Dopamine D1 receptor positive neurons in the lateral nucleus of the cerebellum contribute to cognitive behavior

Abstract

Background

Studies in humans and non-human primates have identified a region of the dentate nucleus of the cerebellum (DCN), or lateral nucleus in rodents (LCN), activated during performance of cognitive tasks involving complex spatial and sequential planning. Whether such a subdivision exists in the rodent is not known. Dopamine and its receptors, which are implicated in cognitive function, are present in the cerebellar nuclei but their function is unknown.

Methods

Utilizing viral and genetic strategies in mice, we examined cellular phenotypes of dopamine D1 receptor positive (D1R+) cells in the LCN with whole-cell patch clamp recordings, mRNA profiling, and immunohistochemistry to examine D1R expression in mouse LCN and human DCN. We used chemogenetics to inhibit D1R+ neurons, and examined behaviors including spatial navigation, social recognition memory, prepulse inhibition of the acoustic startle reflex (PPI), response inhibition, and working memory to test the necessity of these neurons in these behaviors.

Results

We identified a population of D1R+ neurons that are localized to an anatomically distinct region of the LCN. We also observed D1R+ neurons in the human DCN, suggesting an evolutionarily conserved population of dopamine-receptive neurons in this region. The genetic, electrophysiological, and anatomical profile of mouse D1R neurons is consistent with a heterogeneous population of GABAergic, and to a lesser extent glutamatergic cell types. Selective inhibition of D1R+ LCN neurons impairs spatial navigation memory, response inhibition, working memory, and PPI.

Conclusion

Collectively, these data demonstrate a functional link between genetically distinct neurons in the LCN and cognitive behaviors.

Introduction

The cerebellum is well known for its role in coordinating motor output and adapting involuntary reflexes to support sensory prediction-error based learning(1–3). A lesser appreciated, but no less important function of the cerebellum is its role in goal-directed, cognitive functions. Consistent with this role, numerous neuropsychiatric disorders and developmental syndromes are associated with alterations in cerebellar function, including schizophrenia, autism, frontotemporal dementia, and Joubert syndrome(4–7). In many individuals with altered cerebellar anatomy, cognitive and behavioral perturbations occur in the absence of gross motor impairments(8). Several of these disorders are associated with alterations in DCN/LCN anatomy and gene expression(4, 5, 9).

The LCN is the most lateral of the cerebellar nuclei and is the major output nucleus of the neocerebellum(10). In addition to classical cerebellar output pathways to thalamic nuclei and red nucleus, LCN has direct, reciprocal connections with limbic circuitry, such as the ventral tegmental area (VTA)(11–16). Consistent with this connectivity, functional imaging and anatomical mapping studies in primates demonstrate both motor and cognitive regions within the LCN(16–19). For example, in humans, DCN activation increased 3–4 fold during attempts to solve a pegboard puzzle compared to simply moving these pegs in a functional MRI study¹⁸. Neocerebellar lesions in humans can elicit changes in affect, spatial navigation, working memory (WM), behavioral flexibility, and interval timing(20–22), and lesions of lateral cerebellar cortex (CCtx) and LCN in rodents impair spatial learning and memory(23–25), reduce reinforcement learning and motivation(26), and impair absolute timing(27). Several human disorders with cerebellar dysfunction are associated with deficits in social behavior(6, 7). Aberrant cerebellar development in rodent models is also associated with altered WM and social behavior without motor impairment(28, 29). Collectively, these data point to a key role of the cerebellum in the regulation of cognitive and social behaviors. Given the strong correlation between cerebellar function and mental illness, there is a strong desire to understand the relationship between cerebellar anatomy and function and the etiology of symptom domains of these disorders.

Multiple cell types reside within the LCN, which have unique morphological and projection profiles(30–41). Input mapping and gene-mapping studies have revealed distinct cerebellar nuclei subregions, including at least 2 subdivisions of the LCN in rodents(42, 43). However, studies interrogating the functional specificity of these region- or cell type-specific cerebellar divisions are lacking.

Dopamine and dopamine receptors are broadly implicated in mental illness(44) and localize to microanatomical regions of the cerebellum(45–48). Furthermore, in rodents, cerebellar nuclei are reported to have dopamine at concentrations greater than in hippocampus or cerebellar cortex, and similar levels relative to frontal cortex(49). Tyrosine hydroxylase, the rate-limiting enzyme in catecholamine synthesis, is found in fibers projecting to all cerebellar lobules, laminae, and nuclei in mouse(48). D1R is implicated in cognitive functions including WM, spatial navigation and cognitive flexibility, but whether D1Rs in cerebellum could regulate these is unknown(50, 51). Based on these observations, we hypothesized that D1Rs may identify subdivisions within LCN that would allow for manipulation and interrogation of LCN function, with relevance to cognition in neuropsychiatric disorders. Here, we confirm the presence of D1R+ neurons in human DCN and mouse LCN. Molecular profiling, projection analysis, and electrophysiological characterization of these neurons reveal a partially heterogeneous population of neurons that is largely inhibitory. Consistent with a role in cognition, cell-specific chemogenetic silencing of these neurons alters spatial navigation memory, response inhibition, WM, and sensorimotor gating, thus demonstrating a cellular specialization within the LCN for these functions.

