Abstract
Dopamine (DA) is a central monoamine neurotransmitter involved in many physiological and pathological processes. A longstanding yet largely unmet goal is to measure DA changes reliably and specifically with high spatiotemporal precision, particularly in animals executing complex behaviors. Here we report the development of novel genetically-encoded GPCR-Activation-Based-DA (GRABDA) sensors that enable these measurements. In response to extracellular DA rises, GRABDA sensors exhibit large fluorescence increases (ΔF/F0∼90%) with sub-second kinetics, nanomolar to sub-micromolar affinities, and excellent molecular specificity. Importantly, GRABDA sensors can resolve a single-electrical-stimulus evoked DA release in mouse brain slices, and detect endogenous DA release in the intact brains of flies, fish, and mice. In freely-behaving mice, GRABDA sensors readily report optogenetically-elicited nigrostriatal DA release and depict dynamic mesoaccumbens DA changes during Pavlovian conditioning or during sexual behaviors. Thus, GRABDA sensors enable spatiotemporal precise measurements of DA dynamics in a variety of model organisms while exhibiting complex behaviors.
Introduction
Dopamine is a crucial monoamine neurotransmitter in both vertebrates and invertebrates. In the vertebrate central nervous system, DA regulates a wide range of complex processes, including reward signaling (Schultz, 2016; Wise, 2004), reinforcement learning (Holroyd and Coles, 2002), attention (Nieoullon, 2002), motor control (Graybiel et al., 1994), arousal (Kume et al., 2005; Wisor et al., 2001), and stress (Abercrombie et al., 1989). In the human brain, impaired DA transmission is associated with neuropsychiatric disorders and neurodegenerative diseases, including attention deficit hyperactivity disorder (Cook Jr et al., 1995), schizophrenia (Howes and Kapur, 2009) and Parkinson’s disease (Lotharius and Brundin, 2002). Moreover, psychostimulants, such as cocaine and amphetamine, exert their addictive effects by targeting components in the DA signaling pathway and by altering extracellular DA levels (Di Chiara and Imperato, 1988; Giros et al., 1996; Hernandez and Hoebel, 1988; Ritz et al., 1987).
Despite many important roles that DA plays in both physiological and pathological processes, precise measurements of the spatial and temporal patterns of DA release during complex behaviors are lacking. This gap in understanding due in large part to the limitation associated with existing methods for the real-time detection of endogenous DA release in the intact brain. Historically, intracerebral microdialysis has served as the gold standard for quantitative measurements of extracellular DA concentration. However, the relative slow sampling rate afforded by microdialysis is not well suited to detect dynamic changes in DA levels during complex and rapidly evolving behaviors, such as those that characterize social affiliation (Tidey and Miczek, 1996), aggression (van Erp and Miczek, 2000), and mating (Pfaus et al., 1990a, b). Fast-scan cyclic voltammetry (FSCV) is a temperamental electrochemical method that can measure changes in extracellular DA concentrations with 10-ms temporal resolution (Robinson et al., 2008; Rodeberg et al., 2017). However, because FSCV requires oxidization of DA molecules, it is difficult to distinguish DA from other structurally similar transmitters, such as norepinephrine (NE) (Robinson et al., 2003). Moreover, both microdialysis and FSCV require implantation of a relatively large probe or electrode (approximately 70-300 μm in diameter) into a specific brain region, which precludes the ability to obtain spatially precise measurements of endogenous DA release (Jaquins-Gerstl and Michael, 2015).
In lieu of direct measurements of extracellular DA, indirect methods, for example, measuring the activity of dopaminergic neurons (DANs) or the activation of DA receptor downstream targets, have also been used to approximate the dynamics of DA release. Genetic expression of a presynaptic tethered Ca2+ indicator in DANs has been used to indicate potential compartmentalized DA signals in the Drosophila olfactory pathway (Cohn et al., 2015). However, given the highly complex regulation of presynaptic Ca2+, the nonlinear relationship between intracellular Ca2+ and transmitter release, as well as the active uptake of extracellular DA by transporters, it is difficult to quantitatively translate Ca2+ signals precisely into extracellular DA levels. Cell-based DA reporters, such as CNiFERs (Muller et al., 2014), use transplated HEK293 cells constitutively expressing DA receptors together with an intracellular Ca2+ indicator to couple extracellular DA signals with the fluorescence increase. However, This approach requires cell transplantation, which may limit the broad usage of this method. Moreover, CNiFERs are not spatially sensitive to synaptically released DA and thus may be limited to measurements of volumetric transmission of DA. Finally, the TANGO assay and their next-generation versions (Barnea et al., 2008; Inagaki et al., 2012; Kim et al., 2017; Lee et al., 2017) have been used to measure endogenous DA release by coupling the β-arrestin signaling pathway to reporter gene expression. Although this approach enables the cell-type specific expression of the DA reporter and is suitable for in vivo measurements, the long signal amplification time (on the order of hours) required for this assay precludes the ability to monitor rapid, physiologically relevant dynamics of DA signaling in real time.
Here, we report the development of novel, genetically-encoded sensors that enable direct, rapid, sensitive, and cell-type specific detection of extracellular DA. These sensors, which we call GRABDA sensors, were engineered by coupling a conformationally sensitive circular-permutated EGFP (cpEGFP) to a selected human DA receptor. Through iterative engineering and optimization, we yielded two GRABDA sensors: GRABDA1m, with medium DA affinity (EC50 ∼ 130 nM); and GRABDA1h, with high DA affinity (EC50 ∼ 10 nM). We show that these two newly developed GRABDA sensors enable real-time detection of endogenous DA in acute brain slices of mice and in the intact brains of versatile animal models including flies, fish, and mice.
Results
Development and initial characterization of GRABDA sensors
To develop a genetically encoded sensor for DA, we sought to engineer naturally evolved DA receptors to couple a conformationally sensitive cpEGFP. We hypothesized that upon DA binding, the conformational changes in the receptor could alter the arrangement of the associated cpEGFP, resulting in a DA-dependent change in fluorescence. Indeed, a similar strategy was recently applied in creating the genetically encoded acetylcholine sensor GACh (Miao Jing et al., 2018), suggesting that this strategy could potentially be expanded to generate DA sensors.
We used a three-step approach to engineer and optimize GRABDA sensors (Fig. 1A-C): First, as the third intracellular loop (ICL3) of the G protein-coupled receptor (GPCR) links transmembrane V and VI that are thought to undergo large conformational changes upon ligand binding (Kruse et al., 2013), a cpEGFP was first inserted into the ICL3 of all five subtypes of human dopamine receptors (DRs); we subsequently focused on the D2R-cpEGFP chimera due to its superior membrane trafficking and relatively high affinity for DA (Beaulieu and Gainetdinov, 2011; Missale et al., 1998) (Fig. S1A,B). In the second step, the position of the cpEGFP insertion within the ICL3, and individual amino-acids in the linker region between cpEGFP and D2R were systematically screened (Fig. 1A,B). Finally, mutations at residues critical for receptors affinity for DA (Sung et al., 2016) were further introduced to expand the sensor’s response range (Fig. 1C). After screening a total of more than 432 variants, we chose two, GRABDA1m and GRABDA1h, both of which have a ∼90% maximal ΔF/F0 response to the application of saturating levels of DA (Fig. 1B,E), but differ by an order of magnitude with respect to affinity for DA (130 nM for GRABDA1m and 10 nM for GRABDA1h) (Fig. 1C). We also generated a corresponding mutant variant of each sensor (GRABDA1m-mut and GRABDA1h-mut) by introducing C118A and S193N double mutations in the receptor’s putative DA-binding pocket in order to abolish the DA binding (Chien et al., 2010b; Wang et al., 2018) (Fig. S1D). These so-called “dead” mutants exhibited no DA-induced change in fluorescence compared with GRABDA1m and GRABDA1h (Fig. 1D).
