Whole-brain mapping reveals the divergent impact of ketamine on the dopamine system
This pre-print brain-mapping study in mice (2023) shows that repeated ketamine administration decreases dopamine neurons in the midbrain and increases in the hypothalamus. It also reveals further evidence for the plasticity-increasing effects of ketamine.
Authors
- Chauhan, S.
- Chen, Y.
- Cruz, E. D. L.
Published
Abstract
Ketamine is a multifunctional drug with clinical applications as an anesthetic, as a pain management medication and as a transformative fast-acting antidepressant. It is also abused as a recreational drug due to its dissociative property. Recent studies in rodents are revealing the neuronal mechanisms that mediate the complex actions of ketamine, however, its long-term impact due to prolonged exposure remains much less understood with profound scientific and clinical implications. Here, we develop and utilize a high-resolution whole-brain phenotyping approach to show that repeated ketamine administration leads to a dosage-dependent decrease of dopamine (DA) neurons in the behavior state-related midbrain regions and, conversely, an increase within the hypothalamus. Congruently, we show divergently altered innervations of prefrontal cortex, striatum, and sensory areas. Further, we present supporting data for the post-transcriptional regulation of ketamine-induced structural plasticity. Overall, through an unbiased whole-brain analysis, we reveal the divergent brain-wide impact of chronic ketamine exposure on the association and sensory pathways.
Research Summary of 'Whole-brain mapping reveals the divergent impact of ketamine on the dopamine system'
Introduction
Ketamine is a clinically used dissociative anaesthetic and a fast-acting antidepressant, but it also has recreational abuse potential and produces dissociative effects. Pharmacologically it acts broadly in the brain, most prominently as a non-competitive antagonist of N-methyl-D-aspartate receptors (NMDARs) and also via effects on HCN1 channels and possibly opioid receptors. Earlier rodent studies have shown that acute or single-dose ketamine modulates synaptogenesis in prefrontal cortex and increases firing of ventral tegmental area (VTA) dopamine (DA) neurons with enhanced DA release in frontal cortex and striatum, but the long-term, brain-wide consequences of repeated or chronic ketamine exposure across a range of doses remain poorly understood despite clinical and public-health importance. S. and colleagues set out to map systematically how chronic (R,S)-ketamine exposure alters the entire dopaminergic modulatory system in mouse brain. Using sub-hypnotic doses (30 and 100 mg/kg) administered daily and a high-resolution, whole-brain phenotyping pipeline, they aimed to quantify changes in tyrosine hydroxylase (TH)+ cell bodies and TH+ projections across the brain, and to probe whether post-transcriptional regulation of TH mRNA contributes to any observed structural plasticity. The work seeks to reveal dose-dependent and region-specific adaptations of DA-related systems after repeated ketamine treatment.
Methods
S. and colleagues used male TH-2a-CreER mice crossed with a tdTomato reporter line, housed on a 12:12 light:dark cycle. (R,S)-ketamine was administered intraperitoneally at 30 mg/kg or 100 mg/kg once daily for 1, 5, or 10 days; saline injections were used as controls. 4-hydroxytamoxifen (4‑OHT) was given prior to ketamine exposure to induce Cre-mediated tdTomato labelling of pre-existing TH mRNA+ neurons in the inducible TH-CreER line. Locomotor activity was recorded 15 minutes and 1 hour after injections on treatment days. Activated caspase-3 immunostaining was used to assess cell death after the highest-dose, 10-day regimen. Whole brains were processed for protein and mRNA detection. TH protein was detected by α-TH immunolabelling and TH mRNA by hybridization chain reaction (HCR-FISH) using split-initiator probes. Samples were cleared with iDISCO+ or passive CLARITY, and refractive-index matched. Imaging used CLARITY-optimised light-sheet microscopy (COLM) and ClearScope with high-NA objectives. For the inducible‑label experiments, 4‑OHT was administered one week before ketamine to permanently label pretreatment TH mRNA+ cells by tdTomato; after 10 days of ketamine or saline the same brains were immunostained for TH protein and imaged. The authors developed an analysis pipeline (suiteWB) for whole-brain phenotyping. Registered high-resolution images were segmented with a semi-supervised multi-model approach using ilastik for pixel classification (sparse and dense workflows), followed by blob detection (difference of Gaussians) and block-wise processing. Cell detections from channels (tdTomato, α‑TH) were merged and identities assigned by validated thresholding. Projection segmentation used semi-supervised pixel classification and binarisation followed by convolution with 3D Gaussian kernels to estimate locally normalised projection densities. Whole-brain annotations were aligned to the Allen Brain Atlas hierarchy and analyses were performed at multiple annotation levels; voxel-based parcellation approaches (20x20x20 µm3 sampling for parcellation, 25x25x25 µm3 for projection statistics) were also used. Statistical comparisons used ANOVA and two-sided Mann–Whitney U tests with Bonferroni correction on normalised cell counts and densities. The extracted text does not clearly report the number of animals per group.