Methods

Standard techniques for characterization of LCN D1R+ neurons and their role in behavior

We used several standard molecular and cellular techniques including immunohistochemistry, RiboTag, quantitative PCR (qPCR), Slice electrophysiology, and in vivo electrophysiology for characterizing cellular phenotypes, as well as several cognitive, motivational, sensory and motor behavioral tasks. These techniques are described in detail in supplemental materials.

Results

D1R Neurons Occupy a Specific Subregion of the LCN

We confirmed D1R expression in subsets of cells in the DCN of human and mouse. (Figure 1A–C, Supplemental Figure S2A–B). Immunohistochemical analysis suggests that D2R is broadly expressed in the mouse LCN, whereas D1R has a lower, regionally restricted expression (Figure 1D, Supplemental Figures S3A, C, E, S4A–P). D1R staining was not present in striatum or LCN of Drd1a^Cre/Cre knockout mice, but was present in these regions in WT littermate controls, validating the antibody (Supplemental Figure S3A–G, S4A–P). To determine whether D1R would isolate discrete populations of neurons in the mouse LCN, we performed an anatomical analysis of virally-labeled D1R neurons. Cell-specific expression of yellow fluorescent protein (YFP) was achieved by injecting a conditional adeno-associated viral (AAV) vector (AAV-FLEX-YFP) into the LCN of Drd1a^Cre/+mice (Figure 1E). YFP-positive neurons were counted along the rostral-caudal extent of the LCN (Figures 1F–G), with the highest numbers of labeled neurons localized to caudal, medial, and ventral zones of the LCN (Figure 1H–J). Because this method of identifying cells could be confounded by viral spread, we performed a similar experiment using Drd1a^Cre/+mice crossed with a conditional tdTomato reporter line(52). Tomato-positive neurons were counted along the rostral-caudal extent of the LCN (Supplemental Figures 1B–C), with a similar pattern of the highest number of labeled neurons localized to caudal, medial, and ventral zones of the LCN (Supplemental Figures 1D–F). Though a complete stereological assessment of their anatomical distribution in humans could not be performed, these data strongly support a highly conserved population of D1R expressing neurons in the cerebellar nuclei.

Figure 1.

Immunohistochemistry reveals staining for D1R in human DCN and mouse LCN. Mapping reveals subregional localization of D1R neurons in LCN.

A, Immunohistochemistry for the D1R protein in human DCN (in a parasagittal section), square in I is zone at higher magnification in B. Staining for D1R was positive in each of 5 cases.

C, Staining of D1R in the green color channel in lateral nucleus of the cerebellum in a coronal section.

D, Staining of D2R in the red color channel in lateral nucleus of the cerebellum in a coronal section.

E, Illustration depicting the lateral (dentate) nucleus (DN/LAT) from the rostral to caudal extent that was analyzed for D1R neuron location.

F, Illustration of the rostral region of LCN in coronal plane (top), representative image of D1R:Tmto expression in rostral LCN overlaid with divisions of dorsal, ventral, medial and lateral zones (bottom). Dorsal and lateral orientations are denoted in inset. Scale = 120 μm.

G, Illustration of the caudal region of LCN in coronal plane (top), representative image of D1R:Tmto expression in caudal LCN overlaid with divisions of dorsal, ventral, medial and lateral zones (bottom), In addition to the LCN, few cells were observed in the parvocellular region of the LCN and the interposed nuclei. (DN = Dentate (or Lateral) Nucleus; IntA = Interposed Nucleus Anterior part; IntP: Interposed Nucleus, posterior part; PC: Dentate (or Lateral) Nucleus, parvocellular part; Y: Nucleus Y of the vestibular complex).

H, Quantification of distribution of D1R:YFP positive cells in rostral vs. caudal zones (t=7.42, df=29).

I, Quantification of distribution of D1R:GFP positive cells in medial vs. lateral zones (t=19.2, df=56).

J, Quantification of distribution of D1R:GFP positive cells in dorsal vs. ventral zones (t=2.83, df=54). Results were acquired from 4 mice, 4 sections per mouse counted bilaterally and are represented as the average number of cells per section per side or the percentage of cells within each region per section per side.