We next characterized the properties of the two GRABDA sensors in detail in cultured HEK293T cells. Both GRABDA1m and GRABDA1h trafficked efficiently to the plasma membrane of HEK293T cells (Fig. 1D and Fig. S1C). Cells expressing GRABDA1m and GRABDA1h exhibited robust fluorescence increases upon bath application of DA (Fig. 1D-F). Notably, the DA-induced fluorescence increase was completely blocked by the co-application of the D2R antagonist haloperidol (Halo) (Sokoloff et al., 1990), confirming the molecular specificity of these sensors (Fig. 1E). Both mutant forms of GRABDA sensors trafficked normally to the plasma membrane, similar to GRABDA1m and GRABDA1h (Fig. 1D and Fig. S1C), albeit with no detectable fluorescence increase in response to DA application (Fig. 1D,E). To characterize the response kinetics of GRABDA1m and GRABDA1h, agonists or antagonists at high concentrations were locally applied to HEK293T cells expressing these sensors via a rapid perfusion system (Fig. 1F,G). Both GRABDA1m and GRABDA1h showed rapid fluorescence increases (on rate) in response to DA application, with a time constant of ∼100 ms (60 ± 10 ms for GRABDA1m and 140 ± 20 ms for GRABDA1h, respectively), implying that GRABDA sensors are suitable for tracking rapid DA dynamics. The fluorescence decrease (off rate) of the GRABDA sensors in response to applications of the antagonist Halo is slower in GRABDA1h (2.5 ± 0.3 s) compared with GRABDA1m (0.7 ± 0.06 s), consistent with the differences in affinity (Fig. 1G). We also measured the photostability of GRABDA1m and GRABDA1h in HEK293T cells. Under confocal laser illumination, both GRABDA sensors were significantly more photostable than a membrane targeted EGFP or the glutamate sensor iGluSnFR (Marvin et al., 2013) (Fig. S2), suggesting that these new DA sensors are suitable for long-term measurements of DA dynamics.
We also tested the specificity of GRABDA sensors for DA comparing with other neurotransmitters. HEK293T cells expressing GRABDA sensors were sequentially perfused with solutions containing different neurotransmitters (Fig. 1H and Fig. S3; all were applied at 1 μM concentration). Again, bath application of DA induced robust fluorescence increases in both GRABDA1m-and GRABDA1h-expressing cells, and these signals were completely blocked by co-application of D2R antagonists Halo or eticlopride (Etic), but not by the D1R antagonist SCH-23390 (SCH). In contrast, applications of various other neurotransmitters, including serotonin (5-HT), histamine (His), glutamate (Glu), GABA, adenosine (Ade), acetylcholine (ACh), tyramine (Tyr), octopamine (Oct), glycine (Gly), or the DA precursor L-DOPA, did not elicit any detectable fluorescence changes (Fig. 1H and Fig. S3). We found that NE application evoked a small yet significant increase in fluorescence in both GRABDA1m- and GRABDA1h-expressing cells, although this increase was considerable smaller than the response elicited by DA applied at the same concentration (for GRABDA1m, the ΔF/F0 in response to NE is 26% of that to DA). Further characterization of the affinity of GRABDA1m or GRABDA1h to both DA and NE revealed a 10-fold higher affinity to DA against NE (Fig. 1H and Fig. S3), as expected from the specificity of native human D2R to DA and NE (Lanau et al., 1997b). Overall, GRABDA1m and GRABDA1h sensors show rapid and highly sensitive responses to physiological ranges of DA with little or no sensitivity to almost all other neurotransmitters tested here.
Does ectopic expression of GRABDA sensors elicit unintended signaling by coupling to GPCR’s downstream pathways? To examine this, we separately tested the coupling efficacies of GRABDA sensors to either G protein-or β-arrestin-dependent pathways that are known to be downstream of activated D2R (Beaulieu and Gainetdinov, 2011). It is well established that the coupling of a G protein with its cognate GPCR significantly increases the receptor’s binding affinity for its ligand (Kruse et al., 2013). We therefore used pertussis toxin (PTX), which potently and selectively blocks the coupling between Gi proteins and GPCRs, such as D2R (Burns, 1988). In cells expressing either GRABDA1m or GRABDA1h, co-expression of PTX did not significantly alter the sensors’ affinity for DA, indicating that GRABDA sensors do not couple extensively to Gi proteins (Fig. 1I). On the other hand, if the DA-binding of GRABDA sensors activates the β-arrestin-dependent pathway, the resulting internalization of sensors would reduce the fluorescence at the plasma membrane of HEK293T cells (Luttrell and Lefkowitz, 2002). However, we observed stable membrane fluorescence of the GRABDA sensor-expressing cells throughout a 2-hour exposure to DA at a saturation concentration (100 μM), suggesting no detectable activation of β-arrestin-dependent signaling under these conditions (Fig. 1J). Collectively, these data suggest that GRABDA sensors do not engage these major downstream GPCR-mediated signaling pathways.
Characterization of GRABDA in cultured neurons
We next evaluated the expression pattern and functional properties of GRABDA sensors in cultured rat cortical neurons. After 48h expression, both GRABDA1m and GRABDA1h showed basal fluorescence signals in transfected neurons, with the brightest signal in the plasma membrane of somas and neurites, indicated by the colocalization with a membrane marker RFP-CAAX (Hancock et al., 1991) (Fig. 2A,B and Fig. S4). In addition, co-expression of the GRABDA sensors with the postsynaptic (PSD95-RFP) or presynaptic (SYP-RFP) markers revealed that GRABDA sensors were distributed throughout the entire membrane in dendritic shafts, spines, axons and axonal boutons, suggesting the suitability of GRABDA sensors to detect DA dynamics in sub-neuronal compartments (Fig. 2B and Fig. S4). As in HEK293T cells, DA application in transfected cultured neurons induced dose-dependent fluorescence increases in both GRABDA1m-and GRABDA1h-expressing neurons with a similar maximum ΔF/F0 of ∼90%, and with an EC50 of ∼170 nM and ∼8 nM for GRABDA1m and GRABDA1h, respectively (Fig. 2C-E). Moreover, the specificity of both GRABDA sensors for DA was similar in cultured neurons compared to HEK 293T cells (Fig. 2F). Finally, no detectable decrease of surface fluorescence signals of GRABDA sensors-expressing cells was observed in response to a 2-h continuous application of DA at a saturation concentration (100 μM) (Fig. 2G), suggesting that GRABDA sensors were not internalized over this time frame. Collectively, these data demonstrate that GRABDA sensors can monitor DA signals with high sensitivity and specificity in cultured rat cortical neurons.
