The zona incerta negatively regulates the red nucleus during movement cued by sound signals

To determine the role of the red nucleus (RN) in the auditory-cued goal-directed behavior, we recorded RN neuronal activities in mice performing an auditory frequency discrimination task as illustrated in Fig 1A or previously described [3]. In brief, mice learned to self-trigger an auditory cue by poking their noses into the center port, and then moved to one of the side ports for water reward depending on the tone frequencies in the cues (low tones to left port and high tones to right port). Mice’s performance was driven by the water reward because they were water restricted, and they stopped task performance quickly if the water reward was omitted. We performed tetrode recordings in the left RN of adult wild type mice (WT, C57BL/6) and isolated individual excitatory units based on the waveforms of the action potentials as shown in Fig 1B. We aligned the activities of excitatory units to the initiation of movements when mice withdrew from the center port (Time from CenterOut) and moved to the side ports for water reward. The activity of a single neuron during trials has been exampled in Fig 1C. Among the 198 units of excitatory neurons recorded from the left RN of 5 mice (42 from mouse 1, 35 from mouse 2, 58 from mouse 3, 28 from mouse 4, 35 from mouse 5), we found that 61 units were preferentially activated during the movements from the center port toward the right-side port, referring to the contralateral movement (Fig 1D and 1E). These neurons showed the same activity patterns in both correct and error trials (S7A Fig) when they moved to the same side ports, indicating that the RN neuronal activity is correlated with the side of the movement (movement decision) rather than auditory cue identity (sensory representation). In contrast, there were only 13 units showing an ipsilateral preference. Importantly, similar analyses of inhibitory neurons revealed that there was no obvious side preference (S1A and S1B Fig). To analyze whether the 198 units exhibit any preferences in tone frequencies, we aligned the unit activity to the onset of auditory cues (Time from SoundOn). As shown in S2 Fig, only 19 of the 198 units exhibited potential preference to either the high (n = 15) or low (n = 4) frequency tones, suggesting that RN excitatory neurons did not exhibit a preferential activation to the frequencies of auditory cues. Together, these results suggest that excitatory neurons in the RN may be preferentially involved in contralateral movements in the task.

Fig 1. RN excitatory neurons are preferentially activated during the contralateral movements in an auditory task.

(A) Schematic of the auditory frequency discrimination task. In this two-alternative forced-choice task, mice poke the center port to trigger auditory cues. Mice should then poke one of the side ports for water reward depending on the tone frequencies (low frequency to left side, high frequency to right side). (B) Left: schematic of tetrode recording in the RN of a mouse. Right: example waveforms of excitatory (black) and inhibitory (grey) units. (C) One example RN excitatory neuron with contralateral preference. Top: Raster plots of neuronal activity for individual ipsilateral (green) and contralateral (purple) trials aligned to the time when the mouse withdrew from the center port (CenterOut). Bottom: PSTH for ipsilateral (green) and contralateral (purple) trials. (D) Histogram plots show the number of neurons (198 units total) with contralateral preference (purple, 61 units) and ipsilateral preference (green, 13 units). χ² stat = 38.2904; p = 6.096e – 10. (E) Heatmap showing the activity of neurons with contralateral preference (Wilcoxon signed-rank test, p 2 stat = 38.2904; p = 6.096e – 10). (F) Schematic of viral delivery and photoactivation of either left or right RN during the task. 5 ms light pulses at 20 Hz for the whole duration of indicated time window. (G) Left: psychometric curves of mice’ performance in control (black) and photostimulation of left RN (green) trials. Right: psychometric curves of mice’ performance in control (black) and photostimulation of right RN (purple) trials. For the left RN, n = 16 session from 3 mice (5 sessions for mouse 1, 3 for mouse 2, 8 for mouse 3), p = 4.3778e – 04; for the right RN, n = 16 session from 3 mice (7 sessions for mouse 1, 4 for mouse 2, 5 for mouse 3), p = 6.1702e – 07, error bars, mean ± s.e.m. Wilcoxon sign-rank test. Underlying data can be found in the S1 Data.