Illustrations in A–C are from Paxinos and Franklin, 2013. ** P < 0.01, **** P < 0.0001, Student’s t test, two-tailed.

Electrophysiological Properties of D1R LCN Neurons in Slice

Different cell types within the LCN have distinct electrophysiological profiles(32, 36, 37, 40, 41). To characterize the electrophysiological properties of D1R neurons, we performed whole-cell patch clamp recordings on fluorescently identified neurons from acute cerebellar slices. D1R cells had diverse intrinsic properties and could be sorted into two categories (Type I [62% of cells] and Type II [38% of cells]) based on properties of action potential (AP) waveforms, including half-width and time to after hyperpolarization (AHP) (Figure 2A). AP threshold, peak amplitude, and AHP amplitude did not differ between populations (Figure 2B). Type I neurons had broad APs with a slow AHP, while Type II neurons had narrow APs with a fast AHP (Figure 2C–D). The majority of Type I neurons were spontaneously active (15/16), whereas the majority of Type II neurons only fired APs following current injection (7/10) (Figure 2E).

Figure 2.

Electrophysiological characterization of D1R LCN neurons in mice.

A, Average AP waveforms of Type I and Type II neurons, Scale = 20 mV, 2 ms.

B, AP Threshold, AP Peak, and AHP Peak in Type I and Type II neurons.

C, Time to AHP peak in Type I and Type II neurons. ***P < 0.001, Student’s t test, two-tailed, t=4.04, df=24.

D, AP half-width of Type I and Type II neurons. **** P < 0.0001, Student’s t test, two-tailed, t=5.33, df=24.

E, Example Type I (top) and Type II (middle) neurons before, during, and after 50 pA current injection (bottom), Scale = 20 mV, 200 ms.

F, Capacitance of Type I and Type II neurons. G–L: Type I N = 16; Type II N = 10, **** P < 0.0001, Student’s t test, two-tailed, t=5.87, df=24.

G, Distribution of measured surface areas of D1R neurons in the DNC revealing a non-normal distribution (N = 306 cells, Shapiro-Wilk, P < 0.05).

Membrane capacitance of Type I neurons was significantly smaller than Type II neurons (Figure 2F). Consistent with two (or more) distinct cell sizes, histological measurements of cell surface area revealed a significantly skewed, non-normal distribution with the majority of neurons being of smaller size (Figure 2G).

Molecular Characterization and Projections of D1R LCN Neurons

Our electrophysiological results were remarkably similar to established properties of two neuronal populations: local glycinergic/GABAergic neurons and nucleocortical projecting glycinergic neurons(40), and possibly a third type of neuron, putatively glutamatergic(41) (three Type II neurons differed in that they did fire spontaneous APs, similar to a described population⁴¹). To determine the neurotransmitter phenotype of LCN D1R neurons we selectively isolated actively translating mRNA utilizing a RiboTag approach(53). Cell-specific expression of the affinity-tagged ribosomal protein Rpl22-HA was achieved by injecting a conditional AAV vector (AAV-FLEX-Rpl22-HA(53)) into LCN of Drd1a^Cre/+ mice (Figure 3A).

Figure 3.

Translational profiling and immunohistochemistry reveal identities of D1R LCN neurons in mice.

A, Expression of Rpl22-HA in D1R neurons, Scale bar = 60 μm (inset scale bar = 10 μm), in the LCN.

B, Schematic of RiboTag methodology: Following cell lysis (1), HA antibody-coupled magnetic beads immuno-isolate tagged polysomes and associated mRNA (2). mRNA are isolated (3) and cDNA is generated from both input and immunoprecipitated mRNA.

C, qRT-PCR analysis of immunoprecipitate relative to input demonstrating significant enrichment of Drd1a, Drd2, Vgat, Gad1, Glyt2, and Penk (enriched markers in black), relative to Cnp, an oligodendrocyte marker in D1R cells of LCN. Drd3, encoding the dopamine D3 receptor, and oligodendroglial marker (Cnp) were de-enriched (white), while the marker of glutamatergic neurons (Vglut2), was neither enriched or de-enriched (grey). ****P < 0.0001, ** P < 0.01, *P < 0.05, One-way ANOVA, n = 3 pooled samples of 7 mice/pool (F_{9, 20} = 15.3).

D, qRT-PCR analysis of immunoprecipitate relative to input demonstrating significant enrichment of Drd2, Penk, Glyt2, Gad1, and Vgat (enriched markers in black), relative to Cnp, in Vgat+ cells of LCN. Cnp, Vglut2, and Drd3 were de-enriched (white), while Drd1a was neither enriched or de-enriched (grey). ****P < 0.0001, ** P < 0.01, *P < 0.05, One-way ANOVA, n = 4 pooled samples of 4 mice/pool (F_8,32 = 20.1).