Characterization of GRABDA in acute brain slices
A crucial issue is whether GRABDA sensors are sufficiently sensitive to detect and monitor endogenous DA release in native brain tissue. To begin to resolve this issue, we injected AAVs carrying either GRABDA1m or GRABDA1h into the nucleus accumbens (NAc) of mice, then prepared acute brain slices containing NAc two weeks later. Post hoc visualization in fixed brain slices revealed that the fluorescent signal of GRABDA sensors could be detected in the NAc region of virus-injected mice, and the signal was in close proximity to putative DA-releasing fibers that were immunoreactive for tyrosine hydroxylase (TH) (Fig. 3A). Electrical stimulation within NAc core elicited robust and time-locked fluorescence increases in both GRABDA1m-and GRABDA1h-expressing neurons (Fig. 3A-C). Increasing the number of stimulation pulses or stimulation frequency resulted in a progressive increase in the intensity of evoked fluorescence signals in neurons expressing GRABDA1m or GRABDA1h (Fig. 3B,C). These stimulus-evoked fluorescent signals reached plateau levels at high stimulation frequency with large stimulation number (e.g. more than 50 pulses) in GRABDA1h expressing neurons, presumably due to saturation of the sensor arising from GRABDA1h’s high affinity for DA. The rise times of stimulus-evoked fluorescence signals were fast in both GRABDA1m-and GRABDA1h-expressing neurons (Fig. 3D), whereas the decay time of fluorescent signals in GRABDA1h expressing neurons was slower than that in neurons expressing GRABDA1m (Fig. 3D), consistent with the DA affinity and response kinetics measured in cultured cells (see Fig. 1C,G). Repeated electrical stimuli delivered at 5-minute intervals evoked reproducible fluorescence responses, indicating the reliability and stability of GRABDA1m and GRABDA1h in reporting multiple DA releasing events (Fig. 3E). Bath application of the D2R antagonist Halo abolished the electrically stimuli-induced fluorescence responses in either GRABDA1m-or GRABDA1h-expressing neurons (Fig. 3F), verifying the sensor’s specificity when expressed in brain slices. Collectively, both GRABDA1m and GRABDA1h enable sensitive and specific detection of endogenous DA dynamics in acute mouse brain slices.
Imaging DA dynamics in Drosophila
We next examined the ability of GRABDA sensors to detect physiologically relevant DA dynamics in living animals. We selected the fly as an initial test of our sensors, as DA plays a key role in the fly brain, serving as a critical teaching signal in olfactory-associative learning (Burke et al., 2012; Heisenberg, 2003; Liu et al., 2012; Schwaerzel et al., 2003). Transgenic UAS-GRABDA1m flies were generated and first crossed with TH-GAL4 to express GRABDA1m specifically in DANs. Two-photon imaging methods were used to study odor-evoked DA signals in the mushroom body (MB) in living flies (Fig. 4A). We found that the odorant isoamyl acetate (IA) elicited a time-locked fluorescence increase in the MB, most prominently in the β’ lobes, and this IA-evoked response was blocked by Halo application. In contrast, no IA-evoked fluorescence response was observed in flies expressing GRABDA1m-mut (Fig. 4B). To further test the signal specificity, we use the C305a-GAL4 line to express GRABDA1m in Kenyon cells which receives direct input from DANs (Aso et al., 2010). Comparing wild-type (WT) control flies and TH-deficient flies (Cichewicz et al., 2017), which lack DA synthesis in the CNS. Compared with WT flies, we observed no detectable IA-evoked GRABDA1m fluorescence increase was observed in TH-deficient flies (Fig. 4B,C). We also examined whether the ectopic expression of GRABDA sensors may alter physiological properties, such as the neuronal excitability. We observed no significant difference between GRABDA sensors-expressing and non-expressing DANs or Kenyon cells with respect to odor-evoked Ca2+ signals (Fig. S5), suggesting that expression of the sensor does not alter odor-evoked responses in neurons in the fly brain.
To further characterize the sensitivity and kinetics of GRABDA1m in vivo, we electrically stimulated MB DANs while simultaneously monitoring the corresponding fluorescence signals (Fig. 4D,E). We found that GRABDA1m-expressing DANs exhibited a reproducible fluorescence increases in response to repeated trains of electrical stimuli delivered at various frequencies, and even a single stimulus was sufficient to elicit a measurable increase in fluorescence (Fig. 4D,E). In contrast, similar electrical stimulation did not evoke any detectable fluorescence increase in flies expressing GRABDA1m-mut (Fig. 4D,E). With the increase of pulse number delivered at 20 Hz, fluorescence signals of GRABDA1m increased progressively and reached a plateau of peak ΔF/F0 of ∼35% after 10 pulses (Fig. 4F-H). The fluorescence responses evoked by electrical stimulation exhibited sub-second kinetics, with on and off time constants of 0.07 ± 0.01 s and 0.39 ± 0.05 s, respectively (Fig. 4I), indicating the suitability of GRABDA1m for monitoring endogenous transient DA signals. Consistent with the odor-evoked fluorescence increase, application of Halo completely blocked the fluorescence increases elicited by electrical stimuli (Fig. 4J,K), confirming the sensor’s specificity.
The DA transporter (DAT), which is a highly conserved protein across insects and vertebrates, is responsible for the reuptake of DA from extracellular space for subsequent reuse, and is the primary target of many drugs of abuse (Bainton et al., 2000; Ritz et al., 1987). Impairing DAT function results in sustained elevation of extracellular DA levels (Fig. 4L) (Bainton et al., 2000; Ritz et al., 1987). To determine whether our sensor could detect changes in extracellular DA arising from manipulation of the DAT, we applied cocaine, a psychostimulant drug that blocks DAT. Cocaine significantly potentiated the odor-evoked fluorescence increase of GRABDA1m expressed in MB DANs, and this cocaine-potentiated response was accompanied by a prolonged decay in the fluorescence signal (Fig. 4M-O). Genetically knocking down the expression of DAT selectively in DANs phenocopied the effect of cocaine administration (Fig. 4M-O), confirming the ability of GRABDA1m to measure the dynamic regulation of DA release and reuptake in vivo. Together with our results described above, these data demonstrated that GRABDA sensors have the sensitivity, fast kinetics, and specificity to report in vivo DA dynamics in genetically-defined neurons in the intact brain of living flies.
Imaging DA release in the intact zebrafish brain
Zebrafish larvae have an optically transparent brain and is capable of performing a wide range of behaviors, affording a powerful system to explore the structure and function of the vertebrate brain at cellular resolution in a behavioral framework. To test the feasibility of using GRABDA sensors in imaging DA dynamics in zebrafish larvae, we generated the transgenic line Tg(elval3:GRABDA1m/DAT:TRPV1-TagRFP), in which GRABDA1m was expressed pan-neuronally throughout the brain, and TRPV1-TagRFP was expressed specifically in DANs to enable their chemogenetic activation by capsaicin (Fig. 5A). We first applied DA to the fish, and observed fluorescence increases in GRABDA1m-expressing neurons in the head, which were blocked by co-application of antagonist Halo, suggesting the specificity of GRABDA (Fig. 5B-D).
Next, we hope to track the dynamics of endogenous DA signals in zebrafish larvae. We previously reported that the optic tectum of zebrafish larvae receives synaptic input from pretectal DANs (Shang et al., 2015), advancing the tectum as a potential site to monitor DA release. We therefore performed confocal imaging of GRABDA1m-expressing tectal neurons in the live transgenic zebrafish larvae (Fig. 5E). Repeated application of capsaicin (five 100-ms puffs delivered with a 1-min interval) caused the progressive increase in the fluorescence signals in the tectal neuropil (Fig. 5F-H). The capsaicin-induced fluorescence increase was again abolished by the application of Halo (Fig. 5F-H), confirming that the response was specifically due to GRABDA1m activation. Thus, the GRABDA1m is well suited to report in vivo DA dynamics in the brain of zebrafish larvae.