Results
Cell death and behaviour: Activated caspase-3 staining after 10 days of daily 100 mg/kg ketamine showed no significant cell death in the brain. In locomotor measures, the 30 mg/kg group, but not the 100 mg/kg group, exhibited increasing locomotor sensitivity across days at the 15-minute post-injection timepoint; measurements at 1 hour post-injection are reported but no dose-related increase was noted for 100 mg/kg. TH+ cell body counts — dose- and region-specific effects: High-resolution whole-brain TH+ maps were generated after 1, 5 and 10 days of treatment; robust, statistically significant alterations were detected only after 10 days for both 30 and 100 mg/kg, with larger effects at 100 mg/kg. At a coarse (level 6) annotation, both doses produced a dose-dependent decrease in TH+ neuron counts within behaviour-state-related midbrain regions (MBsta) and a concurrent increase within the hypothalamic lateral zone (LZ). At a finer (level 8) annotation, both doses produced decreases in the dorsal raphe (DR) and increases in the zona incerta (ZI). The 100 mg/kg group showed additional decreases in the retrorubral area (RR) and increases in arcuate hypothalamic nucleus (ARH), periventricular hypothalamic posterior part (PVp), tuberomammillary nucleus (TM) and PVHd. The 30 mg/kg group displayed significant increases in the substantia nigra reticular part (SNr) and decreases in the dorsomedial hypothalamic nucleus (DMH) and medial preoptic nucleus (MPN). These results were validated by an independent voxel-based brain parcellation approach (20x20x20 µm3 sampling). TH mRNA versus protein and inducible-labelling results: Whole-brain HCR-FISH revealed that TH mRNA is distributed more widely than TH protein, suggesting a pool of translationally suppressed TH mRNA+ neurons. Using TH-CreER induced tdTomato labelling of pretreatment TH mRNA+ neurons, the authors compared tdTomato labels (pretreatment TH mRNA+) with post-treatment TH protein labelling. After 10 days of 100 mg/kg ketamine, there was a significant increase in neurons co-labelled for TH protein and tdTomato (α‑TH ∩ tdTomato) compared with saline controls, while the overall number of tdTomato+ neurons remained unchanged. Conversely, in midbrain regions that showed decreased TH+ counts, there was a decrease in TH protein+/tdTomato+ cells while tdTomato+ and TH mRNA+ counts remained unchanged. The authors present these results as evidence that recruitment from a pool of untranslated TH mRNA+ neurons can rapidly alter the number of TH protein‑expressing cells regionally. Long-range TH+ projections: Voxel-by-voxel comparison of TH+ projection densities (25x25x25 µm3 sampling) after 10 days of 100 mg/kg ketamine revealed regionally divergent changes. Increased TH+ projection densities were observed in associative cortical regions, including prelimbic cortex (PL), orbital area (ORB), frontal pole (FRP), anterior cingulate area (ACA), and posterior parietal association area (PTLp). Subcortical increases occurred in lateral amygdala (LA), parts of the lateral septal complex (SF and TRS), olfactory-receiving regions (taenia tecta TT and piriform PIR), thalamic lateral posterior nucleus (LP), and hypothalamic VMH and TU. In contrast, decreased TH+ innervation was found in sensory and spatial-processing regions including dorsal auditory area (AUDd), posterolateral visual area (VISpl), entorhinal area (ENT), presubiculum (PRE), and in striatal subregions caudoputamen (Cp) and bed nuclei of the stria terminalis (BST/BNST), as well as specific thalamic and midbrain nuclei (SPA, SPF, SAG, PPN). The authors note that TH+ projections are catecholaminergic and cannot be unambiguously distinguished as dopaminergic versus noradrenergic by TH immunolabelling alone. Method performance metrics reported: segmentation F1 scores of 0.95 (sparse) and 0.86 (dense), with precision 0.889 and 0.976 respectively versus human annotations.