https://doi.org/10.1371/journal.pbio.3003092.g001

We next examined whether excessive activation of RN excitatory neurons right after the CenterOut impacts the movements towards the reward ports. To specifically and temporally activate RN excitatory neurons, we employed the optogenetic method as previously described [6]. In a cohort of adult WT mice, we injected adeno-associated virus (AAV) expressing channel rhodopsin 2 (ChR2) [20] under the CaMKII promoter, termed as AAV-CaMKII-ChR2, into the RN bilaterally. Optic fibers were then implanted right above the RN (Fig 1F). We found that activation of the RN excitatory neurons when mice moved from the center port to a side port induced a substantial contralateral bias (Fig 1G). To rule out the possibility that light stimulation itself caused an effect on the biased behaviors, we expressed green fluorescence protein (GFP) in the left RN and performed the same set of tests with optical stimulation. Light stimulation in the RN itself caused no detectable effect on the task performance (S3 Fig).

To seek neural circuits projecting to the RN that may regulate its activity to impact the movements in this goal-directed auditory task, we used an engineered rabies viral system [21] to retrogradely label brain regions directly projecting to the RN. As illustrated in Fig 2A, to specifically express GFP, TVA, and oG in RN excitatory neurons, we injected AAV-FLEX-TVA-P2A-eGFP-2A-oG into the left RN of adult VGlut2-Cre mice [22]. Three weeks after the AAV injection, we performed the second injection in the same injection site with the EnvA G-deleted Rabies-mCherry, which can infect those RN neurons expressing TVA. GFP and mCherry will both be expressed in RN excitatory neurons, which are defined as starter neurons. The mCherry will retrogradely cross synapses in the presence of oG neurons projecting to the starter neurons. We collected the brain tissue and prepared brain slices for confocal imaging one week after the rabies infusion. We detected mCherry+ neurons in a group of brain regions, including the motor cortex, substantial nigra pars reticulata, subthalamic nucleus and zona incerta (ZI) (Fig 2B).

Fig 2. RN excitatory neurons receive ZI PV neuron projection.

(A) Schematic of monosynaptic retrograde tracing from RN excitatory neurons from 3 mice. (B) Representative mCherry+ neurons in motor cortex (M1, M2), Cingulate cortex (Cg1), Somatosensory cortex (S1), Subthalamic nucleus (STN), Substantia nigra pars reticulata (SNr) and Zona incerta (ZI), that directly project to RN excitatory neurons. (C) Left: Schematic of anatomical location of medial ZI; Right: Immunostaining for PV, Lhx6 and SST cells in the medial ZI. The dash line indicates the border between the dorsal and ventral part of the ZI. Scale bar: 200 μm. (D) Reconstruction of PV (blue), Lhx6 (purple) and SST (orange) cells in the medial ZI. Dash line indicates the border between the dorsal and ventral part of the ZI. Scale bar: 200 μm. (E) ZI projections to the RN from PV-Cre (left), Lhx6-Cre (middle), and SST-Cre (right) mice. Upper row shows the representative images from injection sites (ZI), lower row shows from the projection sites (RN). Underlying data can be found in the S1 Data.

https://doi.org/10.1371/journal.pbio.3003092.g002

Collective evidence has indicated the ZI functions in controlling visceral activity, attention, arousal, posture or locomotion, eating, hunting, defensive and nocifensive behaviors, and potentially encodes actions [10–19]. Clinically, the ZI has been recently suggested as a candidate for deep brain stimulation to treat Parkinson’s Disease [23]. Given its potential role in motor functions, the ZI has been proposed to be an integrative node for generating direct responses to a given sensory stimulus [11,24]. We therefore determined the possible role of the ZI in regulating the activity of the RN in auditory cued goal-directed task.

Most of the retrogradely-traced neurons in the ZI were in the ventral portion of the ventral medial ZI (vZIm) (Fig 2B). The majority of ZI neurons express GAD, an interneurons marker [11]. To verify the interneuron types in the vZIm, we first immunostained the ZI brain slices with different types of interneuron markers. There were three types of inhibitory neurons, parvalbumin (PV)-, somatostatin (SST)-, and LIM homeobox protein 6 (Lhx6)-positive neurons (Fig 2C and 2D), consistent with previous findings [11].