E, qRT-PCR analysis of immunoprecipitate relative to input demonstrating significant enrichment of Drd3, in Vglut2+ cells (black). Glyt2, Gad1, and Vgat, were de-enriched (white), while Drd1a, Drd2, and Penk, were neither enriched nor de-enriched (grey). ****P < 0.0001, ** P < 0.01, *P < 0.05, One-way ANOVA, n = 4 samples of 1 mouse/sample (F_8,32 = 23.3).

F, Venn diagram illustrating distribution of D1R expression in neural subtypes residing in the LCN.

Following immunoprecipitation and isolation of polyribosomal-associated mRNA (Figure 3B), qPCR demonstrated significant enrichment of Drd1a expression relative to total input mRNA (Figure 3C), confirming the efficacy of the enrichment in Drd1a^Cre/+ mice. We also observed enrichment of Drd2 expression (Figure 3C), implying some degree of specific co-expression of D1R and D2R in LCN neurons. We found no enrichment of Drd3 (Figure 3C). Neurons in the striatum co-expressing D1R and D2R express the neuropeptide enkephalin, similar to canonical D2R expressing neurons(54); D1R LCN neurons had enrichment of the proenkephalin (Penk) mRNA (Figure 3C). Additionally, we observed enrichment of several markers of inhibitory neurotransmission, i.e., mRNA for the GABA synthesizing enzyme Gad67 (Gad1), the vesicular GABA and glycine transporter Vgat (Slc32a1), and the membrane glycine transporter GlyT2 (Slc6a5; Figure 3C). Surprisingly, expression of the vesicular glutamate transporter Vglut2 (Slc17a6) was neither enriched nor de-enriched, suggestive of a third, minor population of cells co-expressing both Vglut2 and Drd1a. To clarify this issue, we performed a similar analysis in Vgat^Cre/+ and Vglut2^Cre/+ mice (Figure 3D–E). In Vgat^Cre/+ mice, we observed enrichment of Drd2, but only partial enrichment of Drd1a. In Vglut2^Cre/+ mice, we found enrichment only of Vglut2 and Drd3, implying that Vglut2 neurons make up only a small percentage of the D1R population. These data indicate that D1R neurons are a regionally restricted population of predominantly inhibitory neurons, with a much smaller proportion being glutamatergic (Figure 3F).

To determine projections of D1R neurons, we co-injected Cre-dependent AAVs encoding mCherry (AAV-FLEX-mCherry) to fill cells and a green fluorescent protein (GFP)-tagged synaptophysin (AAV-FLEX-synapto-GFP) to label axon terminals (Figure 4A–O). The most prominent projections were a number of small synapto-GFP positive puncta within LCN (Figure 4D), concordant with local connectivity. We detected large clusters of synapto-GFP in the granular layer of CCtx, most prominently in areas of lateral cerebellar cortex Crus I, Crus II, paraflocculus and flocculus (Figure 4G). Higher magnification imaging of these synapses revealed morphology similar to previously reported nucleocortical rosettes(34, 35, 39) (Figure 4J), consistent with glutamatergic projections to CCtx(30, 34, 35). We also observed a small number of bead-like synapses in the granular and Purkinje cell layers (Figure 4K), consistent with projections from the LCN, which may represent GABA/glycinergic projections(30).

Figure 4.

A–P, Projection patterns of D1R LCN neurons in D1R-Cre mice compared with Vgat-Cre and Vglut2-Cre mice.

A, Illustration depicting the lateral dentate nucleus (DN/LAT) location of injection for Synapto-GFP and mCherry viral constructs (B, C) depicted in D–G.

B, Illustration of Synapto-GFP viral construct.

C, Illustration of mCherry viral construct.

D–F, Histochemistry demonstrating synaptophysin-GFP (Syn-GFP) expression in DNC of Drd1a^Cre/+ Vgat^Cre/+, and Vglut2^Cre/+ mice, respectively; scale bar represents 60 μm.

G–I, mCherry expressing neurons are observed in the cerebellar cortex along with a number of Syn-GFP puncta. Scale bar represents 200 μm.

J–O, Nucleocortical rosette-like or Bead-like synapses in cerebellar cortex in each of three strains of mice. Scale bar represents 2 μm.

