to the CNS, but these constructs are unable to, without additional factors, transduce cells/regions of the brain with specificity—a requirement to avoid mixed or antagonistic effects in potential gene therapies. An emerging method, transcranial focused ultrasound (tFUS), has been posited as a means to circumvent brain surgery altogether, allowing for the delivery of therapeutics to highly focal brain targets by transiently perturbing the BBB.^{5,14,15,16,17,18,19,20,21}

Through the confluence of highly focused sound waves and intravenously injected microbubbles (lipid spheres that are mechanically agitated by sound), the tight endothelial junctions of the BBB can be transiently opened, allowing for the passage of therapeutics that are otherwise minimally effective at crossing the BBB.^4,22 Since its inception by Hynynen et al. (2001),²² microbubble-assisted tFUS BBB opening has been successfully translated across the phylogenetic spectrum (see the following for demonstrations of tFUS BBB disruption in rabbits,²² pigs,²³ mice,²⁴ rats,²⁵ macaques,^20,26,27 and humans^28,29,30,31). We have recently ported the tFUS technique for use in the marmoset monkey (Callithrix jacchus), a small new-world primate species that has great promise for the future of neuroscientific modeling of human brain diseases.³² Despite being roughly the size of a rat (∼350 g as adults), marmosets^21,33 demonstrate a markedly similar functional brain organization to that of humans,^34,35,36 and the recent advent of transgenic marmosets^37,38,39 presents myriad opportunities to test the feasibility of tFUS-mediated genetic technologies in nonhuman primate models of disease. To this end, we expanded on our method of BBB perturbation in the marmoset to include the delivery of AAVs carrying fluorescent reporters to marmoset frontal cortex²¹—a concept instantiated via foundational work in rodents^{14,15,16,17,18} and recent work in macaques.²⁰ By incorporating systemic AAV injection into our primate-specific tFUS BBB disruption method, we demonstrated focal AAV transduction within the target of tFUS application (also termed “insonicated site”) and were able to visualize coursing axons from transduced neurons.²¹ Beyond the delivery of fluorescent reporters, tFUS-mediated viral gene delivery has been used in recent reports to reinstate dopamine transmission in a mouse model of Parkinson’s disease,¹⁹ suggesting the promise of this technique for the transduction of a variety of translationally relevant neuromodulators. Of these, designer receptors exclusively activated by designer drugs (DREADDs)⁴⁰ are promising candidates for noninvasive, reversible circuit modulation.

DREADDs engineered from human muscarinic receptors have been used extensively in the neurosciences since their introduction in the mid 2000s⁴⁰ (see Roth⁴¹ for review). Upon selective activation through administration of their synthetic ligand, these mutant receptors transiently depolarize (Gq)⁴² or hyperpolarize (Gi)⁴³ neurons with concomitant effects on behavior (see Roth⁴¹ and Rogan and Roth⁴⁴ for reviews). DREADDs have been deployed to study a variety of processes in the macaque (see Raper and Galvan⁴⁵ for a review of nonhuman primate chemogenetic studies). Indeed, DREADD actuation has been shown to alter macaque working memory,^46,47,48,49 decision-making,⁴⁸ subjective valuation of reward,^50,51 amygdalar functional connectivity,⁵² emotional reactivity,^53,54,55 socioemotional attention,⁵⁴ false belief attribution,⁵⁶ cutaneous sensation,⁵⁷ and inhibitory control⁵⁸ (see Raper and Galvan⁴⁵ for review). Recently, the efficacy of excitatory DREADDs has also been established in marmoset nonhuman primates via modulation of reward anticipation,⁵⁹ emotional reactivity,⁵⁹ and contralateral rotation.⁶⁰ Beyond circuit dissection for causal analyses of behavior, the use of DREADDs in nonhuman primates is clinically advantageous as DREADD ligands can be radiolabeled, allowing for in vivo assessments of protein expression using positron emission tomography (PET).^{46,47,49,51,53,55,57,60,61,62} Additionally, the extended timescale at which DREADDs are effective (>2 h⁴⁷) allows for noninvasive, circuit-wide indexing of construct function via neuroimaging modalities like [¹⁸F]-fluorodeoxyglucose (FDG) PET (90-min acquisition)—a method providing quantitative measurements of glucose metabolism as a proxy for neuronal activity.⁶³ Moreover, DREADD actuators readily cross the BBB,⁴⁷ reducing the need for repetitive intracranial injections to temporarily excite or inhibit cells of tissues.

Although DREADD ligands can be administered using minimally invasive procedures,^46,47 intracortical injection of viral vectors is currently the preeminent method of focal DREADD delivery to the nonhuman primate brain.⁴⁵ Recently, however, tFUS BBB disruption was used to deliver systemically injected AAVs coding for either inhibitory or excitatory DREADD receptors to the murine brain.⁵ Noninvasively delivered DREADD constructs were demonstrated functional after actuation via increased c-FOS signal (hM3Dq) or inhibition of contextual memory formation (hM4Di).⁵ While tFUS BBB disruption has been effective for noninvasive delivery of viral vectors in marmosets²¹ and macaques,²⁰ this delivery platform has not yet been used for noninvasive chemogenetic delivery to the primate brain.

Here, we demonstrate focal, noninvasive excitatory G(q)-DREADD expression in marmoset frontal cortex via tFUS-mediated AAV delivery. Three marmosets received low intensity focused ultrasound to either A8aD (two marmosets) or A6DR (one marmoset), transiently perturbing the BBB and allowing for entry of systemically administered AAVs (AAV5, two marmosets; AAV9, one marmoset) carrying hM3DG(q) to the insonicated neural tissue. After allowing for DREADD expression at the site of BBB opening, we used FDG PET to index changes in glucose metabolism at the site of viral delivery and connected circuitry—a proxy for neuronal activity levels⁶³—in the presence of a systemically injected DREADD agonist, deschloroclozapine (DCZ).⁴⁷ As compared to vehicle control, DCZ injections resulted in significant increases in glucose metabolism at the site of DREADD delivery. Immunohistochemical staining revealed robust neuronal DREADD expression localized to the area of BBB perturbation. Through comparison of awake resting-state functional connectivity (whole-brain connectivity with the sites of delivery)⁶⁴ and structural connectivity (viral tracing of anterograde neuronal projections, including axons of passage, from sites of delivery),^65,66 we demonstrate that increases in glucose metabolism also occur within regions monosynaptically and polysynaptically connected to the insonicated region only after DREADD actuation. Taken together, these results demonstrate the ability to focally deliver excitatory chemogenetics—without the need for surgery—allowing for activation of long-range frontal cortex circuits of the primate brain.

Results

MRI-based verification of noninvasive BBB disruption for viral administration

Immediately following a single insonication (aided by microbubble cavitation; see “STAR Methods”) to either A8aD (marmosets Pa and V) or A6DR (marmoset S), each animal was transferred to a 9.4 T MRI and injected with a gadolinium-based contrast agent (GBCA).⁶⁴ The GBCA, in combination with a magnetization-prepared rapid gradient echo (MPRAGE) sequence optimized for the decreased T1 relaxation times in neural tissue exposed to gadolinium, allowed for real-time evaluation of BBB disruption. As demonstrated in our previous work, focal hyperintensities arising from rapid gadolinium extravasation mark the site of insonication.^21,33 Focal regions of BBB disruption—i.e., GBCA extravasation within the insonication target—were detected in the gray matter of each animal after a single insonication (Figure 1). The extent of BBB disruption is consistent with the expected peak of focus of the 1.46 MHz transducer (full width at half maximum of the acoustic pressure distribution) used in these experiments, and between animals for whom the insonication target is similar.^21,33 For the exact insonication parameters, see the “STAR Methods” below. Immediately after verification of focal BBB disruption (Figure 1), each animal was injected (through a catheter placed in the lateral tail vein) with an AAV carrying an excitatory DREADD under the synapsin promoter.