We next examined whether all these types of interneurons project to the RN. AAV-DIO-GFP, with a Cre-dependent expression of GFP cassette, was infused into the left ZI of PV-Cre, Lhx6-Cre or SST-Cre mice, respectively (Fig 2E). Three weeks after the viral infusion, we collected the brain tissue and analyzed the GFP signal in the RN by confocal imaging method. We found that there were GFP+ terminals in the ipsilateral RN (indicating projections from the ZI) only from the PV-Cre mice. In the Lhx6- and SST-Cre mice, we did not detect GFP+ signal in the RN (Fig 2E). Together our results indicate that ZI PV neurons, but not Lhx6 and SST neurons, project to the RN.

The projection of PV neurons in the ZI to the ipsilateral RN motivated us to explore its potential role in regulating the RN’s neuronal activity and function in the auditory task. First, we optogenetically activated ZI PV terminals in the RN and recorded RN neuronal activity. As illustrated in Fig 3A, we expressed ChR2 in PV neurons of the left ZI by injecting an AAV carrying DIO-ChR2, a Cre-dependent expression cassette of ChR2, into left ZI of PV-Cre mouse. A cannula was implanted right above the left ZI. A tetrode bundle integrated with an optical fiber, termed optrode, was implanted in the left RN. Three weeks after the surgery, neuronal activity of the left RN upon optical stimulation was recorded using the tetrode method as previously described [2,3]. To prevent the interference of antidromic activation of ZI PV soma, we infused TTX (1 μM, 300 nl) through the cannula in the ZI during the recording. We compared the firing rates between baseline (50 ms before the light onset) and light period (0–500 ms) for each recorded neuron, using Wilcoxon rank-sum test. Among the 44 identified units, 23 of them dramatically decreased firing rates in response to the light stimulation, 3 of them increased the firing rates and the rest 18 units showed no detectable changes in firing rates (Fig 3A and 3B). Overall, it indicates that ZI PV neuronal activation suppressed neuronal activity in the RN.

Fig 3. ZI PV neurons inhibited RN neuronal activity and biased mice’s performance.

(A) Left: Schematic of photostimulation and tetrode recording under the local infusion of TTX into the ZI. Right: representative raster plot and PSTH plot from one recorded unit. (B) A pie chart showing the changes in RN neuronal activity upon ZI PV neuronal terminal opto-stimulation and TTX infusion simultaneously. (C) The experimental design to photostimulate or photosilence ZI PV neurons during task performance. Upper panel: Vial delivery and photo manipulations of PV cells in the ZI. Lower panel: Photo manipulation time window (blue bars, movement epoch). (D) Psychometric curves of mice’ performance in control (black) and photostimulation of left ZI (green) trials or right ZI (purple) trials. n = 19 sessions from 6 mice (2 sessions for mouse 1, 4 for mouse 2, 2 for mouse 3, 3 sessions for mouse 4, 4 for mouse 5, 4 for mouse 6), error bars, mean ± s.e.m. p = 0.0017; Wilcoxon sign-rank test. (E) Psychometric curves of mice’ performance in control (black) and photosilencing of left ZI (green) trials or right ZI (purple) trials. n = 25 session from 4 mice (7 sessions for mouse 1, 4 for mouse 2, 5 for mouse 3, 9 for mouse 4), error bars, mean ± s.e.m. p = 2.e – 05, Wilcoxon sign-rank test has been used for statistical analysis. (F) Left: Schematic of combining TTX fusion into ZI with RN terminal photostimulation. Right: Psychometric curves of mice’ performance in control (black) and photostimulation of right RN + TTX (purple) trials. n = 26 session from 4 mice (5 from mouse 1, 7 from mouse 2, 8 from mouse 3, 6 from mouse 4), error bars, mean ± s.e.m. p = 0.0186, Wilcoxon sign-rank test. (G) Photostimulation of ZI PV terminals in the RN did not change mouse’s movement directions out of the behavioral task. mean ± s.e.m. 42 trials from 3 mice. Underlying data can be found in the S1 Data.

https://doi.org/10.1371/journal.pbio.3003092.g003

We next explored whether such inhibition by ZI PV neuronal activation to the RN affects the movements to reward ports in the task. In a cohort of adult PV-Cre mice, after being well-trained in performing the auditory task, we injected the AAV-DIO-ChR2 into the ZI bilaterally, followed with fiber cannula implantation. Three weeks after the surgery, we optogenetically activated the ChR2 expressing PV neurons either in the left side or right side of the ZI during the movement period. We found that the optical activation of PV neurons of the ZI substantially biased the movements to ipsilateral port (Fig 3C and 3D). Interestingly, optical activation of ZI PV neurons only during the first 200 ms of movement, but not the later phase, induced ipsilateral bias (S4A and S4B Fig).