Discussion
S. and colleagues interpret their results as demonstrating a previously unrecognised, divergent adaptability of the dopamine-related system to repeated ketamine exposure: chronic ketamine reduced TH protein‑expressing neuron counts in behaviour-state-related midbrain domains while increasing TH+ counts in hypothalamic domains, and produced increased innervation of associative cortical centres alongside decreased innervation of sensory and spatial-processing areas. The authors suggest this structural plasticity of a modulatory system could facilitate substantial reconfiguration of neuronal networks in response to external stressors (such as ketamine) or disease states, potentially leading to durable cognitive and behavioural changes. They acknowledge key limitations reported in the paper. TH immunolabelling cannot definitively distinguish dopaminergic from noradrenergic neurons or projections, since TH marks catecholamine-synthesising capacity; noradrenergic somata are known to be spatially restricted in hindbrain regions but projections distribute widely. A loss of TH protein does not necessarily indicate neuronal death — it indicates a loss of the capacity to synthesise L‑DOPA via TH and thus reduced DA production — and the authors observed no evidence of caspase-mediated cell death after the high-dose regimen. The authors propose that the observed reversible‑appearing plasticity is enabled by a substantial pool of neurons that maintain translationally suppressed TH mRNA, permitting rapid region-specific recruitment or withdrawal of TH protein expression in response to chronic ketamine. Mechanistically, they suggest repeated ketamine may drive these changes indirectly via NMDAR antagonism and heterogeneous circuit inputs that activate or deactivate different DA neuronal populations. They note the ketamine‑induced plasticity they observe differs from previously reported neurotransmitter-phenotype changes (for example those driven by altered light cycle), which included loss of TH mRNA expression. Finally, the authors argue their findings highlight the need for unbiased, whole-brain evaluations of ketamine's on‑ and off‑target effects across doses and durations, and they call for development of targeted therapeutic approaches to manage complex brain disorders; they state that all datasets will be made available on request.
View full paper sections
INTRODUCTION
Ketamine is a schedule III (US Food and Drug Administration) substance with clinical applications as a dissociative anesthetic, as a pain management drug and, most recently, as a transformative fast-acting antidepressant. Pharmacologically, ketamine is thought to act broadly in the brain, most prominently as a non-competitive antagonist of the N-methyl-D-aspartate receptor (NMDAR)but also as a blocker of hyperpolarization-activated cyclic nucleotide (HCN1)channels and as a potential activator of opioid receptors. Recent studies in rodent models are unraveling the cellular and neural circuit underpinnings of ketamine's complex action in the brain. For example, an antidepressant dose of ketamine was shown to promote spinogenesis and synaptogenesis in prefrontal cortical circuits to rescue the eliminated spines in a depression mouse model. Another recent study showed that sub-hypnotic doses (50 and 100 mg/kg) of ketamine switches the spontaneous excitatory activity across the neocortex by suppressing the active neurons while activating the previously silent neurons, paralleling its dissociative property. Acute ketamine administration also broadly impacts the dopaminergic modulatory system (via NMDAR antagonism), resulting in increased firing in the ventral tegmental area (VTA) dopamine (DA) neurons and enhanced DA release in the frontal cortex, striatum and nucleus accumbens(Fig.). In contrast, the long-term impact of chronic ketamine exposure on brain networks remains much less understood, with profound scientific and clinical implications. The antidepressant effect of ketamine is known to be transient, especially in treatment-resistant depression patients, thus often requiring maintenance treatments over years. Additionally, the long-term recreational abuse has been associated with cognitive and sensory impairmentsand significant damages have been reported in the frontal, parietal, and occipital cortices in the brain. Recent studies in mice have further revealed significant alterations in neocortical microcircuit synchrony after repeated exposure to ketamine. Therefore, with its broad clinical importance and increasing long-term abuse potential at higher doses, there is a considerable interest in understanding the molecular, cellular, and neural circuit alterations caused by long-term exposure to ketamine over a wide range of doses. We sought to systematically investigate the brain-wide impact of chronic (R,S)-ketamine exposure on the entire dopaminergic system in mice. By utilizing a range of sub-hypnoticdoses (30 and 100 mg/kg) and high-resolution whole-brain phenotyping of DA neurons, we show that chronic ketamine exposure over time results in divergent brain-wide changes in the DA neuron populations and their long-range projections to the prefrontal cortex and sensory areas. Further, we reveal the role of post-transcriptional regulation mechanisms in modulating the ketamineinduced structural plasticity in the DA system. Overall, through an unbiased whole-brain highresolution mapping, we reveal the broad non-monotonic impact of ketamine on the brain-wide DA modulatory system.