Figure 1.

Atlas-based targeting of areas 8aD and 6DR for tFUS and the resultant areas of BBB disruption in all marmosets

(A) Regions targeted for BBB disruption for all marmosets. Each ellipse (white) represents the theoretical volume of opening for the 1.46 MHz transducer used in this study. Cytoarchitectonic boundaries are based on the Paxinos atlas⁶⁷ in Marmoset Brain Mapping template space.^64,68 Scale bar, 1 mm.

(B) Regions of BBB disruption for all marmosets visualized via a magnetization-prepared rapid gradient echo (MPRAGE) MR image. The sites of BBB opening are indicated by focal hyperintensities resulting from gadolinium extravasation into the surrounding tissue. Abbreviations are as follows: R, right; L, left; S, superior; I, inferior; A6M, area 6 of cortex, medial (supplementary motor) part; A6DR, area 6 of cortex, dorsorostral part; A6DC, area 6 of cortex, dorsocaudal part; A6Va, area 6 of cortex, ventral, part a; A6Vb, area 6 of cortex, ventral, part b; A8b, area 8b of cortex; A8C, area 8 of cortex, caudal part; A8aD, area 8a of cortex, dorsal part; A8aV, area 8a of cortex, ventral part; A9, area 9 of cortex; A11, area 11 of cortex; A13b, area 13b of cortex; A13L, area 13 of cortex, lateral part; A13M, area 13 of cortex, medial part; A14R, area 14 of cortex, rostral part; A24a, area 24a of cortex; A24b, area 24b of cortex; A24c, area 24c of cortex; A25, area 25 of cortex; A32, area 32 of cortex; A32V, area 32 of cortex, ventral part; A45, area 45 of cortex; A46D, area 46 of cortex, dorsal part; A46V, area 46 of cortex, ventral part; A47L, area 47 of cortex, lateral part; A47M, area 47 of cortex, medial part; A47O, area 47 of cortex, orbital part; ProM, proisocortical motor region (precentral opercular cortex); OPro, orbital proisocortex; OPAl, orbital periallocortex.

Local chemogenetic neuromodulation via noninvasive AAV transduction

Please see the table in the STAR Methods for the age of each animal, the viral construct administered after tFUS BBB disruption, and the time allowed for viral construct expression before PET imaging. After allowing for a minimum of 2 weeks for DREADD expression in the insonicated brain region, marmosets underwent anesthetized FDG PET imaging either in the presence of DCZ—a potent DREADD agonist with minimal off-target effects on behavior^46,47—or a vehicle control (2.5% dimethyl sulfoxide in sterile saline). Compared to vehicle control, we observed significant increases in FDG uptake within the insonicated brain region after systemic injection of DCZ (Figure 2) as indicated by Wilcoxon signed rank tests corrected for multiple comparisons (marmoset V, insonicated region adjusted p < 0.001; marmoset Pa, insonicated region adjusted p < 0.001; marmoset S, insonicated region adjusted p < 0.001; and group, insonicated region adjusted p < 0.001). Increases in FDG uptake within the insonicated region after DCZ administration were consistent across animals and AAV serotypes used in this study, suggesting the utility of both AAV9 and AAV5 for tFUS-mediated viral gene delivery. Significant increases in FDG uptake were also found in corresponding regions of cortex in the contralateral hemisphere after DCZ administration in all animals (Figure 2; marmoset V, contralateral region adjusted p < 0.001; marmoset Pa, contralateral region adjusted p < 0.001; marmoset S, contralateral region adjusted p < 0.001), as detailed in sections below.

Figure 2.

DREADD actuation increases [¹⁸F]-FDG uptake within the region of BBB disruption and in homotopic cortex

(A) Row 1 contains MR images of BBB disruption in each marmoset. White-dotted lines indicate regions of BBB disruption, which are identified through focal hyperintensities resulting from gadolinium extravasation into the surrounding tissue. Rows 2 and 3 display weight-normalized FDG uptake after administration of either a vehicle control solution (row 2) or DCZ (row 3). Warmer colors represent peaks of FDG uptake and are present in both insonicated and homotopic regions after DREADD actuation. Abbreviations are as follows: R, right; L, left; S, superior; I, inferior.

(B) Bar graphs showing weight-normalized FDG uptake within both insonicated and homotopic cortex (mean ± standard deviation) after administration of either a vehicle control (blue) or DCZ (red). For all animals, Wilcoxon-signed rank tests indicated significant increases in FDG uptake both ipsilateral and contralateral to the insonication site after administration of DCZ as compared to vehicle control (∗∗∗∗p < 0.001). Statistical tests utilized ROI pixel values collected for each scan and include one scan per condition for each animal.

Immunohistochemical validation of DREADD expression

To verify DREADD transduction within the region of BBB disruption, we conducted immunohistochemical analysis on tissue from marmoset V, who received a construct with the mCherry fluorescent reporter. As seen previously with the tFUS-mediated viral gene delivery method,²¹ we found robust mCherry expression within the site of insonication (Figure 3). Within this 50 μm tissue section, we observed mCherry fluorescence in 24 cells with clear neuronal morphology covering over 200 μm² of cortical gray matter. Note that mCherry expression was constrained to the site of insonication (Figure 3; see Figure 1 for the extent of BBB disruption in reference to the transducer focus) and not observed in the contralateral hemisphere, which also showed significant increases in FDG uptake after DCZ injection (Figure 2).

Figure 3.

Focal DREADD expression within neurons at the insonicated site

(A) MR image of marmoset V at the location of BBB disruption in area 8aD (left).

(B) 50-μm tissue section (marmoset V, 20x magnification) at the coronal level of the MR image above. In both the MR image and coronal tissue slice, cyan and magenta boxes demark the locations of images (E, F) and (C, D), respectively. Scale bar, 2,000 μm.

(C and D) Zoomed in view of corresponding tissue region in (B) showing positive DAPI signal (C) but no mCherry expression (D) in cortex contralateral to the site of BBB disruption. Scale bar, 100 μm.

(E and F) Zoomed in view of corresponding tissue region in image (B) showing positive DAPI signal (E) and mCherry signal (F) within the insonicated cortex. Abbreviations are as follows: R, right; L, left; S, superior; I, inferior.

Comparison of functional and structural circuits

In response to our finding of increased FDG uptake in the non-insonicated hemisphere after DREADD actuation (Figure 2) without concomitant hM3D(Gq) expression in this same locus (see Figure 3D), we examined the population functional⁶⁴ and individual long-range structural connectivity^65,66 of insonicated regions via open access marmoset resources. As an index of structural connectivity, data from anterograde neuronal tracer injections included in the Brain/MINDS Marmoset PFC Connectome^65,66 were chosen for analysis based on their proximity to the insonication targets presented in this study (marmoset V, A8aD, injection R01-0054; marmoset Pa, A8aD, injection R01-0059; and marmoset S, A6DR, injection R04-0079). Voxel-wise functional connectivity profiles of each insonication point were downloaded from our Marmoset Brain Connectome, a resting-state functional magnetic resonance imaging (RSfMRI) resource generated with data from 32 marmosets.⁶⁴ To demonstrate DCZ-related activation at corresponding sites in the contralateral cortex, Figure 4 compares these (injection^65,66 and RSfMRI⁶⁴) data to FDG uptake after DREADD actuation in a coronal plane at each animal’s site of insonication. Given the focal transduction of AAVs within the insonicated site (see Figure 3 and previous work²¹), we expected areas of increased radiotracer uptake outside of the insonication target to be functionally connected with and/or contain anterograde projections from the target itself. Indeed, Figure 4 demonstrates consistent overlap between foci of increased glucose utilization, functional connectivity peaks, and anterograde neuronal projections for each insonication location. Overlap between FDG peaks, functional connectivity peaks and anterograde projections is most apparent in contralateral regions of cortex corresponding to insonication targets but is also found in other frontal areas including 8b and A6M (Figure 4).