Combining optogenetics with GRABDA to measure the dynamics of DA in freely moving mice
To test the ability of GRABDA sensors to report DA dynamics in specific circuits in the mouse brain in vivo, we focused on DANs located in the substantia nigra pars compacta (SNc) that project to the dorsal striatum (Str). This nigrostriatal pathway is implicated in complex behavioural functions including motivation, reward, and learning (Balleine et al., 2007; Da Silva et al., 2018; Howe and Dombeck, 2016). We virally expressed DIO-C1V1 (Yizhar et al., 2011) in the SNc of TH-Cre mice, permitting selective optogenetic activation of DANs. Additionally, we virally co-expressed GRABDA1m and tdTomato in the dorsal striatum, allowing us to simultaneously monitor DA release using the green (GRABDA1m) channel while detecting movement-related fluorescence artefacts using the red (tdTomato) channel (Fig. 6A,B, see methods for detail). In free moving mice, the ratio of GRABDA1m to tdTomato fluorescence was elevated upon the administration of methylphenidate, a known DAT blocker (Volkow et al., 1999) (Fig. 6C, top), and was suppressed by subsequent administration of Etic (Fig. 6C, top), a D2R blocker, implying the ability of GRABDA sensors in reporting DA dynamics to pharmacological treatments. Interestingly, we observed fluctuations in the ratio of GRABDA1m to tdTomato fluorescence that likely reflect the spontaneous DA release during the animal movements (Balleine et al., 2007; Da Silva et al., 2018; Howe and Dombeck, 2016), as they were prolonged during methylphenidate application and largely diminished by Etic administration (Fig. 6C, bottom).
Combining optogenetic stimulation and fiber photometry in freely moving mice, we observed that C1V1-evoked activation of DANs in the SNc generated a time-locked transient fluorescence increase in the dorsal striatum specific to the GRABDA1m channel (Fig. 6D), consistent with the evoked DA release from SNc DAN terminals. Systemic administration of the DA transporter blocker methylphenidate significantly prolonged the decay of optogenetically-evoked GRABDA1m fluorescence responses in the dorsal striatum (Fig. 6D,E). Furthermore, administration of D2R antagonist Etic largely abolished the response (Fig. 6D,F). Therefore, the GRABDA-and C1V1-based all-optical approach is effective for monitoring spontaneous and optogenetically-evoked DA release in the nigrostriatal pathway of mice.
Bi-directional modulation of DA dynamics in the nucleus accumbens (NAc) during Pavlovian conditioning
In addition to the nigrostriatal pathway, the dopaminergic projection from the ventral tegmental area (VTA) to the nucleus accumbens (NAc) regulates a variety of important functions including the reinforcement learning (Daw and Tobler, 2013; Glimcher, 2011; Gonzales et al., 2004), wherein an animal learns to associate an initially neutral sensory cue, such as a short tone burst, with an ensuing reward or punishment. Previous studies show that DANs in the VTA fire phasically to unpredicted rewards or reward-predicting cues and, with conditioning, the reward-evoked response diminishes as it becomes fully predicted by the cue (Bayer and Glimcher, 2005; Schultz, 2006; Schultz et al., 1997). To test whether our sensor can detect behaviourally relevant changes in endogenous DA release, we first expressed GRABDA1h in the NAc of head-fixed, water-restricted mice and trained them to associate a brief auditory cue with subsequent delivery of a liquid sucrose water reward (Fig. 7A). Each mouse experienced two distinct auditory cues, one that predicted delivery of a reward within a variable delay of 500-1500 ms after the end of the cue, as well as a second, randomly interleaved control cue (No Water, N.W.) that was not associated with water reward. Mice were trained daily for ∼10 days, and the GRABDA1h signal was recorded using in vivo fiber photometry in both the early and late stages of training. Consistent with the expected DA signaling in this classical paradigm, elevations in the GRABDA1h fluorescent signal aligned to reward delivery in every mouse, immediately in the first session without requiring training (Fig. 7C,E). After 6-10 days of training, mice selectively learned to associate the reward-predictive cue with delivery of reward. Consistent with the established reward prediction error theory of DA function, DA signals emerged in trained mice in response to the reward-predicting cue before the delivery of the reward (Fig. 7B,D-F). These results demonstrate that the signal-to-noise and temporal resolution of GRABDA1h is sufficient to detect physiologically relevant DA signaling in vivo in awake, behaving mice. Furthermore, these results serve to validate the GRABDA1h recorded DA signals by anchoring them to decades of established basal ganglia and the physiology of DA neurons.
In addition to transient elevations in DA signal in response to unexpected reward, DA neuron firing dips briefly in response to aversive stimuli (Brooks and Berns, 2013; Schultz, 2007; Ungless et al., 2004). We expressed the lower affinity GRABDA1m in the NAc of mice and trained them to associate a tone burst with an ensuing reward (a drop of water), or a punishment (a brief air puff to the face) to test whether our sensor can detect bi-directional changes in DA tone. During training, mice learned the association between an auditory cue and several outcomes (Fig. 7A). In vivo fiber photometry recording of GRABDA signals revealed that during early stages of training, reward or punishment delivery triggered a robust increase or decrease in the fluorescence signal in the NAc (Fig. 7G,H). Over the course of training, the magnitude of this reward-evoked response decreased, while a response of similar sign developed to the associated cue (Fig. 7G,H). In summary, the GRABDA sensor can be used to report the dynamic bi-directional changes in DA release over the course of Pavlovian conditioning.
Monitoring DA release in the NAc of mice during male mating behaviors
In contrast to the well-established involvement of DA in Pavlovian conditioning, DA dynamics during naturally rewarding social behaviors (Berridge and Robinson, 1998), such as courtship and mating, remains largely a subject of debate. Microdialysis and voltammetric measurements have previously revealed a relatively slow change in DA concentration during sexual behaviors: DA levels in the male’s NAc start to increase after a female is introduced into his cage and continues to rise throughout the course of sexual behaviors and ejaculation (∼20 minutes), then slowly (∼30 minutes) return to baseline after the female is removed (Damsma et al., 1992; Mas et al., 1995; Pfaus et al., 1990b). In contrast, a more recent study using FSCV showed that DA is transiently released in the male’s NAc when the female is introduced but few changes in DA levels were detected during subsequent sexual behaviors (Robinson et al., 2002; Robinson et al., 2001). To better understand DA dynamics during sexual behaviors, we virally expressed GRABDA1h in the NAc of male mice and used fiber photometry to record DA signals during sexual behaviors (Fig. 8A). Four weeks after injection, a sexually receptive C57BL/6 female was introduced into the home cage of the male mouse. Upon a female introduction, the male quickly approached and investigated the female, and then initiated mounting within the first minute. The GRABDA1h signals measured in the male’s NAc acutely and consistently increased with investigation of the female, mounting, intromission, ejaculation, and penile grooming (Fig. 8B). When we aligned the GRABDA1h signals with the manually annotated behaviors for each male, we observed that fluorescence signals increased prior to the corresponding behavior, peaked at the behavior’s onset, and then gradually declined (Fig. 8C). Among all of the sexual behaviors we annotated, the largest fluorescence increase occurred during intromission and ejaculation (Fig. 8C,D). These results indicate that DA in the NAc is acutely released during episodes of sexual behaviors and may carry information regarding specific features of courtship and/or mating.