HIGH-RESOLUTION WHOLE-BRAIN PHENOTYPING OF KETAMINE-TREATED ANIMALS
We established a complete pipeline for whole-brain labeling, high-resolution imaging and comparative phenotyping of the entire dopaminergic modulatory system after 1, 5 and 10 days of daily (R,S)-ketamine (30 and 100 mg/kg) and saline control intraperitoneal (i.p.) injections (Fig.). The cellular toxicity of the ketamine exposure was assessed by α-activated caspase antibody staining after 10 days of 100 mg/kg daily i.p. injections, revealing no significant cell death in the brain (Supplementary Fig.). The locomotion of injected animals was video-recorded and quantified at 15' and 60' post injections (Supplementary Fig.). The 30 mg/kg group, but not 100 mg/kg, exhibited increasing (with days of exposure) locomotion sensitivity 15'-post injection. The extracted intact brains were stained with α-tyrosine hydroxylase (TH; rate-limiting enzyme for DA synthesis), which is a widely used marker of DA neurons 27 (Fig., Supplementary Video 1). Note that even though DA may further get converted to other catecholamines (norepinephrine and/or epinephrine) in the downstream pathways, the distribution of these noradrenergic neurons is very well characterized and known to be localized within the hindbrain regions (pons and medulla). Nevertheless, due to their brain-wide projections (e.g., to cerebral cortex, hippocampus, amygdala and hypothalamus), the precise identity of the TH+ neuronal projections (as DA or noradrenergic) may only be inferred as catecholaminergic. For investigating the TH mRNA expression, we established and used a whole-brain staining method based on hybridization chain reaction, and also utilized a well-characterized inducible Cre line, TH-CreER 32 , crossed with a tdTomato reporter line (Supplementary Fig.). The intact brain samples were cleared with either iDISCO+or passive CLARITY 34 methods, and imaged at high-resolution with COLMor light sheet theta microscopy (LSTM 35 ) (Fig.). Finally, we developed a set of accurate large data analysis methods (suiteWB, Fig.) for high-resolution phenotyping of the entire dopaminergic system -both at the levels of TH+ cell bodies as well as TH+ brain-wide projections (Fig.). To this end, we first generated a local average reference (from 7 brains), which was annotated by the registration of Allen brain atlas (ABA, ccfv3) annotations. In addition, we developed a multi-model image segmentation approach to accurately detect the TH+ cell bodies and their brain-wide projections (Fig., Supplementary Video 3). ANOVA and two-sided Mann Whitney U tests with Bonferroni correction were utilized for statistical comparisons. Overall, suiteWB methods allow accurate high-resolution phenotyping of the brain-wide structural plasticity.
DOSAGE-DEPENDENT DIVERGENT IMPACT OF CHRONIC KETAMINE EXPOSURE ON THE DA DOMAINS
We generated high-resolution whole-brain maps of TH+ neurons after 1, 5, and 10 days of daily ketamine and saline i.p. injections (Fig., Supplementary Video 3). Using suiteWB, TH+ neuron counts were calculated across all the brain regions (Fig.) and were statistically compared across treatment groups at multiple scales by utilizing different graph-cut levels of the hierarchical ABA annotation tree. Robust, statistically significant alterations were only detected after 10 days of ketamine exposure for both 30 and 100 mg/kg ketamine treatment groups, therefore, 1-and 5days treatment datasets were not analyzed further. In addition, as expected, 100 mg/kg treatment group exhibited much more changes than the 30 mg/kg treatment group (Fig.). At a higher-level annotation (i.e., larger brain regions ROIs; level 6), we found an overall dose-dependent statistically significant decrease in TH+ neuron counts within the behavior-state related mid-brain regions (MBsta) and, conversely, an overall increase within the hypothalamic lateral zone (LZ) in both the 30 and 100 mg/kg (10 days) treatment groups (Fig.). In addition, for the 100 