Figure 4.

Comparison of anterograde tracing, RSfMRI, and [¹⁸F]-FDG PET data for each insonicated region

(A) Anterograde tracer data collected from the Brain/MINDS Marmoset PFC Connectome^65,66 demonstrating structural connectivity from injections closest to insonication targets (marmoset V, injection R01-0054; marmoset Pa, injection R01-0059; marmoset S, injection R04-0079).

(B) Resting-state functional connectivity maps collected from the Marmoset Brain Connectome (Z score of 3.28 corresponds to p = 0.001).⁶⁴ Single voxel seeds were chosen based on the location of BBB disruption. Labels in rows 1 and 2 are derived from the Paxinos atlas parcellation.⁶⁷

(C) Weight-normalized FDG uptake after DCZ injection for all animals. White dotted lines indicate the extent of BBB disruption in the coronal plane. Abbreviations are as follows: A6M, area 6 of cortex, medial (supplementary motor) part; A6DR, area 6 of cortex, dorsorostral part; A6DC, area 6 of cortex, dorsocaudal part; A6Va, area 6 of cortex, ventral, part a; A6Vb, area 6 of cortex, ventral, part b; A8b, area 8b of cortex; A8C, area 8 of cortex, caudal part; A8aD, area 8a of cortex, dorsal part; A8aV, area 8a of cortex, ventral part; A9, area 9 of cortex; A11, area 11 of cortex; A13a, area 13a of cortex; A13b, area 13b of cortex; A13L, area 13 of cortex, lateral part; A13M, area 13 of cortex, medial part; A14C, area 14 of cortex, caudal part; A14R, area 14 of cortex, rostral part; A24a, area 24a of cortex; A24b, area 24b of cortex; A24c, area 24c of cortex; A25, area 25 of cortex; A32, area 32 of cortex; A32V, area 32 of cortex, ventral part; A45, area 45 of cortex; A46V, area 46 of cortex, ventral part; A47L, area 47 of cortex, lateral part; A47M, area 47 of cortex, medial part; A47O, area 47 of cortex, orbital part; AON, anterior olfactory nucleus; Cd, caudate nucleus; ProM, proisocortical motor region (precentral opercular cortex); OPro, orbital proisocortex; OPAl, orbital periallocortex.

Long-range chemogenetic neuromodulation via noninvasive AAV transduction

Because we observed increased activation—measured via FDG uptake—in connected regions in the contralateral hemisphere after DREADD activation, we hypothesized that we would observe similar increases in activation within other brain areas functionally connected to the insonication target. To test this, we employed the “fingerprinting” technique, which has been used previously to compare patterns of activity across a circuit (i.e., interareal connectivity),³⁶ rather than comparing intensity values with any single region. By vectorizing and scaling the patterns (here patterns of FDG uptake across connected regions), the fingerprint plots allow for visual comparison of overall patterns of activity. Using resting-state functional connectivity maps of insonicated regions collected from the Marmoset Brain Connectome,⁶⁴ we placed regions of interest (ROIs) along nodes of the functional circuit (as identified by sites of peak seed-based correlation values) of each insonication location (Figure 5). From these ROIs, we extracted weight-normalized FDG uptake from functionally connected cortical and subcortical locations for each animal and condition (vehicle vs. DCZ) to generate condition-dependent activation fingerprints. As expected, FDG uptake in regions connected to the insonicated location was elevated after DCZ administration as compared to vehicle injections (Figure 5). Indeed, elevations of activation in DCZ trials were apparent in nearly every functionally connected region to that receiving BBB disruption, especially areas containing dense projections from insonicated targets.⁶⁶ DREADD-mediated increases in FDG uptake were also observed within functionally connected regions in which there was no overlap with structural connections (Figure 5) suggesting that focal DREADD activation in discrete frontocortical targets results in large-scale, interhemispheric increases in activation within the larger functional network—likely representing polysynaptic connections. Critically, in marmosets Pa and S, changes in FDG uptake between DCZ and vehicle controls were lowest for the entorhinal cortex, a control region neither functionally (Z score < 2) nor directly structurally connected (lack of fluorescent reporter) to insonication targets in A8aD and A6DR, supporting the specificity of modulation.

Figure 5.

DCZ-mediated increases in glucose uptake are found across regions structurally and/or functionally connected to insonicated cortex

(A) Schematic outlining the process used to determine if ROIs functionally connected with insonicated cortex were also structurally connected with insonicated cortex. The leftmost image displays resting-state functional connectivity with the insonicated region (white asterisk) along with 1.5 mm diameter spherical ROIs (gray circles) centered at connectivity peaks (and mirrored to the contralateral hemisphere). RSfMRI⁶⁴ and viral anterograde tracer data (from injections approximating the location of insonication targets; image 2)^65,66 were binarized and combined to create overlap volumes (image 3; anterograde projections, green; functional connectivity, magenta; overlap, purple). Within each ROI, we determined the presence of overlap between binarized functional and anterograde projection data to designate ROIs as a “structural and functional” connection (overlap between structural and functional data, yellow circles) or a “functional” connection (no overlap between structural and functional data, cyan circles). An example overlap volume (image 4) and categorized ROIs on a surface of the marmoset brain.⁶⁸

(B) Surface maps demonstrating functional connectivity with the insonicated region (magenta), anterograde projections from the insonicated region (green), and functional connectivity/anterograde projection overlap (purple) for each animal. The white asterisk denotes the region of BBB disruption. White borders on the surface of the brain are from the Paxinos atlas⁶⁷ in Marmoset Brain Mapping space.^64,68 Color-coded ROIs are also included in surface maps for each animal (structural and functional connection, yellow; functional connection, cyan; control region demonstrating no marked structural or functional connectivity with the insonicated region, white). Scale bar, 10 mm.

(C) RSfMRI (left) and PET (right) interareal fingerprints for each animal included in this study. Note that functional connectivity (white outline) and weight-normalized FDG uptake (vehicle control, blue outline; DCZ, red outline) data are independently scaled between 0 (lowest values) and 1 (highest values) to facilitate comparisons of patterns between modalities (RSfMRI vs. PET) and experimental trials (DCZ vs. vehicle control). The nature of connectivity between each ROI and its corresponding insonication target is reflected in the color of its label (see above). ROI location relative to the area of BBB disruption is denoted by enclosing semi-circles (contralateral hemisphere, dark gray; ipsilateral hemisphere, light gray). Abbreviations are as follows: A6DR, area 6 of cortex, dorsorostral part; A8aD, area 8a of cortex, dorsal part; A8b, area 8b of cortex; A11, area 11 of cortex; A13M, area 13 of cortex, medial part; A23b, area 23b of cortex; A30, area 30 of cortex; Cd, caudate nucleus; Thal, thalamus; LIP, lateral intraparietal area of cortex; EC, entorhinal cortex.