Discussion
Here we describe the development and characterization of a pair of novel genetically-encoded sensors that enable specific, real-time detection of endogenous DA dynamics in various ex vivo and in vivo preparations and model systems. In acute mouse brain slices, GRABDA sensors were well suited to monitor stimulus-evoked DA release in mesolimbic pathway. In flies, GRABDA sensors were sufficiently sensitive to detect DA release in the MB triggered by odorants presented at physiologically relevant concentrations, and could also readily resolve DA release evoked by a single electrical stimulus. In transgenic zebrafish, GRABDA sensors were able to report DA release in the optic tectum in response to the chemogenetic activation of pretectal DANs. In mice, combining optogenetic stimulation with GRABDA sensors enabled the simultaneous optical manipulation and detection of DA signals in vivo. Finally, GRABDA sensors revealed real-time DA dynamics in the NAc of freely behaving mice as they underwent Pavlovian conditioning or engaged in sexual behaviors.
Compared to current methods used to measure DA release, our GRABDA sensors described here exhibit several clear advantages. First, GRABDA sensors are genetically-encoded by relatively small genes (∼2 kb), making them highly amenable to transgenic approaches and viral packaging. Second, GRABDA sensors have high sensitivity to DA. In response to DA, GRABDA1m and GRABDA1h sensors are capable of achieving maximal ΔF/F ∼ 90% with ∼10 nM and ∼100 nM affinities, respectively. In contrast, conventional GPCR-based FRET probes for detecting neurotransmitters are usually limited to a maximum FRET signal changes of ∼ 5% under optimal conditions, and less than that in vivo (Vilardaga et al., 2003; Ziegler et al., 2011). Third, GRABDA sensors have high specificity to DA: A range of experimental approaches, including application of multiple neurotransmitters and D2R antagonists, perturbation of DATs, or manipulation of DA synthesis pathways unequivocally support the molecular specificity of GRABDA sensors for DA. Notably, similar to the human D2R upon which they are based, both GRABDA sensors have a 10-fold higher affinity for DA than for NE (Lanau et al., 1997b); in contrast, FSCV is unable to discriminate between these two catecholamines (Fox and Wightman, 2016; Park et al., 2009). Finally, GRABDA sensors have extremely rapid response kinetics: GRABDA sensors report increases in DA levels with a rise time of = 100 ms (Fig. 1G, 3D and 4I). Although this response time of GRABDA sensors is slower than voltammetry methods, it is still sufficiently rapid for reporting physiologically relevant DA dynamics and share response kinetics similar to WT GPCRs (Lohse et al., 2008), providing an accurate readout of their activation.
Because GRABDA sensors were engineered using the D2R, a Gi-coupled GPCR, a potential concern is that overexpressing the GRABDA sensors may inadvertently activate pathways downstream of D2R. However, several lines of evidence argue against this possibility, including the negligible coupling between GRABDA sensors and both G protein-dependent and arrestin-dependent intracellular signaling pathways, alleviating this concern (Fig. 1I, 1J). This lack of coupling is presumably due to the steric hindrance imposed by the bulky cpEGFP moiety that replaces parts of the ICL3, which is the critical position for G protein or arrestin to interact with the GPCR (Luttrell and Lefkowitz, 2002; Neves et al., 2002). Consistent with minimal coupling between GRABDA sensors and downstream signaling pathways, in vivo Ca2+ imaging experiments using the Ca2+ sensor jRCaMP1a revealed no measurable alteration in Ca2+ signaling in neurons of transgenic flies that overexpressing GRABDA sensors (see Fig. S5).
In flies, DA signaling is critical for olfactory learning (Burke et al., 2012; Heisenberg, 2003; Liu et al., 2012; Schwaerzel et al., 2003). Here show that GRABDA sensors can be targeted to specific cells in the fly brain and used to probe odor-evoked DA dynamics in the MB of the living fly in real-time. It is reported that different types of DANs innervate different compartments of MBs, and these different dopaminergic pathways may play distinct roles in appetitive and aversive olfactory-dependent behaviors (Aso and Rubin, 2016; Cognigni et al., 2017). The GRABDA sensors developed here create new opportunities for exploring how distinct DA dynamics can correspond with specific compartments in the MB of the intact fly, particularly as the animal engages in different behavioral paradigms. Experiments in flies also illustrate the power of GRABDA sensors to probe DAT function in vivo by directly measuring extracellular DA level in real time. Similarly, we showed that GRABDA sensors readily respond to DA transients in the intact brain of the zebrafish larvae, providing a robust and convenient tool to examine DA dynamics in this classic vertebrate model system.
In addition to the morphological distinctions that can be made between various DANs in the mammalian CNS, breakthroughs in single-cell sequencing have further divided these neurons into a wide variety of cell types with distinct molecular features, suggesting high functional heterogeneity (Nair-Roberts et al., 2008; Ungless and Grace, 2012). Therefore, genetically-encoded GRABDA sensors provide a novel tool to explore patterns of DA release from genetically distinct DAN types, thereby facilitating the understanding of how they may be functionally specialized for different physiological and behavioral processes (Lammel et al., 2014; Pignatelli and Bonci, 2015). In fact, GRABDA sensors applied in the mouse faithfully reported the expected bi-directional regulation of DA levels in the NAc during different forms of Pavlovian conditioning. Consistent with the notion that DA can predict an anticipated reward (Schultz et al., 1997), the GRABDA sensor detected a robust, phasic increase in DA levels that shifted from reward onset to cue onset over the course of reward learning. Conversely, in naïve animals, a transient decrease in DA levels was triggered by the air puff onset and this decrease shifted to cue onset during aversive conditioning. An important goal of future studies will be to use GRABDA sensors to determine whether the same or different types of DANs are involved in appetitive vs. aversive learning.
Our experiments in freely behaving male mice provide new perspectives with respect to DA dynamics during sexual behaviors. Contrary to traditional views, GRABDA sensors expressed in the NAc revealed striking time-locked DA elevation immediately prior to and peak at the onset of various distinct sexual behaviors, consistent with a model where DA encodes behavioral motivation, anticipation, or arousal. These rapid changes in DA levels are in contrast with previous dialysis studies showing that DA levels in the NAc slowly increase during sexual behaviors, a difference that likely relates to the slow readout associated with dialysis-based methods. The targeted expression of GRABDA sensors could therefore provide a critical window into the coding strategy of DA release in complex behaviors. Moreover, because GRABDA sensors readily discriminate between DA and NE, they may be useful in studying cortical and subcortical regions in which dopaminergic and adrenergic inputs are intertwined.
Given that the crystal structure of the D2R was recently solved (Wang et al., 2018), future efforts can use this structural information to further tune the affinity, enhance the selectivity, and increase the signal-to-noise ratio in the next-generation of GRABDA sensors. Moreover, by adding a red fluorescent protein, GRABDA sensors can be readily transformed into ratiometric indicators, which could prove useful for more quantitative measurements of DA release across different experiments and preparations. Finally, a GPCR-based strategy was recently used to develop a genetically-encoded sensor (GACh) with high sensitivity and high selectivity for acetylcholine (ACh) (Miao Jing et al., 2018). Although the GACh sensor differs from the GRABDA sensors described here in that is based on the muscarinic Gq-coupled GPCR receptor M3R and an ICL3 loop derived from a Gs-coupled beta adrenergic receptor (Levitzki, 1988; Rasmussen et al., 2011), a feature common to this sensor is that the conformational changes in a GPCR induced by ligand binding are successfully harnessed and converted into a sizable increase in cpEGFP florescence. Given the diverse ligand-specificity of different GPCRs, a future goal will be to explore whether this principle can be expanded even further in order to develop sensors for the entire range of neurotransmitters and neuromodulators.