mg/kg group, we observed a tendency for increase within the hypothalamic periventricular zone (PVZ) and, for the 30 mg/kg group, significant decrease within the periventricular region (PVR) (Fig.). Next, we compared the brains at lower-level (level 8) graph cut of the ABA annotation tree. For both 30 and 100 mg/kg treatment groups, we observed a robust decrease in the dorsal raphe (DR) and increase in the lateral hypothalamic region zona incerta (ZI). In addition, in the 100 mg/kg treatment group, we observed a robust decrease in the reticular nucleus retrorubral area (RR) and increases in the arcuate hypothalamic nucleus (ARH), the periventricular hypothalamic nucleus posterior part (PVp), tuberomammillary nucleus (TM) and the periventricular hypothalamic nucleus descending division (PVHd) (Fig.). Whereas, in the 30 mg/kg treatment group, we found significant increases in mid-brain region SNr (Substantia nigra reticular part) and a significant decrease in the dorsomedial nucleus of the hypothalamus (DMH) and the medial hypothalamic region MPN 41 (preoptic nucleus). These brain-wide alterations were further validated by an independent voxels-based (20x20x20 µm 3 sampling) brain parcellation approach (Fig.). Representative example volume renderings are shown for RR, DR, ARH, PVp and PVHd (Fig., 100 mg/kg ketamine-treated group). Overall, an unbiased whole-brain comparison of TH+ neuron counts revealed divergent and dosage-dependent brain-wide impact of chronic ketamine exposure (Fig.) -increases within multiple hypothalamic domains (e.g., ARH, containing TH+ DA neurons with orexigenic function in energy homeostasis) and decreases within the behavioral state related midbrain regions (e.g., DR, containing TH+ DA neuron which modulate the social isolation behaviors 37 ; RR, containing TH+ DA neurons with role in fear and aversive signaling).
UNTRANSLATED TH MRNA+ NEURONS FACILITATE KETAMINE-INDUCED CELLULAR PLASTICITY
We sought to further investigate the mechanistic basis of the chronic ketamine-induced brain-wide cellular plasticity. Utilizing our whole-brain mRNA labeling protocol, we first mapped the expression of TH mRNA (Fig., Supplementary Video 4), revealing a much broader distribution than the corresponding TH protein. Such discrepancy in the TH mRNA/protein expression (Supplementary Video 5) suggests a potential role for post-transcriptional mechanisms in rapid modulation of the brain-wide DA system. To test this hypothesis further, we utilized a well-characterized inducible TH-CreER transgenic line 32 (Supplementary Fig.), crossed with a tdTomato reporter line, to investigate if the newly acquired TH+ neurons originated from the untranslated TH mRNA+ neurons. TH-CreER induction (by 4-OHT i.p. injections) was performed 1-week prior to the start of the ketamine exposure, thus, permanently labeling the pretreatment TH mRNA+ neurons by tdTomato. After 10 days of chronic ketamine (100 mg/kg) exposure, the brains were harvested, cleared with passive CLARITY, and labeled with α-TH antibody for a direct comparison of the after-treatment TH+ neurons with before-treatment TH mRNA+ neurons, within the exact same brain samples. As shown in Fig., we found a significant increase in TH protein+/tdTomato+ co-labeled neurons (α-TH ∩ tdTomato) in the ketaminetreated group (compared to saline controls), while the overall number of tdTomato+ neurons remained unchanged, consistent with unchanged TH mRNA+ neurons (Fig.). Conversely, we observed a decrease in TH protein+/tdTomato+ co-labeled neurons in the midbrain regions (which showed reduction in TH+ DA neuron counts), while the tdTomato+ and TH mRNA+ neuron counts remain unchanged compared to the saline controls (Fig.). Altogether, these results suggest that the cellular plasticity in the dopaminergic system may be facilitated by the existence of a much larger pools of untranslated TH mRNA+ neurons to rapidly modulate the number of available TH+ DA neurons in various regions of the brain.