Discussion

Here, we demonstrate noninvasive, focal transduction of functional excitatory chemogenetic proteins in marmoset frontal cortex. Using tFUS aided by systemically injected microbubbles, we transiently and focally disrupted the BBB, allowing for passage of viral constructs that transduced neurons with excitatory G(q)-DREADD proteins (Figure 1; see Parks et al.^21,33 for our previous work establishing this technique in the marmoset). Excitatory DREADDs were activated by noninvasively administered ligands and, using radiolabeled glucose as a proxy for neural activity,⁶³ we demonstrated modulation of frontal cortex regions in three nonhuman primates (Figure 2). Results from FDG PET experiments were supported by immunohistochemistry showing selective reporter expression at the site of insonication (Figure 3). Increased glucose metabolism was also observed in homotopic regions of cortex (Figures 2 and 4) as well as other connected circuitry (Figure 5) after DCZ administration. Using fully awake RSfMRI⁶⁴ and data from directly injected viral anterograde tracers,^65,66 we demonstrate that DCZ-driven increases in glucose metabolism were present within both structurally and functionally as well as only functionally connected regions as compared to control regions (those without significant functional connectivity; Figure 5). Taken together, these data demonstrate the efficacy of tFUS BBB disruption for noninvasive delivery of neuromodulators that, when actuated, alter activation levels across the functional circuit of insonication targets—including functionally connected regions not containing direct projections from regions in which the BBB was disrupted.

The central focus of this study was to expand tFUS-mediated viral gene delivery to include neuromodulators with both scientific and clinical applications; moreover, to demonstrate that tFUS-mediated gene delivery and modulation is viable in frontal cortex targets, with other extant reports (apart from our demonstration of tFUS AAV delivery in marmosets²¹) targeting subcortical regions.²⁰ Accordingly, three marmosets received a single insonication to disrupt the BBB within frontal cortex (Figure 1). The parameters of which were optimized from those utilized in Parks et al.,²¹ where we established the efficacy of noninvasive AAV delivery in the marmoset. Specifically, there were minor differences in the pressure delivered to the brain (1.27–1.38 vs. 1.17 MPa), the pulse duration (20–25 ms vs. 10 ms burst duration), and the microbubble dose given to the marmosets (350 vs. 200 μL/kg).²¹ Further modification to tFUS parameter sets should reflect several factors including the desired strength of opening, depth of targeting, and the attenuation of the signal in that specific animal; we have found that the skull derates acoustic pressure at ∼50%, but this is an average value measured in ex vivo samples.³³ Acoustic modeling, such as that shown in Di Ianni et al.,⁶⁹ in the marmoset will allow for more precise selection of parameters on an individual animal basis. Parameter optimization notwithstanding, this procedure is very rapid. In experienced hands, BBB disruption takes only a few minutes. The acquisition of GBCA MR images (Figure 1), however, takes significantly longer to complete (transport of animal into the MRI, MRI setup and image acquisition, etc.). While not strictly necessary, verification of BBB disruption with MRI is highly recommended; without employing MRI-based verification of BBB disruption, one incurs the risk of administering virus with an intact or minimally perturbed BBB. If an MRI scanner is not easily accessible, hydrophone measured cavitation feedback may be used to assess the likelihood of BBB disruption—albeit with greater uncertainty.^33,70 Consequently, following verification of BBB opening with GBCA MRI marmosets were immediately systemically injected with either AAV5 or AAV9 constructed to express the excitatory DREADD receptors: hM3D(Gq).

After allowing several weeks for viral expression, we conducted FDG PET imaging on all marmosets under two conditions: administration of the DREADD agonist DCZ⁴⁷ or a vehicle control. As compared to vehicle trials, FDG uptake was significantly higher within insonicated regions after systemic DCZ administration (Figure 2). These results are consistent with those of Mimura and colleagues,⁶⁰ who demonstrated increased relative FDG uptake at the target of a unilateral DREADD injection following receptor actuation. In support of our finding of increased FDG uptake within insonicated regions, immunohistochemical staining of tissue from marmoset V revealed neuronal cell bodies positive for mCherry that were limited to the region of BBB disruption (Figure 3). These data indicate successful noninvasive DREADD protein expression within a focal region of the primate brain. While recent work has advanced the use of tFUS BBB disruption to facilitate AAV-based gene transfer, these studies generated openings either broad in spatial extent and/or within rodents.^{5,14,15,16,17,18,19,20} Here, we demonstrate BBB disruption and transduction patterns with precision on the order of millimeters. Given the spatial extent of BBB disruption as measured via MRI (∼1 mm radially x ∼2.5 mm axially), the transduced cell count presented here, which arises from a single 50 μm coronal section, likely underestimates the total number of cells transduced within the target region due to sampling bias; these estimates may be further confounded by a low detection sensitivity for the mCherry fluorescent tag as evidenced by our extended staining protocol. As reliable alterations in macaque behavior have been shown after minimal (as low as approximately 3%⁷¹) cellular DREADD expression,^50,71 we consider the sparse pattern of DREADD transduction resulting from tFUS BBB opening to be advantageous for precise manipulation of neural circuits. Additionally, these data, which are consistent with Galvan et al.,⁷² highlight the importance of selecting gene payload combinations based on the model organism employed in experiments; indeed, DREADD constructs that readily localize to the cell membrane in primates—i.e., those that utilize an HA tag or employ a G(q) DREADD fused to mCherry—may be more effective than similarly fused inhibitory DREADDs.⁷² Future studies exploring FUS-based delivery of inhibitory DREADDs and quantifying the ultrastructural localization of transduced DREADDs across primate species (both inhibitory and excitatory, delivered by FUS) would be of great value. Similarly, future work should explore the use of alternative AAV serotypes, especially those specifically engineered for use with tFUS BBB disruption and detargeted from peripheral organs⁷³—a preferable property due to the promiscuity of many AAV serotypes.^74,75 Such studies would also benefit from work to optimize the timing of viral delivery, which may improve the transduction profile within the region of BBB disruption. Finally, work with enhancer elements, such as those designed to limit AAV expression to cortical inhibitory interneurons,⁷⁶ may also be advantageous to broaden the potential impact of this technique for both circuit modulation and the treatment of a variety of diseases/disorders (see Paterno et al.⁷⁷ for review).

In addition to increased FDG uptake at the site of insonication, we also observed marked elevations in FDG uptake within homotopic regions of cortex after DCZ injection (Figure 2) which we explored via examination of the functional and structural connectivity of insonicated targets. Using the Brain/MINDS Marmoset PFC Connectome,^65,66 the Marmoset Brain Connectome,⁶⁴ and PET data presented here, we compared the topography of anterograde projections from injection regions approximating insonicated cortex, group resting-state functional connectivity profiles from single voxels within insonicated regions, and individual animal FDG uptake data (Figure 4). Overlap between anterograde projections, resting-state functional connectivity, and DCZ-mediated elevations in FDG uptake was found for all animals (Figure 4); in combination with the focal cellular expression of hMD3(Gq) only within insonicated cortex (Figure 3), these data suggest monosynaptic communication with corresponding regions in the contralateral hemisphere may be responsible for the propagation of increased activation levels after DCZ administration. These results are in line with observations from our lab of transgene expression in anterograde projections after tFUS-mediated viral gene delivery.²¹ Albeit mediated by a different viral construct, these results are also consistent with those observed during an optogenetic neuromodulation paradigm whereby stimulation of cell bodies in one hemisphere could elicit increases in oxidative metabolism within homotopic cortex.⁷⁸