To exclude any potential effect from optical stimulation, we performed a similar set of tests by injecting AAV-DIO-GFP, in which optical stimulations exhibited no detected effect on the movement (S4C Fig). To rule out the possibility of ZI PV neuron collateral activation effect, we repeated the above behavioral test from a different cohort of mice in the following condition: (1) infused TTX (1 μM, 300 nl) into the right-side of ZI through an implanted cannula, and (2) photo-stimulating ZI PV neuronal terminals in the RN through optic fibers implanted in the right-side of RN (Fig 3F, left panel). Under this condition, photo-stimulation is specific to the ZI PV to the RN projection and there should not be antidromic activation of ZI PV soma. We found that this manipulation induced rightward bias in mouse behavior (Fig 3F, right panel), indicating that ZI PV projection to the RN could modulate movements directed by auditory cue.

To further validate this observation, in another cohort of adult PV-Cre mice, we infused AAV-DIO-ArchT, a Cre-dependent expression cassette of ArchT, an inhibitory optogenetic gene [25], into the ZI. The experimental procedures were performed as those of ChR2 tests. As expected, when we optogenetically inhibited the PV neurons during the movement period, the performance was biased to the contralateral port (Fig 3E). This is consistent with the findings when we activated the RN neurons (Fig 1F and 1G).

To assess whether the observed behavioral effect is task related, we performed a similar ZI PV terminal photo-stimulation in mice that are not performing the task. We found that stimulating ChR2+ ZI PV terminals in left RN induced no clear directional movements (Fig 3G), indicating that the ZI PV to the RN projection is likely to modulate goal-directed movements.

It has been shown before that PV and Lhx6 neurons in external globus pallidus contribute oppositely to motor suppression [26]. The substantial impact of activation or inhibition of ZI PV neurons projecting to the RN on the task movement motivated us to examine the other population of ZI interneurons. We then determined whether the Lhx6+ neurons in ZI are involved in regulating the task movements. Similar to the experiments on testing the ZI PV neurons, we used Lhx6-Cre mice for optogenetic manipulations. We found that activation of Lhx6 neurons caused a dramatic contralateral bias. In contrast, in a cohort of SST-Cre mice, we found that activation of vZIm SST neurons showed no detectable effect on the task movement (S4D–S4G Fig). Note that the biased effect of LhX6 activation is opposite to that of PV neuronal activation (Fig 3C).

To further understand how ZI PV and Lhx6 neurons are involved in the movements, we used a ChR2-assisting tagging approach [27] to record ZI PV or Lhx6 neurons from mice performing the task. Similar to the strategy described above, we specifically express ChR2 in ZI PV (or Lhx6) neurons using PV-Cre (or Lhx6-Cre) mice. These neurons can be activated by blue light pulses with short latencies and follow high-frequency light stimulations (S5A–S5C Fig). We found that 43% identified PV neurons (n =  30 neurons) displayed contralateral movement preference, 7% displayed ipsilateral movement preference, and the rest with no preference (S5D Fig). On the contrast, 42.9% identified Lhx6 neurons (n =  8 neurons) displayed ipsilateral movement preference, 14.2% displayed contralateral preference, and the rest with no preference (S5E Fig). All the identified SST neurons (n = 11) displayed no preference (S5F Fig). Together, these results indicated that PV and Lhx6, but not SST neurons in the vZIm are preferentially activated in contralateral and ipsilateral movements directed by auditory decisions, respectively. This is consistent with optogenetic manipulation results in Fig 3.

The findings that ZI Lhx6 neurons do not directly project to the RN (Fig 2E) and activations of ZI PV and Lhx6 neurons caused opposite bias directions in task movements (Figs 3 and S4), led us to determine the possibility that ZI Lhx6 neurons regulate task movement through inhibiting ZI PV neurons locally.