ALTERED LONG-RANGE TH+ PROJECTIONS AFTER CHRONIC KETAMINE EXPOSURE
Taking advantage of the high-resolution of our datasets, we sought to map the brain-wide changes in TH+ neuronal projections. Note that TH+ neuronal projections may be DA or noradrenergic, therefore they can only be inferred as catecholaminergic. Using the suiteWB pipeline, we estimated the projection densities in the ketamine (100 mg/kg, 10 days) treated and saline control whole-brain datasets and performed a voxel-by-voxel (at 25x25x25 µm 3 sampling) statistical comparisons across groups. As shown in Fig., Supplementary Video 6, chronic ketamine exposure resulted in robust brain-wide changes in TH+ projections. We observed increased densities in multiple associative cortical regions including the prefrontal cortex (PFC)related prelimbic area (PL, Supplementary Video 7, Fig.), orbital area (ORB, Supplementary Video 8, Fig.), frontal pole (FRP, Fig.) and anterior cingulate area (ACA, Fig.), and posterior parietal association area (PTLp, Fig.). Further, the lateral amygdala (LA, Supplementary Video 9, Fig.), which is crucial for processing of threatening stimuli and fear behavior, specific regions of the lateral septal complex (septofimbrial nucleus (SF), involved in reward and reinforcement) and the triangular nucleus of septum (TRS) also showed increased projection densities (Fig.). Furthermore, brain regions receiving olfactory inputs, including the taenia tecta (TT) and piriform area (PIR), and regions of the thalamus (lateral posterior nucleus of the thalamus (LP)) and hypothalamus (VMH and TU) also showed increases in the TH+ neuronal innervations. In contrast, the dorsal auditory area (AUDd, Fig., Supplementary Video 10), posterolateral visual area (VISpl, Fig., Supplementary Video 11), entorhinal area (ENT) and presubiculum (PRE, involved in spatial information processing; Fig.) exhibited decreased TH+ projection densities. Striatum sub-regions caudoputamen (Cp) and bed nuclei of the stria terminalis (BST/BNST, Fig., Supplementary Video 12) showed reductions consistent with the loss of TH+ neurons in DR, RR and SNc (Fig.). Finally, thalamic regions subparafascicular area (SPA) and subparafascicular nucleus (SPF), and midbrain regions nucleus sagulum (SAG) and pedunculopontine nucleus (PPN) also showed decreased projections. Overall, consistent with the divergent changes in TH+ neuron counts in the midbrain and hypothalamic dopaminergic domains (Fig.), chronic ketamine exposure resulted in increased TH+ neuronal projection densities in the associative brain centers, including PFC-related regions, and decreased innervations of visual, auditory, and spatial information processing regions (Fig.).
DISCUSSION
Utilizing an unbiased, high-resolution whole-brain mapping approach, we revealed hitherto unknown divergence in the adaptability of the DA system to repeated ketamine exposure (summarized in Fig.) -reduced number of TH+ neurons within the behavior state-related midbrain domains and increased within the hypothalamic domains, along with altered long-range innervation of the association and sensory areas by TH+ neuronal projections. Such structural plasticity of brain-wide modulatory system may facilitate significant reconfiguration of the neuronal networks in response to external stresses such as ketamine exposure or diseased conditions (e.g., schizophrenia), to eventually result in long-lasting cognitive behavioral changes. Note that, DA may further get processed in the noradrenergic neurons, but their cellular distribution is well known to be restricted to specific domains in hindbrain regions. However, the TH+ neuronal projections may still not be precisely identified as DA or noradrenergic by TH immunolabeling. In addition, the loss of TH protein in DA neurons may not result in a permanent loss of DA neurons, nevertheless, it does indicate the loss of DA producing capacity due to the required role of TH in the L-DOPA (DA precursor) biosynthesis. We also show that the adaptability of the DA system is facilitated by a large pool of neurons that stably maintain translationally suppressed TH mRNA (Fig.). These findings were validated by a direct comparison of the newly recruited or lost TH+ DA neurons with the before-exposure TH mRNA expressing neuron populations by utilizing an inducible TH-CreER transgenic mice line (Fig.). Mechanistically, these observed changes in DA system may, in part, be the result of repeated ketamine-induced activation/deactivation (indirect, via NMDAR antagonism) of the DA neurons due to the heterogeneous neuronal inputs. Intriguingly, the ketamine-induced cellular plasticity is distinct from the previously reported neurotransmitter phenotypic plasticity of the hypothalamic DA neurons after changed day/night durations in rodents, and included loss of TH mRNA expression. Overall, our data suggest that the maintenance of translationally suppressed mRNA, even though energetically costly process, may allow for faster brain-wide adaptations to various external stresses. The development and use of unbiased, high-resolution whole-brain phenotyping of the entire DA system allowed discovery of the divergent impact of chronic ketamine exposure. Such non-monolithic brain-wide impact further underscores the need for unbiased investigations of on/off-target effects of ketamine treatment at a range of doses, as well as the urgency to develop targeted pharmacological intervention approaches (e.g., focused ultra sound-based approaches) for treatment of complex brain disorders. All datasets will be made available on request.
ANIMALS.