The multimodal combination of open access RSfMRI data⁶⁴ and anterograde tracer data^65,66 also enabled us to assess whether DCZ mediated alterations in glucose uptake were present in other regions only functionally connected with insonication targets. By determining overlap between functional connectivity and anterograde projections, we deduced whether ROIs were directly structurally connected with areas in which the BBB was perturbed. Explicitly, we used a binary measure of overlap between functional connectivity and anterograde projection maps from each insonication location to determine if regions functionally connected were also structurally connected with insonication targets (the presence of any overlap indicated a structurally and functionally connected region, if no overlap was found between anterograde projections from and functional connectivity with insonicated cortex, the ROI was classified as a functional connection; see “identification of anterograde structural connections”). In contrast to regions contralateral to the site of BBB disruption, regions ipsilateral to the insonication point presented high correspondence between anterograde projections and peaks of functional connectivity—i.e., were structurally and functionally connected with the region of insonication (Figure 5). DCZ-mediated increases in glucose uptake were apparent in nearly all regions both structurally and functionally connected with insonicated targets (Figure 5). These data support the findings of Michaelides and colleagues,⁶³ who demonstrated that local DREADD activation can modulate FDG uptake in connected regions of the rat brain. Further, we observed increases in FDG uptake after DREADD actuation in ROIs where overlap between functional connectivity and projections from insonicated cortex was absent, evidencing modulation of polysynaptically connected regions after actuation of excitatory chemogenetics within a single target area (Figure 5). It is important to note that our finding of modulation within regions polysynaptically connected to the insonication target is reliant on our method of ROI classification, which utilizes viral anterograde tracing data^65,66; as such, we cannot make definitive claims on the possibility that regions are connected in a retrograde fashion nor if detected fluorescence corresponds to synaptic terminations or axons of passage. It is also critical to note that ROIs were selected using group resting state connectivity maps,⁶⁴ while tracer data from each injection stem from a single animal^65,66 precluding our ability to account for inter-animal variations in structural connectivity.⁷⁹ Indeed, inter-animal variability may also account for points where results we present here diverge from regions canonically viewed to be monosynaptically connected (Figure 5)—e.g., A8aD and the lateral intraparietal area (LIP)^66,80—as peaks in group functional connectivity maps may not reflect the exact point of entry of projections from cells captured in a single injection. Overall, however, we demonstrate long-range alterations in glucose metabolism within functionally connected circuitries as the result of DCZ activation.

As is evident in Figure 2, FDG uptake increased most dramatically at the site of viral delivery and in connected circuitry, but there were also more subtle widespread increases (comparing vehicle and DCZ scans for each animal; Figure 2). Although the potential neuromodulatory effects of DCZ administration alone are not explicitly tested here, others have demonstrated that a DCZ dose of 0.1 mg/kg (as used in this study) has negligible effects on whole brain FDG uptake⁴⁷ and resting state functional connectivity.⁸¹ Rather, these data and those presented in other work employing invasive methods of cortical stimulation⁸² suggest that spreading activation may be a hallmark of focal neuromodulation. Moreover, using fMRI others have noted altered functional connectivity⁵² and task evoked activity⁵⁷ between regions directly and indirectly structurally connected to injection targets after DREADD actuation. These results are complemented by those of Hirabayashi and colleagues⁴⁹ who, using functional PET, found that chemogenetic inhibition of the orbitofrontal cortex modulated task-related activation in other structurally/functionally connected regions of the macaque brain. Further, Hirabayashi et al.⁴⁹ demonstrated that DREADD-mediated reductions in activity of orbitofrontal cortex are sufficient to abolish delay related activity for preferred stimuli in anterior ventral temporal cortex—an anatomically connected region—during an object memory task. Although we utilize an excitatory G(q)-DREADD construct in our experiments, we note a qualitatively similar spread of modulated activity from the site of insonication, particularly in connected regions (Figures 2, 4, and 5). The prospect of spreading depolarization in response to actuation of focal excitatory chemogenetics may be of great importance to adequately place in context any causal findings from behavioral paradigms incorporating DREADDs. Nevertheless, similarities between results in macaques and those presented here are particularly notable given the difference in method employed by each study—e.g., tFUS-mediated viral gene delivery—a noninvasive technique, recapitulates effects observed after direct intracranial injections.

In this work, we present the use of tFUS and lipid microbubbles to disrupt discrete regions of the BBB in marmoset frontal cortex, promoting focal, excitatory DREADD transduction via systemically delivered viral constructs. Across three marmosets, actuation of Gq-DREADD constructs via IV DCZ administration consistently increased radiolabeled glucose metabolism not only within insonicated regions, but in structurally and/or functionally connected circuitry as well. Taken together, these results reinforce the feasibility of tFUS BBB opening for the delivery of biologics in marmosets²¹ and represent a completely noninvasive suite of tools for chemogenetic circuit modulation in nonhuman primates. For these reasons, we expect this technique, which is dramatically simpler and faster than performing a neurosurgical injection, could make more practicable the field of nonhuman primate chemogenetics by increasing the feasibility with which other labs can access this foundational method of circuit modulation. Moreover, so long as the vector is impermeant to the BBB and remains intact after application of ultrasound, tFUS-mediated BBB disruption may prove an effective alternative to intracranial injection for targeted delivery of a variety of biologics—carrying genetic material for myriad proteins—to the primate CNS. Thus, this technique may empower the design of brain region-based, noninvasive gene therapies for the treatment of complex neurological diseases and disorders across the phylogenetic spectrum.

Limitations of the study

While we were unable to test for DREADD-mediated behavioral alterations in this study as all animals were selected based on a pre-designation for humane euthanasia (not allowing sufficient time for behavioral training), we specifically selected insonication targets based on their behavioral and translational relevance. Indeed, both insonicated regions presented in this study are involved in the control of saccades in marmosets.^83,84 Future studies validating robust and reliable behavioral effects will be critical for demonstrating the validity of the approach described here. Behavioral validation notwithstanding, our results demonstrate the feasibility of an approach that is dramatically simpler and faster than performing a neurosurgical injection. Moreover, this report is one of the first to demonstrate noninvasive, focal AAV delivery to primate cortex and we look forward to a formal assessment on how tFUS-mediated viral gene delivery can be used for the modulation of well characterized behaviors.

Resource availability

Lead contact

Requests for further information and resources should be made to the lead contact, Dr. David J. Schaeffer (dschaeff@pitt.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We wish to thank Brianne L. Stein, Lauren Dubberley, and Dr. Julia Oluoch for animal care and preparation. This work was supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under award number R21NS125372 (D.J.S.) and the National Institute of General Medical Sciences of the National Institutes of Health under award number T32 GM142630 (M.R.C.).