Previous studies have shown that GABAergic cells in the LCN project to inferior olive(37) and locally(36, 40), whereas glycinergic cells project to CCtx with small, bead-like synapses(30). In contrast, glutamatergic cells project to thalamic targets and to CCtx where they form large mossy fiber-like rosette-synapses(34, 35). Analysis of projections from Vglut2^Cre/+, and Vgat^Cre/+ mouse lines (Figure 4E–O) confirmed nucleocortical rosettes in Vglut2^Cre/+ mice. In Vgat^Cre/+ mice we observed local projections and bead-like synapses in CCtx (Figure 4D–O). D1R+ projections were found in extracerebellar targets such as ventromedial thalamic nucleus, VTA, and locus ceruleus. A complete list of extracerebellar projections in Drd1a^Cre/+, Vglut2^Cre/+, and Vgat^Cre/+ mice is presented in Table 1 (example sections with synapto-GFP positive puncta shown in Supplemental Figure S5).

Table 1.

Brain Regions Targeted by Lateral Nucleus of the Cerebellum Cell Populations

Structure	Genotype
	D1R-Cre	Vgat-Cre	Vglut2-Cre
Diencephalon
Nucleus of the Vertical Limb of the Diagonal Band	+	−	+
Lateral Preoptic Area	+	−	+
Medial Forebrain Bundle	+	−	+
Nucleus Reuniens	+	−	+
Zona Incerta	++	−	++
Ventral Anterior/Ventral Lateral Nucleus of the
Thalamus	−	−	++
Ventral Medial Nucleus of the Thalamus	+	−	+++
Prerubral Field	+	−	++
Midbrain
Superior Cerebellar Peduncle (fibers)	++	++	+++
Prosomere 1 Reticular Formation	++	−	+
Mesencephalic Reticular Nucleus	++	−	+
Red Nucleus	++	+	+++
Retrorubral Fields	+	−	++
Ventral Tegmental Area	+	+	+
Pons
Reticulotegmental Nucleus of the Pons	+	+	+++
Basilar Pontine Nuclei	++	+	+
Pedunculopontine Tegmental Nucleus	+	+	+
Medial Lemniscus	+	++	+
Gigantocellular Reticular Nucleus	+	+	+
Locus Ceruleus	+	+	+
Medulla
Superior Vestibular Nucleus	+	+	+
Medial Vestibular Nucleus	+	+	−
Lateral Vestibular Nucleus	+	+	−
Inferior Olive, Primary Nucleus	++	++	−

+ = Weak, ++ = Moderate, +++ = Strong Projection in 4/4 animals examined in each group.

Inhibition of D1R LCN Neurons with DREADDs

To assess the function of D1R LCN neurons, we inhibited these cells through conditional viral-mediated expression of the inhibitory Designer Receptor Exclusively Activated by a Designer Drug (DREADD) Hm4Di(55) fused to YFP (AAV-FLEX-Hm4Di-YFP) (Supplemental Figure S6A–B). Bath application of the selective ligand clozapine-n-oxide (CNO; 5 μM) to acute cerebellar slices reduced the firing frequency of spontaneously active YFP-positive cells by an average of 44.3 ± 11.9% (12 cells) and hyperpolarized non-spontaneous cells by an average of 4.0 ± 1.3 mV (6 cells) (Supplemental Figure S6C).

Analysis of the in vivo effects of CNO-mediated inhibition (1 mg/kg; intraperitoneal injection of CNO) using chronic tetrode recordings revealed a number of cells (13/42) that were inhibited, while a similar number (13/42) were excited, and the remainder were unaffected (17/42) (Supplemental Figure S6D–G). Based on observations of local, likely inhibitory, projections of D1R neurons within the LCN, neurons activated by CNO are likely a reflection of disinhibition caused by silencing inhibitory D1R cells.

D1R LCN Neurons Influence Spatial Navigation

Lesions to LCN disrupt spatial memory(23–25). To establish whether D1R LCN neurons influence spatial memory, we inhibited D1R LCN neurons during a Barnes Maze task. Experimental D1R:Hm4Di and Control D1R:GFP mice were both pretreated with CNO (1 mg/kg; intraperitoneal injection) prior to behavioral assessment during both training and memory recall. During training, both groups showed improvements in performance (decreases in total distance traveled, center crossings and latency to goal; Supplemental Figure S7A–C), but D1R:Hm4Di mice showed deficits in measures of learning such as time spent and nose pokes in the target quadrant during training trials relative to littermate D1R:GFP controls (Figure 5A). We next probed mice for memory recall following removal of the escape tunnel; D1R:Hm4Di mice had reduced nose pokes in the goal and the goal quadrant (Figure 5B–C). Since we observed differences between groups in learning across training, in a separate cohort of mice we injected CNO only prior to the probe trial. In this experiment, D1R:Hm4Di mice showed a difference on only one measure during training (D1R:Hm4Di mice had increased duration in target quadrant on day 2), but not nose pokes in target quadrant (Supplemental Figure S7D). There were no differences in performance on the probe trial when CNO was administered (Supplemental Figure S7E–F). Thus, the effect of CNO treatment on time spent in the target quadrant during training is not specific to neuronal silencing. In contrast, the effect of silencing on nose poke accuracy is specific to neuronal silencing as differences in this measure throughout training (and in the probe trial) occurred only when CNO was given throughout training.