To test this hypothesis, we expressed ChR2 in ZI Lhx6 neurons and recorded light-evoked neuronal activity in the ZI using tetrode recording method (Fig 4A). Similar to the optogenetic tagging method with optrode as described previously [27], neurons activated by light pulses with short latencies and also able to follow with high-frequency light stimulations were considered as ChR2 expressing Lhx6 neurons (see section “Materials and methods”). Out of the total 72 detected units, we found 2 units were activated by the light pulses and followed the 20 Hz light stimulation (Fig 4B, upper panel), thus they were defined as ChR2+ neurons. The activity of units that were inhibited or unchanged by the light pulses are defined as ChR2– neurons (Fig 4B, lower panel). Among these 70 ChR2– neurons, 54 units showed strong light-evoked inhibitory effects (Fig 4C and 4D). Because most inhibitory neurons in the vZIm are PV expressing neurons (Fig 2D), we expected most of the 54 units were PV expressing neurons, suggesting that Lhx6 neuron activation likely suppressed PV neuron activity.

Fig 4. ZI-Lhx6 neurons inhibited other ZI neurons.

(A) Schematic of tetrode recording in ZI where Lhx6 neurons express ChR2. (B) Raster plots show two neurons’ responses to photostimulation. Upper panel: ChR2+ neuron; lower panel: ChR2– neuron. 5 ms light pulses for 500 ms at 20 Hz. (C) All recorded neurons from the experiment illustrated in (A). Color-coded neural traces were normalized to the peak of each cell’s mean firing rate and sorted by modulation index during the light stimulation. (D) Average z-score of ChR2+ neurons (red, n = 2 of 72 across 4 mice), and inhibited ChR2– neurons (blue, n = 54 of 72) during photostimulation. 2/72 with activation vs. 54/72 with silencing; χ² stat = 54.3354; p = 1.6909e – 13. (E) Schematic of tetrode recording in ZI where PV neurons express ChR2. (F) Raster plots show two neurons’ responses to photostimulation. Upper panel: ChR2+ neuron; lower panel: ChR2– neuron. 500 ms light stimulation at 20 Hz. (G) All recorded neurons from the experiment illustrated in (E). Color-coded neural traces were normalized to the peak of each cell’s mean firing rate and sorted by modulation index during the light stimulation. (H) Average z-score of ChR2+ neurons (red, n = 8 of 88 across 4 mice), and inhibited ChR2– neurons (blue, n = 1 of 88) during photostimulation. 8/88 with activation vs. 1/88 with silencing; χ² stat = 5.7379; p = 0.0166. (I) Left: An example of whole-cell patch recording from ZI inhibitory neurons where Lhx6 neurons express ChR2. Blue tick, 5 ms photo stimulus to activate ChR2. PTX: picrotoxin applied through bath solution. Right: Population summary of inhibitory postsynaptic current (IPSC) amplitude (n = 7). (J) Left: An example of whole-cell patch recording from ZI inhibitory neurons where PV neurons express ChR2. Blue tick, 5 ms photo stimulus to activate ChR2. Right: Population summary of inhibitory postsynaptic current (IPSC) amplitude (n = 9). (K) Proposed working model of ZI-RN circuit in controlling contralateral movements. Underlying data can be found in the S1 Data.

https://doi.org/10.1371/journal.pbio.3003092.g004

To examine whether the intra-ZI connections are a common feature of ZI interneurons, we performed the same tests as for Lhx6 on PV neurons. Among the 88 recorded neurons, 8 of them were ChR2+ and 80 were ChR2–. In contrast to the Lhx6 activation experiment, when we activated ChR2 expressing ZI PV neurons, we found no obvious effect on ChR2 negative neurons in the ZI (Fig 4E–4H). Only one neuron showed a potential inhibitory response (Fig 4H, lower panel). To validate the possibility that Lhx6 neurons form robust synapses with local neurons, we performed in vitro slice recordings. We found that the activation of Lhx6 but not PV neurons induced local postsynaptic activation (Fig 4I and 4J), suggesting that Lhx6 neurons form strong intra-ZI synaptic connections. Because most inhibitory neurons in the vZIm are PV expressing neurons (Fig 2D), we speculated that some of the 43 inhibited neurons in Fig 4C and 4D were PV expressing neurons, suggesting that Lhx6 neuron activation likely suppressed PV neuron activity.