Author contributions

M.R.C., experimentation, data preprocessing, data analysis, immunohistochemistry and microscopy, figure generation, writing, and editing; T.V.P., experimentation, data preprocessing, data analysis, immunohistochemistry and microscopy, figure generation, writing, and editing; S.S.G., analysis, figure generation, and editing; L.L., immunohistochemistry and microscopy, writing, and editing; I.Z.R., experimentation and editing; S.K.A.P., analysis and editing; V.P.C., data preprocessing (MRI denoising), writing, and editing; D.S., pulse sequence optimization and editing; D.J.S., study conception, experimentation, analysis, immunohistochemistry and microscopy, figure generation, writing, and editing.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rat mCherry Monoclonal Antibody (16D7), Alexa Fluor™ 594	ThermoFisher Scientific	Cat#M11240; RRID: AB_2536614
DAPI (4′,6-Diamidino-2-Phenylindole)	ThermoFisher Scientific	Cat#62248
Bacterial and virus strains
AAV5-SYN1-HA-hM3D(Gq)	Galvan et al.72; Addgene	Cat#121539-AAV5; RRID: Addgene_121539
AAV9-hSyn-hM3D(Gq)-mCherry	Addgene	Cat#50474-AAV9; RRID: Addgene_50474
Chemicals, peptides, and recombinant proteins
Deschloroclozapine powder	Sigma-Aldrich, Inc.	Cat#SML3651-5MG
Deschloroclozapine dihydrochloride powder	Hello Bio Inc	Cat#HB9126-10mg
Deposited data
Brain/MINDS Marmoset PFC Connectome (DataID: 6358)	Skibbe et al.65; Watakabe et al.66	https://doi.org/10.24475/bminds.mti.6358
Marmoset Functional Connectivity Resource	Schaeffer et al.64	https://www.marmosetbrainconnectome.org
Experimental models: Organisms/strains
Common marmoset (Callithrix jacchus)	University of Pittsburgh	N/A
Software and algorithms
Analysis of Functional NeuroImages (AFNI) version AFNI_23.0.03	Cox85; Cox and Hyde86	https://afni.nimh.nih.gov
FMRIB Software Library (FSL)	Jenkinson et al.87; Smith et al.88	https://fsl.fmrib.ox.ac.uk/fsl/docs/
R	R Core Team89	https://www.R-project.org/
PMOD version 4.4	PMOD Technologies LLC	https://www.pmod.com/web/
ZEN 3.5, blue edition	Zeiss	https://www.zeiss.com/microscopy/us/products/software/zeiss-zen.html
MATLAB version 9.14.0.2206163 (R2023a)	MathWorks	https://www.mathworks.com
Other
Focused Ultrasound Apparatus	FUS Instruments Incorporated	RK-50 Marmoset

Experimental model and study participant details

Animals

Data were collected from three adult female marmosets (Callithrix jacchus; see Table - Subject and viral construct details) housed at the University of Pittsburgh.

Subject and viral construct details.

Marmoset	Age (months)	Weight (g)	Sex	Viral Construct	Titer (GC/mL)	Vol. (μL)	Dose (vg/kg)	Time to Scan (DCZ)
Pa	7	175	F	AAV5-SYN1-HA-hM3D(Gq)72; Addgene 121539-AAV5	1.1 x 1013	170	1.07 x 1013	17 days
V	51	255	F	AAV9-hSyn-hM3D(Gq)-mCherry; Addgene 50474-AAV9	2.8 x 1013	200	2.20 x 1013	27 days
S	25	260	F	AAV5-SYN1-HA-hM3D(Gq)72; Addgene 121539-AAV5	1.1 x 1013	170	7.19 x 1012	38 days

Female animals were chosen for these foundational experiments based on availability for terminal procedures. Marmosets were either opposite-sex pair housed or family group housed, fed a diet consisting of twice daily provisions of commercial chow including purified diet, and supplemented with fresh fruit and vegetables daily with drinking water provided ad libitum. Foraging materials and enrichment were provided daily. Prior to tFUS, MRI and PET procedures, marmosets received intramuscular injections of meloxicam (0.2 mg/kg; 07-893-5916, Patterson Vet Supply, Inc., Loveland, CO, USA) and glycopyrrolate (0.005 mg/kg; 1226607, McKesson Medical-Surgical, Richmond, VA, USA). Animals were anesthetized (induced and maintained) with 2% isoflurane delivered via mask and fitted with a 26-gauge venous catheter for tFUS, MRI, and PET procedures. During the procedures, heart rate, blood oxygen, respiration, and rectal temperature were monitored. Body temperature was maintained with heated water blankets or heated air. All experimental procedures were in accordance with ethical guidelines for animal testing established by the University of Pittsburgh Institutional Animal Care and Use Committee.

Method details

Focused ultrasound BBB disruption

To transiently disrupt the BBB within a focal region of cortex for noninvasive viral vector delivery, marmosets were placed in an RK-50 Marmoset tFUS apparatus (FUS Instruments Incorporated, Toronto, ON, Canada) incorporating a marmoset stereotax (Model SR-AC; Narishige International Incorporated, Amityville, New York, USA) and an MRI-based atlas for stereotactic targeting^64,67,68 (https://www.marmosetbrainconnectome.org). The tFUS apparatus used here is described in detail in³³ and summarized in.²¹ Briefly, an automated 3-axis position system guided a 35 mm spherically focused 1.46 MHz transducer (FUS Instruments Incorporated, Toronto, ON, Canada) that was rigidly mounted to the positioning system to allow for precise control of the sonication location. The number of pulses, repetition period, and number of bursts commanded by the software were generated by an external waveform generator (Siglent SDG 1032X, Siglent Technologies, Solon, Ohio, USA) and sent through a 15 W amplifier to the transducer. The transducer was sealed using a 3D printed cover incorporating an o-ring seal and a polyimide film face that held degassed water. The transducer and cover were immersed in a ∼300 mL tank containing degassed water, the polyimide bottom of which was coupled to the head using degassed, warmed ultrasound transmission gel (01–50, Parker Laboratories, INC., Fairfield, NJ, USA). Water degassing was accomplished using a FUS-DS-50 portable water degasser (FUS Instruments Incorporated, Toronto, ON, Canada).

Transcranial insonication was based on stereotactic position, of which the zero point (x, y, z = 0, 0, 0) corresponded to a location equidistant from the center of the ear bars and in the axial plane of the inferior most aspect of the orbit bars. tFUS positioning software (Morpheus framework; FUS Instruments Incorporated, Toronto, ON, Canada), which incorporates the marmosetbrainconnectome.org and⁶⁴ the marmosetbrainmapping.org atlases⁶⁸, was used to select predefined insonication targets based on the Paxinos parcellation of the marmoset brain.⁶⁷ These locations were further refined based on preprocedural MRI (using the same parameters defined below). Two regions were selected for sonication, area 8aD for its role in controlling eye movements in the marmoset,^83,84 and area 6DR which, in addition to its role in the neural control of eye movements,^83,84 is a central functional connectivity hub in the marmoset.^90,91

At the beginning of all tFUS procedures, the marmoset scalp was cleared of hair using a combination of electric clippers and depilatory cream (Nair Shower Cream, Church & Dwight Co, Inc., Ewing, NJ, USA). Immediately prior to insonication, a bolus of lipid microbubbles (350 μL/kg; Definity, Lantheus Medical Imaging, Billerica, MA, USA) - prepared by mixing 0.86 mL of sterile saline with 0.1 mL of microbubbles - was administered via the implanted catheter to aid in BBB disruption. To ensure complete clearance of the microbubble solution from the catheter hub, a flush of 200 μL sterile saline was administered to all marmosets. Following microbubble injection, the 1.46 MHz transducer was oriented directly above the insonication target for a single sonication with the following parameters: acoustic pressure = 1.27–1.38 MPa (estimated with 47% deration³³), burst duration = 20-25 ms, burst period = 1000 ms, number of bursts = 60 bursts.