Figure 5.

DREADD Receptor expression in D1R LCN neurons and CNO application results in altered performance on the Barnes Maze.

A, Schema for when CNO was injected IP prior to training and probe tests above performance of each group during training trials, as measured by average duration in the goal quadrant (left) and as measured by nose pokes in target quadrant holes and goal hole (right). N = 17 D1R:Hm4Di mice, and N = 23 littermate D1R:GFP control mice. For goal quadrant duration, Error bars are SEM. * P < 0.05 (Two-way rmANOVA, 2 factors were significant: Interaction between training and presence of Hm4Di (F_3,114 = 3.14, *P < 0.05) and training (F_3,114 = 5.3, **P < 0.01). the factor of presence of Hm4Di alone was not significant. Holm-Sidak’s post hoc multiple comparisons test did not indicate any difference in target quadrant duration for any days of training. For nose pokes in goal quadrant, (Two-way rmANOVA) 2 factors were significant: Interaction between training and presence of Hm4Di (F_3,114 = 3.92, *P < 0.05) and Training (F_3,114 = 8.49, *P < 0.0001); The factor of presence of Hm4Di alone was not significant. Holm-Sidak’s post hoc multiple comparisons test indicated that the difference in nose pokes between groups during training was only significantly different on day 4 of training, *P < 0.05, DF =152.

B, Performance of each group during memory recall during the probe trial of the Barnes maze, as measured by nose pokes in target quadrant holes. N = 17, D1R:Hm4Di mice and N = 23 littermate D1R:GFP control mice. Error bars are SEM. * P < 0.05 Unpaired Student’s t test, two tailed, t=2.52, df=38.

C, Performance of each group during memory recall during the probe trial of the Barnes maze, as measured by nose pokes in goal hole. D1R:GFP had significantly more nose pokes in goal hole than D1R:Hm4Di mice on probe trial, * P < 0.05, Unpaired Student’s t test, two tailed, t=2.23, df=38.

D, Representative path traces of D1R:GFP after IP injection of CNO and E, D1R:Hm4Di after IP injection of CNO on the probe trial on Day 5 of the Barnes Maze.

F, Social approach as measured by time spent in arena with a novel mouse or novel object. Both groups preferred social interaction (F_1,72 = 48.52, P < 0.0001). (N = 22, D1R:GFP mice, and N= 17, D1R:Hm4Di mice, Two-way ANOVA). No significant differences were found for presence of Hm4Di or interaction (presence of Hm4Di X zone).

G, Social preference as measured by time spent in arena with a novel mouse or familiar mouse. Factors for interaction (Presence of Hm4Di X zone, F_1,72 = 4.59, *P < 0.05) and zone (F_1,72 = 4.19, P < 0.05) were significant (N = 22, D1R:GFP mice, and N= 17, D1R:Hm4Di mice, **P < 0.01, Two-way ANOVA, Sidak’s multiple comparisons test).

H, Prepulse inhibition of the acoustic startle reflex (N = 22, D1R:GFP mice, and N= 17, D1R:Hm4Di mice), was significantly different for presence of Hm4Di (F_1,37 = 6.95, *P < 0.05) and dB of prepulse, P < 0.0001, F_2,74 = 44.22), but not interaction, Two-way rmANOVA.

To establish whether spatial navigation deficits in mice with inhibited D1R LCN neurons is due to motor incoordination, CNO-treated D1R:Hm4Di and D1R:GFP mice were assayed on an accelerating rotarod task. D1R:Hm4Di mice showed no deficits in performance relative to D1R:GFP mice across multiple days of training (Supplemental Figure S7G). We also analyzed gait variability and found no difference between groups on measures of stance and stride (Supplemental Figure S7H).

D1R LCN Neurons Influence Social Recognition Memory and Sensorimotor Gating Behaviors

Mental illnesses with associated changes in cerebellar function are associated with decreased social cognition, and alterations in prepulse inhibition of the ASR (PPI)(56, 57). To test social approach, interaction, and preference we used a three-chamber assay(56). Both CNO-treated D1R:Hm4Di and D1R:GFP mice showed preference for exploring a chamber and in sniffing zone containing a mouse compared to one containing a novel object (Figure 5F, Supplementary Figure S7J). We next replaced the object with a novel mouse, establishing a familiar mouse in one chamber and an unfamiliar mouse in the second chamber. While D1R:GFP mice spent significantly more time in the chamber with the novel mouse, D1R:Hm4Di mice failed to discriminate in this task (Figure 5G, Supplemental Figure S7K). Deficits in social preference in D1R:HM4 mice were not associated with an overall reduction in exploratory behavior (Supplemental Figure S7L–M).