Male TH-2a-CreER 32 mice were acquired from Dr. David Ginty's lab and were bred with B6;129S6-Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze /J (Ai14; JAX Strain #:007908). All mice were group-housed in a 12:12 light:dark cycle at 22ºC. Food and water were provided ad libitum. (R,S)ketamine i.p. injections were performed during the light phase. All experimental procedures were approved by the IACUC at Columbia University. Locomotor activity was recorded after 1, 5, and 10 days of treatment. The mice were placed within a novel home cage and were recorded (camera: GoHZQ, 1920 x 1080 pixels,30 fps transmission rate) 15 minutes and 1 hour post-intraperitoneal ketamine injections. The total distance traveled was quantified using the ANY-maze tracking software (ANY-maze, RRID:SCR_014289, Stoelting, Wood Dale, IL, United States). Drugs. 4-hydroxytamoxifen (4-OHT; Sigma, H7904) was dissolved in corn oil/ethanol (90% corn oil, 10% ethanol) via the use of a vortex, ultrasonication, and 55°C heating for <15 min. Mice were intraperitoneally injected with 2 mg of 4-OHT. (R,S)-ketamine (Covetrus) was used for all ketamine exposure experiments. 3 and 10 mg/mL stocks were prepared in saline (0.9% NaCl). One dose of 30 mg/kg or 100 mg/kg of (R,S)-ketamine was intraperitoneally injected in a 24 hour period for 1 day, 5 days, or 10 days treated animals. 8-10 weeks old mice were used for all experiments. Whole brain clearing and labelling. For iDISCO clearing, the iDISCO+ protocol was followed as previously described. The brains were pre-treated with methanol, placed in 66% dichloromethane/ 33% methanol overnight, bleached with 5% H2O2/methanol, and then rehydrated. Afterwards, the whole brains were permeabilized for 2 days at 37°C in 1x phosphate buffered saline (PBS), 0.2% Triton X-100, 0.3 M glycine, and 20% dimethyl sulfoxide (DMSO). The brains were blocked with 1xPBS, 0.2% Triton X-100, 10% DMSO, 6% donkey serum for 2 days at 37°C, followed by incubation in 1xPBS, 0.1% Triton X-100, 3% donkey serum, and a 1:200-1:500 dilution of the primary antibody sheep α-TH (ab113, Abcam) for 10 days. After washing in 1xPBS/0.1% Triton X-100, the brains were placed in the secondary antibody solution containing 1xPBS/0.1% Triton X-100/3% donkey serum and a 1:1000 dilution of donkey antisheep 647 (A-21448, Thermofisher) for 10 days. We used the passive CLARITY method as detailed previously. The tissue was first encapsulated in a hydrogel monomer (HM) solution consisting of 1% (wt/vol) acrylamide, 0.05% (wt/vol) bisacrylamide, 4% paraformaldehyde, 1x phosphate buffered saline (1xPBS), deionized water, and 0.25% of thermal initiator (VA-044, Fisher Scientific), followed by clearing with SBC buffer (4% (wt/vol) SDS, 0.2 M boric acid, pH 8.5 and deionized water) at 37°C with shaking. The SBC buffer was replaced every 2 days. After clearing, the SBC buffer was washed off with 0.2 M boric acid pH 8.5 with 0.1% Triton X-100. The cleared tissue was immunostained in 0.2 M boric acid pH 7.5 with 0.1% Triton X-100. The final refractive index matching was performed in RapiClear (SunJin lab, RI=1.47). For HCR-FISH, we used split-initiator DNA probes (IDT) for detecting the tyrosine hydroxylase (Th, NM_012740.3) mRNA. During probe hybridization/detection, the brains underwent 1) equilibration with 5xSSCTw buffer, 2) acetylation with 0.25% acetic anhydride solution, 3) equilibration with probe hybridization buffer (30% formamide, 5xSSC, 0.5 mg/mL yeast tRNA, 10% dextran sulfate), and 4) probe incubation (50 nM) in probe hybridization buffer at 37C. The time/amount of solution varied depending on the thickness of the brain slices. The slices were then washed in probe wash buffer (30% formamide, 5X SSC, 9 mM citric acid, and 0.1% tween 20) at 37ºC as well as two rounds of washes in 5xSSCTw. Afterwards, the slices were equilibrated in amplification buffer (5xSSC, 10% dextran sulfate, and 0.1% Tween-20), and then incubated in amplification buffer with 50-150 nM of hairpins H1 and H2 conjugated with AF-647 (Molecular Instruments). The hairpin/amplification buffer mixture was then washed off with several rounds of 5xSSCTw. For activated caspase staining, the TH-2a-CreER;Ai14 brains were sliced with a vibratome (Leica VT1000 S Vibrating blade microtome) into 50 µm sections. These sections were placed in a blocking buffer (1xPBS/0.1% bovine serum albumin/0.1% Triton X-100) for 30 minutes and then incubated in a 1:200-1:400 dilution of the primary antibody, rabbit anti-active caspase-3 (BD Biosciences, catalog no. 559565), with 1xPBS/0.5% bovine serum albumin /0.1% Triton X-100 overnight. After washing in 1xPBS/0.1% Triton X-100, the sections were incubated, for two hours, in a 1:500 dilution of the secondary antibody, goat anti-rabbit 647 (ThermoFisher, Catalog #A-21245) with 1xPBS/0.5% bovine serum albumin /0.1% Triton X-100. The secondary antibody was washed off in 1xPBS/0.1% Triton X-100, and the slices were incubated in a 1:1000 dilution of a 1 mg/mL stock of DAPI with 1xPBS for 15 minutes. After washing off with 1xPBS/0.1% Triton X-100, the slices were mounted in 65% glycerol.