Altogether, the findings in Figs 3 and 4 suggest Lhx6 neurons may likely exert their regulation to the task movement via inhibiting ZI PV neurons locally. In this task, PV neurons in the ZI may integrate local activities (i.e., from Lhx6 neurons) and then impact RN activity to regulate the movement to the reward ports, as proposed in Fig 4K.

Animal procedures were approved by the Stony Brook University Animal Care and Use Committee and carried out in accordance with the National Institutes of Health standards (approval number: 824,397). Male and female C57BL/6J mice (Charles River), PV^Cre (JAX strain #017320), SOM^Cre (JAX strain #013044), Lhx6^Cre ([33]), and GAD^Cre (JAX strain #010802) were housed with free access to food, but water restricted after the start of behavioral training. During training, water was available based on task performance (2.5 µl for each correct trial); whereas during non-training days water bottles were provided to the mice for at least 1 h per day.

The mice were placed on a water deprivation schedule and trained to perform an auditory 2AFC task in a single-walled sound-attenuating training chamber as described previously. Behavioral system is controlled, and data is analyzed by Bpod system (Sanworks, https://sanworks.io/ provides all source files). In brief, freely moving mice were trained to initiate a trial by poking into the center port of a three-port operant chamber, which triggered the presentation of a stimulus. Subjects then selected the left or right goal port. The cloud-of-tones stimulus consisted of a stream of 30 ms overlapping pure tones presented at 100 Hz. The stream of tones continued until the mouse withdrew from the center port. Eighteen possible tone frequencies were logarithmically spaced from 5 to 40 kHz. For each trial, either the low stimulus (5–10 kHz) or high stimulus (20–40 kHz) was selected as the target stimulus, and the mice were trained to report low or high by choosing the correct side of port for water reward. Correct responses were rewarded with water (2.5 µl for each correct trial), and error trials were punished with a 4 s timeout. Sound intensity was calibrated at 65 dB SPL. The evidence strength r determined the difference in the rate of high and low octave tone in the stimulus. Tones were drawn from the target octave with a probability of 1 + 2r/100/3. To quantify mice’s performance in the task, we used a logistic regression model described before [2,3,6]. log(p/(1 – p))   β₀ + r*β₁, where p is the fraction of choices towards the port associated with high frequencies. Parameters β₀ and β₁ measure the bias and slope of the psychometric curve. Reaction time was calculated as the time between the onset of tone and the time of withdraw from the center port. Movement time was calculated as the time between the time of withdraw from the center port and the onset of poking the side port.

Electrical signals in ZI or RN were recorded using Neuralynx Cheetah 32-channel hybrid system and cheetah data acquisition software. Signals were filtered 600–6,000 Hz. Single units were isolated offline using Spike3D and MClust3.5. Clusters with isolation distance >20 and L-Ratio 

To control for the potential influence of visual stimulation effects, during all trials, we included a masking light stimulation using the same presentation time window and wavelength as photo stimulation trials delivered through a bulb placed above the center port, as previously employed [4,6].

At the beginning of each experiment involving identified ChR2-expressing neurons, the probe was lowered to the presumed depth of the ZI through the burr hole and allowed to settle for 10–45 mins. The fiber attached to the probe was coupled to a 473 nm diode-pumped solid-state laser (Shanghai Dream Lasers, Shanghai, China) using an optical multimode fiber (200 µm, 0.39 NA FC/PC, Thorlabs part no. M83L01). Once a stable recording was established, blue light was flashed for 10 ms at 2–10 mW through the fiber into the brain at 20 Hz for 200 repetitions. Laser power was adjusted to minimize the latency of activation while also minimizing optical artifacts. To identify neurons with ChR2-mediated responses, we performed the following three analyses as previously reported [27,34]: (1) distribution of spiking timing; (2) Latency; (3) spike waveform correlation. A unit (neuron) was determined as ChR2-expressing if: (1) we can detect a reliable increase of spike precisely time-locked to the onset of light pulses at 20 Hz, (2) the light response has a latency less than 10 ms, and (3) the spike waveforms have correlation of more than 0.95. Once a neuron was identified as possibly being light active, the locomotor/stimulation session would proceed after which a post identification session would be carried out to ensure another unit had not moved into the recording space. Final clustering was performed post hoc.