Although motor coordination was not altered in D1R:Hm4Di mice, we did observe a significant deficit in PPI (Figure 5H), primarily at the lowest amplitude prepulse (70dB). This is consistent with a deficit in sensorimotor gating and attentional processes(57). The amplitude of responses to different amplitudes of startle pulses in the absence of acoustic prepulse did not differ between groups, indicating intact basic motor reflexes in D1R:Hm4Di mice (Supplementary Figure S7I). Cerebellar modulation of PPI in mice has been documented previously(58).

D1R LCN Neurons Influence Temporally-Dependent Response Inhibition

A classic function attributed to dopamine in the prefrontal cortex is temporally-dependent response inhibition(59). To determine whether D1R LCN neurons modulate temporally-dependent response inhibition, we used a paradigm known as differential reinforcement of low rate responses (DRL) which reinforces a subject’s ability to refrain from responding for a set time period. Optimal performance requires a subject to accurately time the interval between lever presses, and inhibit responses before a set time has elapsed (Figure 6A–B). CNO-injected D1R:Hm4Di male mice had right-shifted responses compared to controls, increased response inhibition in the 10-second interval version of this task (DRL-10) (Figure 6C). The peak (mode) of the response distribution for each animal was averaged by group and was significantly increased for D1R:Hm4Di mice on DRL-10 trials (inset, Figure 6C). Assessment of this behavior in female mice showed similar deficits in D1R:Hm4Di mice, though the overall training time required for controls to perform the task was greater than in males (Supplemental Figure S8A–F).

Figure 6.

Peak Interval Timing and Working Memory are regulated by D1R neurons in the LCN.

A and B, Schematics describing contingencies of DRL operant timing task.

C, Distribution of latencies of lever presses after reward in the DRL-10 for D1R:GFP control mice (black circles) and D1R:Hm4Di mice (red diamonds). Group means ± SEM are presented for latencies binned in 2s intervals. Dotted line represents minimum response latency that is rewarded for the DRL paradigm (2-way RM ANOVA, Presence of Hm4Di x time interaction: Week 3: F_29,261 = 5.61, P < 0.0001; N = 6, D1R:Hm4Di mice and N = 5, D1R:GFP mice). Inset is the mean peak of response distribution of each group for this week of training in this 10 second DRL paradigm.

D, Schematic describing the delayed alternation operant task.

E–J, Performance of D1R:GFP control mice (black lines) and D1R:Hm4Di mice (red lines) on increasingly difficult versions of the task as measured by proportion correct (E–G) and pellets rewarded (H–J). Group means ± SEM are presented in E–J. E, 2 second delay, only significant factor was for training: F_5,140 = 101.5, P < 0.0001; no significant differences between groups, or interaction between training and presence of Hm4Di. N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice. F, 8 second delay, significant differences for factors of training: F_4,112 = 16.16, P < 0.0001, and between groups, F_1,28 = 7.2, *P < 0.05, but not interaction. N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice. G, 16-second delay, factor of presence of Hm4Di was significant: F_1,28 = 4.83, *P < 0.05. Interaction between groups and training were not significant. N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice. H, 2 second delay, only significant factor was for training: F_5,140 = 50.92, P < 0.0001; no significant differences between groups or interaction between presence of training and presence of Hm4Di. N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice. I, 8 second delay, no significant differences for factors of training, Hm4Di groups, or interaction between groups and training were found. N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice. J, 16-second delay, factor of presence of Hm4Di was significant: F_1,28 = 4.35, *P < 0.05. Interaction between groups and training was not significant. Two-way rmANOVA N = 12, D1R:Hm4Di mice and N = 18, D1R:GFP mice.

The altered performance of the D1R:Hm4Di male mice was not due to changes in motivation, or ability to lever press, as performance in instrumental conditioning (FR1), a progressive ratio reinforcement task, as well as overall pellets rewarded in DRL was equivalent between groups (Supplemental Figure S9A–C). Female D1R:Hm4Di mice had significantly fewer pellets rewarded when performing DRL (Supplemental Figure S9D).

D1R LCN Neurons Influence Working Memory

Lateral CCtx and LCN in species ranging from rodents to humans are also implicated in WM(18, 19, 23). Thus, we chose the delayed alternation (DA) protocol to test WM in rodents, which also requires dopamine(60, 61). This operant paradigm requires a subject to learn to press one of two levers, wait for a delay, and then press the other lever (Figure 6D). CNO-injected D1R:Hm4Di mice showed no differences on acquisition of the task with a 2s delay (Figure 6E, H), but had poorer performance on one measure (proportion correct) at 8s delay (Figure 6F) and on two measures (proportion correct, pellets rewarded) with delay of 16s (Figure 6G, J), indicating D1R cells are necessary for when WM demands are higher.