MRI acquisition and denoising

To visualize BBB opening after insonication, anesthetized marmosets were transferred from the tFUS apparatus to a 9.4 T 30 cm horizontal bore MRI scanner (Bruker BioSpin Corp, Billerica, MA, USA), equipped with a Bruker BioSpec Avance Neo console and the software package Paravision-360 (version 3.6; Bruker BioSpin Corp, Billerica, MA, USA). Marmosets were placed into a custom-designed MRI-compatible stereotax for imaging (see https://www.marmosetbrainconnectome.org/apparatus/ for computer-aided design files),^92,93 after which 100 μL of a gadolinium-based MRI contrast agent (GBCA; Gadavist, gadobutrol; Bayer Healthcare Pharmaceuticals, Leverkusen, Germany) was systemically injected through the implanted catheter. GBCA was cleared from the catheter hub using 200 μL of sterile saline. During imaging, radiofrequency transmission was accomplished with a custom 150 mm inner diameter coil. A custom-built 14-channel phased-array marmoset-specific coil was used for radiofrequency receiving.⁹² A magnetization-prepared rapid gradient echo (MPRAGE) sequence was used to detect reduced T1 relaxation times resulting from gadolinium extravasation at the site of tFUS BBB permeation with the following parameters: TR = 6,000 ms, TE = 3.30 ms, field of view = 42 × 35 × 25 mm, matrix size = 168 × 140 × 100, voxel size = 0.250 × 0.250 × 0.250 mm, bandwidth = 50 kHz, flip angle = 12°, total scan time = 20 min, 6 s. To enhance the signal-to-noise ratio (SNR) of the anatomical MR images, we applied a denoising procedure as in.^94,95 We used a variance-stabilizing transformation (VST) framework tailored for Rician-distributed noise, characteristic of magnitude MR images.⁹⁶ The forward VST was applied to transform the heteroscedastic Rician noise into approximately homoscedastic Gaussian noise. The transformed images were then denoised using the block-matching and 4D filtering (BM4D) algorithm.⁹⁷ Finally, the inverse VST was applied to restore the denoised images to their original intensity range.

Systemic AAV injections

Immediately after BBB opening was verified by GBCA extravasation at the targeted site, either AAV9 or AAV5 (Addgene, Watertown, MA, USA; see the table in the STAR Methods for details and dosages. Subject and viral construct details, for details and dosages) were injected as a single bolus through the implanted catheter. Following AAV injection, a 200 μL flush of sterile saline cleared virus from the catheter hub. At the conclusion of experiments, the scalp was covered in Aquaphor ointment (Beiersdorf Inc., Wilton, CT, USA) to prevent irritation from hair removal. After animals had sufficiently recovered from anesthesia, they were returned to their home cages.

[¹⁸F]-FDG PET imaging of DREADD actuation

In the weeks following viral delivery (Marmoset Pa, 17 days; Marmoset V, 27 days; Marmoset S, 38 days), all marmosets underwent anesthetized FDG PET scans to noninvasively probe the effects of DREADD actuation. Marmosets were fixed in stereotactic plane using a custom-built marmoset PET stereotax (https://www.marmosetbrainconnectome.org/apparatus/)⁹² and fit with a 26-gauge venous catheter. Immediately prior to FDG imaging, animals received either a single dose of DCZ (100 μg/kg) or 2.5% dimethyl sulfoxide (DMSO, D2650-100ML, Sigma-Aldrich, Inc., St. Louis, MO, USA) diluted in sterile saline (vehicle solution) via the implanted catheter. Two preparations of DCZ were used in this study: 1) DCZ powder (SML3651-5MG, Sigma-Aldrich, Inc., St. Louis, MO, USA) dissolved in DMSO at a concentration of 2 mg/mL, and then diluted to a final concentration of 2.5% DMSO in sterile saline (Marmosets Pa and S) or 2) DCZ dihydrochloride powder (HB9126-10mg, Hello Bio Inc, Princeton, NJ, USA) dissolved in distilled water at a concentration of 1 mg/mL (Marmoset V). At the time of DCZ/vehicle solution administration, marmosets also received a bolus of FDG at a dose of 80 MBq (2.16 mCi). Immediately following both radio-and-synthetic ligand dosing, PET data were collected using a Bruker Si-78 small animal PET/CT (Bruker BioSpin Corp, Billerica, MA, USA), with a console running ParaVision 360 (version 3.6, Bruker BioSpin Corp, Billerica, MA, USA). PET scans were acquired over 90-min with a field-of-view of 120 × 80 × 80 mm. PET scans were dynamically reconstructed using a maximum a posteriori projection (MAP 0.5) in the following bins: 3x20; 4x30; 2x60; 5x300; 6x600 s. At the conclusion of PET imaging, a computed tomography (CT) image was acquired to aid in image registration with the following parameters: field of view = 79.6 × 81.3 mm, pixel size = 200 μm, X-ray source filter = 1 mm aluminum, frame averages = 1, scanning mode = step and shoot, rotation angle = 1°.

Immunofluorescence staining

To conduct immunofluorescence staining, Marmoset V was euthanized via injection of a solution containing 390 mg/mL pentobarbital sodium and 50 mg/mL phenytoin sodium (100 mg/kg, EUTHASOL, Virbac, Westlake, TX, USA). Following euthanasia, the animal was perfused transcardially with cold phosphate buffer saline (PBS) and then 4% paraformaldehyde (PFA). The brain was extracted, post-fixed in 4% PFA, and cryoprotected in ascending concentrations of sucrose (10%–30%). Following cryoprotection, the brain was embedded in tissue freezing medium (TFM-C, General Data Healthcare, Cincinnati, OH, USA), frozen, and sectioned at 50 μm via a Leica CM1950 cryostat (Leica Biosystems, Deer Park, IL, USA). Sections were stored at −20°C in cryoprotectant solution consisting of 30% glycerol, 30% etholene glycol, 30% distilled water and 10% pH 7.2 0.2M phosphate buffer. To begin immunofluorescence staining, cryoprotectant solution was removed from free floating sections via a series of 1x PBS washes (3 × 10 min). Sections were then gently rocked in blocking solution (5% Donkey Serum, 1% Bovine Serum Albumin, 0.2% Triton-X, and PBS) at room temperature for 60 min. Following the blocking step, sections were incubated at 4°C in primary antibody solution containing a rat anti-mCherry amplifier conjugated to a 594 nm fluorophore (1:250, M11240, ThermoFisher Scientific, Waltham, MA, USA), 1% Donkey Serum and 1x PBS. Sections were gently agitated in primary antibody solution for a total of 72 h. Upon completion of the primary antibody incubation, tissue sections were washed in 1x PBS (3 × 10 min). All tissue sections were incubated with DAPI (4′,6-Diamidino-2-Phenylindole, 1:1000, 62248, ThermoFisher Scientific, Waltham, MA, USA) in 1x PBS for 10 min followed by additional 1x PBS washes (2 × 10 min). Tissue sections were mounted on Superfrost Plus slides (22037246, ThermoFisher Scientific, Waltham, MA, USA) and allowed to dry. Slides were coverslipped using Immu-Mount (9990402, ThermoFisher Scientific, Waltham, MA, USA) prior to imaging.

Quantification and statistical analysis

[¹⁸F]-FDG PET image processing and analysis

To prepare data for an analysis of FDG uptake after DREADD actuation, all PET and MRI scans were preprocessed using PMOD quantification software (version 4.4, PMOD Technologies LLC, Zurich, Switzerland). In brief, the anatomical orientation of each marmoset in their PET images was verified using lateralized soft and bony tissue landmarks (PMOD’s View). The orientation of the brain in anatomical MR images was verified using the site of insonication - a feature demarcating the left hemisphere (PMOD’s View). PET scans were time corrected (PMOD’s View) and brought into alignment with their anatomical MR image (PMOD’s Fusion). After initial alignment, PET images from each animal were rigidly registered to their anatomical MRI (PMOD’s Fusion). PET scans were then averaged 30 min–90 min (bins 15 to 20, PMOD’s Fusion) and converted to standard uptake value normalized to body weight (SUVbw) images (PMOD’s Fusion). At the conclusion of pre-processing, PET images were converted from digital imaging in communications and medicine (DICOM) format to neuroimaging informatics technology initiative (NIfTI) files using the PMOD View tool. Post-tFUS MRI images were converted to NIfTI format using Analysis of Functional NeuroImages (AFNI, version AFNI_23.0.03) software (dcm2niix_afni)^85,86,98 for further analysis outside of PMOD.