Methods
Animals
Wild-type Sprague-Dawley rat pups (P0) were used to prepare cultured cortical neurons. Wild-type C57BL/6 and TH-Cre mice (B6.FVB(Cg)-Tg(Th-cre)FI172Gsat/Mmucd obtained from MMRRC) were used to prepare the acute brain slices and in vivo experiments. All animals were maintained in the animal facilities and were family-or pair-housed in a temperature-controlled room with a 12-h/12-h light/dark cycle. All procedures for animal surgery and maintenance were performed using protocols that were approved by the Animal Care & Use Committees at Peking University, Chinese Academy of Sciences (CAS), New York University, University of California, San Francisco, and US National Institutes of Health, and were performed in accordance with the guidelines established by US National Institutes of Health guidelines.
To generate transgenic zebrafish, plasmids containing pTol2-elval3:GRABDA1m (25 ng/μL) and Tol2 mRNA (25 ng/μL) were co-injected into fertilized eggs, and founders were screened three months later. Transgenic zebrafish adults and larvae were maintained at 28 °C on a 14-h/10-h light/dark cycle.
To generate transgenic flies, the coding sequence of GRABDA1m was integrated into the pUAST vector using Gibson Assembly (Gibson et al., 2009), which was then used in P-element-mediated random insertion. Transgenic Drosophila lines carrying GRABDA1m on the chromosomes 2 and 3 with the strongest expression level after crossing with TH-GAL4 were used. The coding sequence of GRABDA1m-mut was incorporated into pJFRC28 (Pfeiffer et al., 2012) (Addgene plasmid #36431) using Gibson Assembly, and this plasmid was used to generate transgenic flies using PhiC31-mediated site-directed integration into attp40. The embryo injections were performed at Core Facility of Drosophila Resource and Technology, Shanghai Institute of Biochemistry and Cell Biology, CAS. Transgenic flies were raised on conventional corn meal at 25°C, with ∼70% humidity, under 12:12-h light-dark cycle.
The following Drosophila lines used in this study:
TH-GAL4, a gift and unpublished line generated by appending 2A-GAL4 to the last exon of TH, from Yi Rao, Peking University. C305a-GAL4 and 30y-GAL4, also gifts from Yi Rao. DTHFS+/-ple2/TM6B (Cichewicz et al., 2017), a gift from Jay Hirsh, University of Virginia. UAS-DAT-RNAi (TH01470.N), from Tsinghua Fly center, Tsinghua University. UAS-jRCaMP1a (Bloomington #63792), a gift from Chuan Zhou, Institute of Zoology, Chinese Academy of Sciences.
The following genotypes were used in the following figures: Fig. 4A-C
UAS-GRABDA1m/cyo; TH-GAL4 (DANs)/TM6B
UAS-GRABDA1m-mut/+; TH-GAL4/+
c305a-GAL4 (a’ and β’ Kenyon cells)/UAS-GRABDA1m; DTHFS+/-ple2/+ (WT group)
c305a-GAL4/UAS-GRABDA1m; DTHFS+/-ple2 (TH-deficient group)
UAS-GRABDA1m/cyo; TH-GAL4/TM6B
UAS-GRABDA1m-mut/+; TH-GAL4/+
UAS-GRABDA1m/cyo; TH-GAL4/TM6B
UAS-GRAB DA1m/+; TH-GAL4/UAS-DAT-RNAi
TH-GAL4/ UAS-jRCaMP1a
UAS-GRAB DA1m/+; TH-GAL4/UAS-jRCaMP1a
30y-GAL4/ UAS-jRCaMP1a
UAS-GRAB DA1m/+; 30y-GAL4/UAS-jRCaMP1a
Molecular biology
Plasmids were generated using Gibson Assembly. DNA fragments were generated using PCR amplification with primers (Thermo Fisher Scientific) with 30-bp overlap. The fragments were assembled using T5-exonuclease (New England Biolabs), Phusion DNA polymerase (Thermo Fisher Scientific), and Taq ligase (iCloning). All sequences were verified using Sanger sequencing (Sequencing platform in the School of Life Sciences of Peking University). DNA encoding the various DA receptor subtypes (D1R-D5R) was generated using PCR amplification of the full-length human GPCR cDNAs (hORFeome database 8.1). For characterization in HEK293T cells, the GRABDA constructs were cloned into the pDisplay vector (Invitrogen), with an IgK leader sequence inserted upstream of the coding region. The IRES-mCherry gene was attached downstream of GRABDA and was used as a reference of membrane marker to calibrate of the signal intensity. Site-directed mutagenesis of the N- and C- terminal linker sequences in cpEGFP was performed using primers containing randomized NNB codons (48 codons in total, encoding the 20 possible amino acids; Thermo Fisher Scientific). Site-directed mutagenesis of the D2R gene was performed using primers containing the target sites. For the characterization in cultured neurons, the GRABDA constructs were cloned into the pAAV vector under the TRE promoter or the human synapsin promoter. The marker constructs RFP(mScarlet)-CAAX, EGFP-CAAX, KDELR1-EGFP, PSD95-mScarlet and synaptophysin-mScarlet were cloned into pEGFP-N3 vector.
Expression of GRABDA in cultured cells and in vivo
HEK293T cells were cultured in DMEM supplemented with 10% (v/v) FBS (Gibco) and 1% penicillin-streptomycin (Gibco) at 37°C in 5% CO2. The cells were plated on 12-mm glass coverslips in 24-well plates and grown to ∼50% confluence for transfection. Transfection was performed by incubating the HEK293T cells with a mixture containing 1 μg of DNA and 3 μg of PEI for 6 h. Imaging was performed 24-48 h after transfection.
Rat cortical neurons were prepared from postnatal 0-day old (P0) Sprague-Dawley rat pups as previously described (Zhang et al., 2009). In brief, the cortical neurons were dissociated from the dissected rat brains in 0.25% Trypsin-EDTA (Gibco), and plated on 12-mm glass coverslips coated with poly-D-lysine (Sigma-Aldrich) in neurobasal medium containing 2% B-27 supplement, 1% GlutaMax, and 1% penicillin-streptomycin (Gibco). The cells were transfected 7-9 days later using the calcium phosphate transfection method. Imaging was performed 48-72 h after transfection.
For in vivo expression, Wild-type C57/BL6 mice with the age of P42-60 were first anesthetized by 2,2,2-Tribromoethanol (Avetin, 500 mg/kg) through intraperitoneal injection, or by isoflurane (RWD Life Science), and then placed in a stereotaxic frame to inject AAVs of GRABDA sensors into NAc with a microsyringe pump (Nanoliter 2000 Injector, WPI), or a microinjection pipette injector (Nanoject II, Drummond Scientific). The coordination of NAc was set as AP: −1.40 mm from Bregma, ML: 1.00 mm from the midline, DV: 3.90 mm from the brain surface. The injection was made unilateral with ∼300-500 nL per animal. In dual-color optical recordings experiments, the AAVs of hsyn-GRABDA1m and hsyn-tdTomato were injected in the dorsal striatum (AP = - 0.5mm, ML = ±2.5mm from bregma, and DV = −2.2mm from the brain surface), and AAV of Ef1a-DIO-C1V1-YFP was injected in the substantia nigra pars compacta (SNc) (AP = −3.1mm, ML = ±1.5mm from bregma, and DV = −4.0mm from the brain surface) in TH-Cre mice.