IMAGING.
All imaging experiments were performed with CLARITY-optimized light sheet microscopy (COLM)or ClearScope (MBF Biosciences)Olympus 10x/0.6NA/8 mmWD or ASI 16x/12mmWD detection objectives were used for most of the whole brain imaging with COLM. Olympus Macro 4×/0.28NA or Nikon 20x/1.0NA were used for ClearScope imaging., however, some of them are optimized for confocal images with smaller data sizes, others utilize deep learning models with pre-trained parameters that do not accurately generalize to images with different signal quality distribution and also require dense annotation training datasets. We chose to develop a semi-supervised multi-model learning approach to address the challenges of whole brain segmentation. This approach does not require dense annotated training and also allows the flexibility of using different optimal parameter sets for different regions (i.e. with dense or sparse object densities). We utilized open-source toolkit ilastik 51 for semi-supervised learning of classifiers based on image features. The high-resolution registered whole brain images were split into densely and sparsely populated regions by visual inspection, followed by pixel classification to generate probability maps. For the sparse region, a standard pixel classification workflow in ilastik was applied to generate cell probability maps. First, image features were extracted by using multiple types of filters with different kernel sizes, including intensity filters (Gaussian Smoothing), edge detection filters (Laplacian of Gaussian, Gaussian Gradient Magnitude, Difference of Gaussians), and texture detection filters (Structure Tensor Eigenvalues, Hessian of Gaussian Eigenvalues). The image features were then used to estimate the probabilities of pixels belonging to a cell or background. For denser regions, we utilized the Autocontext workflow in ilastik. In the first step, the image pixels were classified into five categories: empty space, brain background, fibers (neuronal projections), cell cytoplasm and cell nucleus. In the second step, the classification results were combined with image features (with filters described above for sparse regions) to generate probability maps of cells against background. Next, we applied blob detection on the dense and sparse probability maps separately to detect cells. This was achieved by using the difference of Gaussian (DOG) algorithm (Python scikit-image package 52 . To optimize memory usage, images were processed block-wise with carefully resolved boundary conditions. Cells detected in dense and sparse regions were merged and were further tuned by using Napari viewer. F1 for sparse and dense region segmentations reached 0.95 and 0.86, with precision 0.889 and 0.976 respectively, compared to human-annotated datasets (Fig.). For comparison, ClearMap 48 , applied to immediate early gene datasets, reported a precision of 0.83 and 0.75 compared to two human annotators. For colocalization analysis of the multi-channel (i.e. tdTomato and α-TH immunostaining signal) CLARITY images, cell detection (as described above) was applied separately to different channels. The full cell body labels were generated by taking the union of the detected cells in the two channels that were less than 25µm apart. The cell identities (i.e. if expressing tdTomato or α-TH or both) were assigned by manually validated thresholding of probability maps of each channel. Neuronal projection segmentation was performed by semi-supervised pixel classification by using image features (as discussed above) to generate the probability maps distinguishing projections and background. The probability threshold for binarization was determined by careful inspection throughout various regions of the brain. The binarized data was then convolved with 3D Gaussian kernels of 200 µm and 500 µm. The ratio of the two gaussians resulted in locally normalized neuronal projection densities. Statistical comparisons between Ketamine and Saline treated samples were done with twosided Mann Whitney U test (using Python package Scipy 53 ) on normalized cell counts (cell count in a region divided by total cell count of that whole brain) and cell densities (estimated by 3D Gaussian kernel of size 50 µm in each dimension).