Analysis was conducted on data from animals separately and as a group. For individual animals, comparisons of FDG uptake within the insonicated and corresponding site in the contralateral hemisphere between DCZ and vehicle control trials were conducted using pre-processed PET scans in native space. Using PMOD software (PMOD’s Fusion), 1.5 mm diameter spheres were placed in the region of BBB disruption and mirrored to the contralateral (non-insonicated) region of cortex. Values from each pixel within insonicated and contralateral non-insonicated ROIs were extracted (PMOD’s Fusion, Pixel Dump), pre-processed, and visualized via plot of mean and standard deviation using R⁸⁹ in RStudio (2023.06.2 + 561, Posit Software, PBC).^{99,100,101,102,103,104} In R, data were assessed for normality via univariate Shapiro-Wilk tests (shapiro_test)¹⁰⁵ with p < 0.05 indicating a non-normal distribution. Due to a high prevalence of non-normal distributions (data not shown), comparisons of weight-normalized FDG uptake within insonicated cortex and the corresponding region in the contralateral (non-insonicated) cortex between trials (DCZ and vehicle) were conducted using a paired samples Wilcoxon signed rank test in R (wilcox_test)¹⁰⁵; a p < 0.05 indicated a significant difference between trial type.

A group analysis of FDG uptake within the insonicated site between trials was conducted using PET images registered and resampled to a marmoset brain template image^64,68 (see the proceeding section for a more detailed description of processing). Using the FMRIB Software Library (FSL),^87,88 a 1.5 mm diameter sphere was placed within the region of BBB disruption for each animal (fslmaths). Weight normalized FDG uptake was extracted from each voxel within the sphere ROI (AFNI’s 3dmaskdump) for each experimental condition and combined across animals. As above, normality testing was conducted on the data using R in RStudio (shapiro_test)¹⁰⁵ and indicated a non-normal distribution (data not shown). A paired samples Wilcoxon signed rank test was conducted in R (wilcox_test)¹⁰⁵ to compare weight normalized FDG uptake within the region receiving tFUS across animals and between conditions in which either DCZ or vehicle control was administered; the criterion for significance remained p < 0.05. For all statistical tests, multiple comparison correction was conducted using the Bonferroni method.

RSfMRI and [¹⁸F]-FDG PET fingerprinting analysis

To examine the effects of DREADD actuation within the functional circuit of each insonication target, we generated region specific fingerprints^{36,106,107,108} using open access RSfMRI data⁶⁴ and FDG PET data collected for this manuscript. Preprocessed, MPRAGE images were skull-stripped using UNet Studio software (version for Windows 7+)¹⁰⁹ and linearly registered to a T1-weighted template image^64,68 using the FSL FLIRT command. MR images in template space were then nonlinearly registered to the same template^64,68 using Advanced Normalization Tools (ANTs).¹¹⁰ Transformation matrices from both linear and nonlinear registration were saved for each animal and applied to their respective PET images yielding files for group analysis.

ROIs were selected using RSfMRI correlation maps generated by selection of a single voxel within each animal’s insonicated region via the Marmoset Brain Connectome webpage (https://www.marmosetbrainconnectome.org) - an open access resource of RSfMRI data from 32 marmosets.⁶⁴ Discrete regions of cortex/subcortex were selected based on the presence of robust, well-defined functional connectivity peaks with the insonicated region (threshold = Z score > 2). A control region was also selected for each animal based on a lack of both significant correlation and anticorrelation with the insonicated region (threshold = within Z score −2 to 2). Note that each point was mirrored in the contralateral hemisphere after selection, yielding ROIs in both the right and left hemispheres of the brain. Based on these criteria, 11 to 13 ROIs were selected for each animal included in this study. At each point, we generated a 1.5 mm diameter sphere (FSL’s fslmaths) from which an average value was extracted (AFNI’s 3dmaskave) from RSfMRI and PET images resampled to the T1-weighted template image referenced above (AFNI’s 3dresample). Average RSfMRI and PET (DCZ and vehicle control trials) values were rescaled from 0 (lowest functional connectivity or FDG uptake) to 1 (highest functional connectivity or FDG uptake) in MATLAB (version 9.14.0.2206163 (R2023a), MathWorks, Natick, MA, USA). Rescaling data allowed us to compare patterns of activation between modalities directly when presented radially (a customized version of radarplot¹¹¹).

Identification of anterograde structural connections

To examine the anatomical underpinnings of insonicated region functional circuits, we compared the topographies of RSfMRI correlation maps⁶⁴ and viral anterograde tracer data^65,66 downloaded from open access web-based resources. Anterograde tracing data were collected from three injections included in the Brain/MINDS Marmoset PFC Connectome (DataID: 6358, https://doi.org/10.24475/bminds.mti.6358).^65,66 Injection locations were chosen based on their proximity to the insonication target of each marmoset; two injections from A8aD (Marmosets Pa and V) and one from A6M (approximating insonication of A6DR in Marmoset S). Injection selections, by marmoset, were as follows: Marmoset V, thy1-tTA 1/TRE-clover 1/TRE-Vamp2mPFC, 0.2 μL injected, injection R01-0054; Marmoset Pa, thy1-tTA 1/TRE-clover 1/TRE-Vamp2mPFC, 0.2 μL injected, injection R01-0059; Marmoset S, thy1-tTA 1/TRE-clover 1/TRE-Vamp2mPFC, 0.2 μL injected, injection R04-0079. See https://dataportal.brainminds.jp/marmoset-tracer-injection/viewer for further injection details.^65,66 Tracer data were then registered to the Marmoset Brain Mapping template^64,68 using ANTs (antsApplyTransforms). RSfMRI maps collected for each insonication location (see RSfMRI and [18F]-FDG PET fingerprinting analysis) were then resampled to the tracer data in template space (AFNI’s 3dresample). Thresholds of 3.28 (Z score, p = 0.001 for RSfMRI maps) and 0.2 were applied to functional and tracer data, respectively. A threshold of 0.2 was chosen for anterograde tracer data to reduce noise induced as a result of registration to MBM space, thereby ensuring the spatial extent of the fluorescent signal in template space^64,68 approximated that in native space.^65,66 After applying these thresholds, functional connectivity and anterograde projection data were binarized (AFNI’s 3dcalc). We then determined whether binarized functional and anterograde projection data were overlapping in the space contained within each ROI defined for fingerprinting analyses (AFNI’s 3ddot). ROIs were categorized as being “structurally and functionally” connected to the insonication target if any overlap between anterograde projection and functional connectivity data was detected within the ROI. ROIs in which no overlap was detected between functional and structural connectivity data were categorized as “functional” connections. To visualize the location of ROIs in relation to resting-state functional connectivity and anterograde projections, overlap maps were generated in Connectome Workbench (version 1.5.0)¹¹² and, along with ROIs, displayed on surface maps of the marmoset brain.⁶⁸

Microscopy and cell counting

To quantify the extent of transduction resulting from focal BBB disruption, stained sections were imaged on an AxioImager M2 epifluorescence microscope (Carl Zeiss, White Plains, NY, USA). Acquisition settings were kept constant for all images within a staining group. Margins for tiled layout images were hand delineated. Large-scale tiled images were obtained at 20x and stitched using ZEN microscopy software (ZEN 3.5, blue edition, Zeiss, Oberkochen, Germany). As described above, insonicated regions were located in vivo via extravasation of GBCA into the surrounding neuropil. An MR image demonstrating the locus of BBB disruption in Marmoset V was used to identify this same location in sectioned tissue. In the region of BBB opening, cells with clear neuronal morphology that were also positive for mCherry were manually counted.

Published: May 4, 2026