Fluorescence imaging of HEK293T cells and cultured neurons
GRABDA-expressing HEK293T cells and cultured neurons were imaged using an inverted Ti-E A1 confocal microscope (Nikon) and the Opera Phenix high content screening system (PerkinElmer). The Nikon confocal microscope was equipped with a 40×/1.35 NA oil immersion objective, a 488-nm laser and a 561-nm laser. During imaging, the HEK293T cells and cultured neurons were bathed or perfused in a chamber with Tyrode’s solution containing (in mM): 150 NaCl, 4 KCl, 2 MgCl2, 2 CaCl2, 10 HEPES and 10 glucose (pH 7.4). Solutions containing the drug/compound of interest (e.g., DA, Halo, 5-HT, histamine, NE, or ACh) were delivered via a custom-made perfusion system or via bath application. The chamber was fully cleaned with Tyrode’s solution and 75% ethanol between experiments. The GFP signals (e.g., the GRABDA sensors, the iGluSnFR or EGFP) were recorded using a 525/50 nm emission filter, and the RFP signals were collected using a 595/50 nm emission filter. The photostabilities of GRABDA1m, GRABDA1h, EGFP, and iGluSnFR were measured using 350-μW 488-nm laser illumination. Photobleaching was applied to the entire sensor-expressing HEK293T cell. The Opera Phenix high content screening system was equipped with a 60×/ 1.15 NA water immersion objective, a 488-nm laser, and a 561-m laser. The GRABDA signals were collected using a 525/50 nm emission filter, and the mCherry signals were collected using a 600/30 nm emission filter. Where indicated, the culture medium was replaced with 100 μl of Tyrode’s solution containing various concentrations of the indicated drug/compound. The fluorescence intensities of the GRABDA sensors were calibrated using mCherry as the reference.
Fluorescence imaging of GRABDA in brain slices
Two weeks after the virus injection, the animals were anesthetized with Avetin and then decapitated. The brains were removed immediately and placed directly in cold slicing buffer containing (in mM): 110 choline-Cl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 7 MgCl2, 25 glucose, and 2 CaCl2. The brains were then sectioned into 200-μm thick slices using a VT1200 vibratome (Leica, Germany), and the sections were transferred into the oxygenated Ringer’s buffer containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 1.3 MgCl2, 25 glucose, and 2 CaCl2; the slices were then allowed to recover in 34 °C for at least 40 minutes. For fluorescence imaging, the slices were transferred to an imaging chamber in an Olympus FV1000MPE two-photon microscope equipped with a 40×/0.80 NA water-immersion objective and a mode-locked Mai Tai Ti:Sapphire laser (Spectra-Physics) tuned to 920 nm for the excitation of GRABDA sensors and a 495∼540 nm filter for signal collection. For electrical stimulation, a concentric electrode (model #CBAEC75, FHC) was positioned near the NAc core under the fluorescence guidance, and the imaging and stimulation were synchronized using an Arduino board with custom programs. The stimulation voltage was set at 5-6 V, and the duration of each stimulation was set at 2 ms.
For immunostaining of brain sections, GRABDA-expressing mice were anesthetized with Avetin, and the heart was perfused with 0.9% NaCl followed with 4% paraformaldehyde (PFA). The brain was then removed and placed in 4% PFA for 4 hours, then cryoprotected in 30% (w/v) sucrose for 24 hours. The brain was embedded into tissue-freezing medium, and 50-μm-thick coronal sections were cut using a Leica CM1900 cryostat (Leica, Germany). To label dopaminergic terminals and GRABDA in the NAc, tissue sections were rinsed and then immunostained with rabbit anti-TH antibody (1:100, Millipore, #ab152) and chicken anti-GFP antibody (1:500, Abcam, #ab13970), followed by an Alexa-555-conjugated goat-anti-rabbit and Alexa-488-conjugated goat-anti-chicken secondary antibodies. The immunostained tissue sections were imaged using the same Nikon confocal microscope in cell imaging.
Fluorescence imaging of transgenic flies
Adult Drosophila (within 3 weeks of eclosion) were used for imaging experiments. The mounting and dissection protocols were as previously described (Liang et al., 2013). In brief, a section of rectangular cuticle between the eyes was removed to expose the brain, which was then bathed in saline, so called adult hemolymph-like solution (AHLS). The same Olympus two-photon microscope used for brain slices imaging was also used here. For GRABDA sensors imaging, 920-nm excitation laser and 495∼540-nm filter were used. For jRCaMP1a, 1000-nm excitation laser and 575∼630-nm filter were used. For olfactory stimulation, the odorant isoamyl acetate (Sigma-Aldrich; Cat. #306967) was firstly diluted 200-fold in mineral oil in a bottle; this dilution was subsequently diluted 5-fold in air, which was then delivered to the fly’s antenna at a rate of 1000 ml/min. Compounds such as Halo and cocaine were added directly to the AHLS to their final concentration, and the following experiments were performed 10 min after compound application. For electrical stimulation, a glass electrode (resistance ∼0.2 MO) was placed in the region of the DANs in the MB and the stimulation voltage was set at 20∼80 V. Arduino was used to synchronized stimulation delivery and imaging with custom code. The sampling rates during olfactory stimulation and electrical stimulation was 2.7 Hz and 12 Hz, respectively.
Fluorescence imaging and chemogenetics in zebrafish
All experiments were performed on 5 days-post-fertilization (5 dpf) larvae in 10% Hank’s solution containing (in mM): 140 NaCl, 5.4 KCl, 0.25 Na2HPO4, 0.44 KH2PO4, 1.3 CaCl2, 1.0 MgSO4, and 4.2 NaHCO3 (pH 7.2). Imaging of Tg (elval3:GRABDA1m/DAT:TRPV1-TagRFP) larvae at 5 dpf was performed with an inverted confocal microscope (Olympus FV3000, Japan) using a 30X oil-immersion objective (1.05 N.A., morphology imaging) or an upright confocal microscope (Olympus FV1000, Japan) using 40X water-immersion objective (0.8 NA, time-lapse imaging). After the larvae were paralyzed with a-bungarotoxin (100?μg/ml, Sigma), they were mounted dorsal side up in 1.5% low melting-point agarose (Sigma) and then immersed in an extracellular solution consisting of (in mM): 134 NaCl, 2.9 KCl, 4 CaCl2, 10 HEPES and 10 glucose (290?mOsmol/L, pH 7.8). For imaging the morphology, images were acquired with a field of view consisting of 1,024 pixels × 1,024 pixels with spatial resolution of 0.414 × 0.414 × 1 μm3 (x × y × z). For bath application of compounds, dopamine (100 μM in 1 mM ascorbic acid solution, Sigma) was added by pipette at ∼ 4 min and haloperidol (50 μM in DMSO, Tocris) at ∼12 min. These images were acquired with a view field of 640 × 640 pixels with spatial resolution of 0.497 × 0.497 μm2 (x × y) at ∼1.5 Hz. For functional imaging, small anterior dissections initiated in ventricles were made, after which a glass pipette containing the TRPV1 agonist capsaicin (50 μM in absolute ethanol, Tocris) was advanced through the incision and placed near the cell bodies of the DANs. To activate the DANs, 5 pulses of puffs (9-10 psi, 100-ms) were delivered with 1 min interval. The larvae were bath in Halo (50 μM in DMSO, Tocris) for 10 min before imaging. These images were acquired with a field of view consisting of 800 × 800 pixels with spatial resolution of 0.397 × 0.397 μm2 (x × y) at ∼1 Hz.
Fiber Photometry recording in freely moving mice
In all-optic experiments in Fig. 6, optical fiber probes (105 μm core /125 μm cladding) were implanted in the dorsal striatum and in SNc 4 weeks after the virus injection. Fiber photometry recording in the dorsal striatum was performed using a 50-μW 470-nm LED, and C1V1 in the SNc was stimulated using a 9.9- mW 561-nm laser. The measured emission spectra of GRABDA1m and tdTomato were fitted using a linear unmixing algorithm https://www.niehs.nih.gov/research/atniehs/labs/ln/pi/iv/tools/index.cfm. The coefficients of GRABDA1m and tdTomato generated by the unmixing algorithm were used to represent the fluorescence intensities of GRABDA1m and tdTomato, respectively. To evoke C1V1-mediated DA release in the dorsal lateral striatum, pulse trains (10-ms pulses at 10 Hz for 1 s) were delivered to the SNc using a 9.9-mW, 561-nm laser. In other experiments in Fig. 7 and 8, an optic fiber (Thorlabs, FT200UMT,FT400UMT or BFH48-400) was attached to the implanted ferrule (Thorlabs, SF440-10) via a ceramic sleeve. A 400-Hz sinusoidal blue LED light (30 μW) (LED light: M470F1; LED driver: LEDD1B; both from Thorlabs) was bandpass filtered (passing band: 472 ± 15 nm, Semrock, FF02-472/30-25 in Fig.8 or 460-490nm in Fig.7) and delivered to the brain to excite GRABDA. The emission light then traveled through the same optic fiber, was bandpass filtered (passing band: 534 ± 25 nm, Semrock, FF01-535/50 in Fig.8 or 500-550nm in Fig.7), detected by a Femtowatt Silicon Photoreceiver (Newport, 2151) and recorded using a real-time processor (RZ5, TDT). The envelope of the 400-Hz signals that reflects the intensity of the fluorescence signals was extracted in real-time using a custom TDT program.
Behaviors
For the auditory conditioning task, mice were recovered for >3 days after surgery, and then water-restricted until reaching 85-90% of its original body weight and then prepared for behavior training. In the first Pavlovian task, the mice were trained on two frequency modulated pure tone auditory cues of 500 ms in duration, centered around 2.5 kHz and 11 kHz. For each mouse, one of the two tones was pseudo-randomly assigned to be the reward-predictive tone. Reward (water sweetened with 10% sucrose) was delivered through a water spout in front of the mouth following the reward-predictive cue with a variable 500-1500 ms delay. Rewarded and unrewarded trials were randomly interleaved with a variable inter-trial interval of 8-20 s. Mice experienced 200 trials (∼100 rewards) per day in sessions lasting ∼45 min.
In the subsequent Pavlovian conditioning task, the mice were trained on an auditory conditioning task, in which three pairs of auditory cues → outcomes pairs (or CS-US pairs; 8 kHz pure tone → 9 μl water; white noise → brief air puff to the face; and 2 kHz pure tone → no response) were delivered at random with a 10 −20 s randomized inter-trial interval. The duration and intensity of each auditory cue was 1 s and 70dB, respectively. The respective outcomes was delivered 1 s after the end of each auditory cue. The behavioral setup consisted of a custom-built apparatus allowing head fixation of the mouse’s head to a Styrofoam rod (diameter: 15 cm). Rotation of the Styrofoam rod, which corresponds to the animal’s running speed, was detected using an optical rotatory encoder. Licking behavior was detected when the mouse’s tongue contacted the water delivery tube. Each lick signal was processed using an Arduino UNO board with custom code and sent digitally to the training program (written in MATLAB) via a serial port. Water delivery was precisely controlled using a stepping motor pump, and the air puff (15 psi, 25-ms duration) was controlled using a solenoid valve. Timing of the pump and valve was controlled using the same Arduino UNO board used for lick detection, which also provided synchronization between the training program and the data acquisition system (RZ2 processor, Tucker-Davis Technologies). During first two days of each training session, the outcomes were delivered without the prediction cues.
The sexual behaviors are defined following conventions in previous literature (Hull and Rodriguez-Manzo, 2009). In details, sniffing female was defined as the male’s nose coming in close proximity to the female’s facial, body, and/or urogenital areas. “Mount” was defined as when the male posed his forelegs over the female’s back and with his hindfeet on the ground accompanying shallow pelvic thrusts. The mounting onset was defined as the moment at which the male tried to clasp female back. “Intromission” was defined as a deep rhythmic thrust following mounting. The onset of intromission was defined as the time at which the male performed the first deep thrusting toward the female with vaginal penetration. “Penile grooming” was defined when a male animal repeated grooming for his urogenital area after intromission and ejaculation. Ejaculation is detected when the male stopped thrusting and freeze for seconds. The putative ejaculation event was confirmed by the presence of vaginal copulatory plug.
Data analysis
For imaging experiments in HEK293T cells, neurons, acute brain slices and transgenic flies, images were first analyzed using Image J software (National Institutes of Health), and then analyzed using Origin 9.1 (OriginLab) and MATLAB (MathWorks) with custom-written scripts. The data in acute brain slices and flies were first binned by 2x and averaged to generate representative traces.
For fiber photometry data, the signal baseline was first obtained by the MATLAB function “msbackadj” with a moving window of 25% of the total recording duration in Fig.8, or by subtracting 2nd order exponential fitted data from the raw data after 10.17 Hz binning in Fig. 7. The fluorescence responses were indicated by ΔF/F0 in Fig.8, which was calculated as (Fraw –Fbaseline)/Fbaseline, or by Z score in Fig.7. To analyze event-evoked changes in DA release, we aligned each trial to the onset or offset of the behavior, and calculated the peri-stimulus time histogram (PSTH). To compare PSTH changes during different phases of the training, we used data from the 2nd day as naive, the 5-10th day as trained and >10th day as well-trained, and normalized the PSTH of each animal by water-evoked response during early training. The peak response during a behavior was calculated as the maximum ΔF/F0 during the behavior minus the average ΔF/F0 in the duration-matched period 2s prior to the behavior onset in Fig.8. The response to the CS was defined as the peak of the normalized PSTH between the CS onset and the US onset, and the response to US was calculated similarly using data from the US onset to data collected 2 s after the US onset in Fig.7.
Except where indicated otherwise, group differences were analyzed using the Student’s t-test, test, sign-rank test, One-Way ANOVA or post-hoc Tukey’s test.
Except where indicated otherwise, all summary data presented as the mean ± SEM.
Author Contributions
Y.L. conceived and supervised the project. F.S., M.J., and J.Z. performed experiments related to sensor development, optimization and characterization in culture HEK cells, culture neurons, brain slices and transgenic flies, with the initial work from YC. L. and J. F., and help from Z.Y. F.L. and J.D. designed and performed experiments on transgenic fish. J.Z., Y.G., T.Y., W.P., S.O., L.W., S.Z., D.L., M.X., A.K., and G.C. designed and performed experiments in behaving mice. All authors contributed to data interpretation and data analysis. Y.L. wrote the manuscript with input from F.S., J.Z., M.J., D.L., S.O., M.X. and help from other authors.
Acknowledgements
This work was supported by the National Basic Research Program of China (973 Program; grant 2015CB856402), the General Program of National Natural Science Foundation of China (project 31671118 and project 31371442), an NIH brain initiative grant NS103558, and the Junior Thousand Talents Program of China to Y.Li and to S.Zhang. Additional support was provided by the Ministry of Science and Technology of the People’s Republic of China (2017YFA0505703) to M.Xu. We thank Y. Rao for sharing the 2-photon microscope and X. Lei for the platform support of Opera Phenix high content screening system at PKU-CLS. We thank the Sequencing platform in the School of Life Sciences of Peking University. We thank R. Mooney, Y. Huan and L. Luo for valuable feedback of the manuscript.