Discussion

We have identified D1R expression in human DCN, a population of D1R neurons in the mouse ventro-caudal LCN that demonstrate restricted anatomical localization, and converging evidence that activity in this population is required for normal performance in several cognitive domains. This population of LCN D1R neurons in the mouse is similar to previous reports of neurons in ventro-caudal DCN that make up a cognitive domain in monkeys and humans(16, 17, 19, 62). Prefrontal D1Rs have been shown to modulate cognitive behaviors such as interval timing, risk-based decision making, WM, and behavioral flexibility(63–67). Previous studies linking cerebellum to these functions proposed modulation of catecholamines in prefrontal cortex via connections through the thalamus or VTA(68, 69); this is the first to link a specific cerebellar cell population to these behaviors.

D1R LCN neurons appear to be a heterogeneous population containing principally inhibitory neurons, and a small number of excitatory neurons. In cerebellar slices, we observed electrophysiological properties consistent with putative glycinergic, GABA/glycinergic, and glutamatergic neurons(30, 40, 41, 70). D1R cells are enriched with markers of inhibitory neurotransmission, and we observed synapses both within LCN and in CCtx consistent with inhibitory local and nucleocortical projections. D1R cells were neither enriched nor de-enriched in the glutamatergic marker Vglut2, and we observed rosette-like synapses in the cerebellar cortex similar to those seen in Vglut2^Cre/+ mice. We also observed a small number of larger cells in slice that were consistent with glutamatergic neurons. These observations are consistent with a smaller proportion of D1R LCN cells being glutamatergic. The extracerebellar targets of D1R+ and Vglut2+ cell projections lend further evidence to the idea that these cells can regulate cognitive functions, via regulation of brain targets associated with them. The restricted distribution of D1R LCN neurons suggests that this population of cells may represent a specific behavioral control segment involved in cognitive operations. Why encompass inhibitory and excitatory neurons? One possibility is that the glutamatergic D1R neurons represent the minimal essential component for regulation of specific cognitive functions, with the larger number of inhibitory neurons projecting locally within the LCN and back to the cerebellar cortex to constrain output from other behavioral control regions. In this way, performance of specific spatial, WM and temporally-dependent tasks can proceed without competing behavioral processes running concurrently.

When we induced functional inhibition of D1R neurons in the LCN, similar numbers of neurons were inhibited as were excited. The implication of this is that LCN D1R cells are significantly contributing to local inhibition. Previous studies have found that only a few neurons in cerebellar nuclei are sufficient to transmit temporally precise information for a given effector system, and that inhibitory Purkinje cell input is suitable to modulate neurons in the cerebellar nuclei at this level of precision (71). Furthermore, we reported D1R+ nucleocortical synapses in the granular layer of Crus I, Crus II, paraflocculus and flocculus. Crus I and Crus II are classically associated with prefrontal and parietal cortices(72–75) and paraflocculus and flocculus are associated with modulation of eye movements and adaptation of the vestibulo-ocular reflex(76, 77), so D1R cells could possibly exert differential control over different behavioral control regions at the level of cerebellar cortex. Thus, it is likely that D1R cells regulate adjacent neurons via local inhibition, and more distant cerebellar cortical neurons via nucleocortical projections, suggesting more levels of cerebellar output regulation, potentially by extracerebellar catecholaminergic inputs.

A remaining question is whether the D1R cells we describe are responding to dopamine or some other neurotransmitter in the behaviors we examined. Distinct dopamine and norepinephrine uptake mechanisms into synaptosomes isolated from cerebellum have been identified(78). Adrenergic afferents to the cerebellar nuclei likely come from locus ceruleus (catecholamine group A6) and an adjacent nucleus subceruleus (catecholamine group A4), but not other adrenergic nuclei(79, 80). While it is clear that the LCN has receptors for dopamine, there is some disagreement about where it may be coming from. In primates, there is robust distinction between Dopamine tranporter (DAT)-positive and Dopamine-β-Hydroxylase-positive axons in different parts of the cerebellar cortex(47). Retrograde mapping with Cav2-Cre in a tomato reporter line failed to reveal any tomato labeling in the VTA of mice(81). In rats, VTA sends glutamatergic projections to LCN, but not dopaminergic projections(13). All cerebellar nuclei are reported to have DAT-like binding, which is not blocked by norepinephrine(46, 82). Other possible sources of catecholamines in the LCN include a small population of Purkinje cells in the caudal cerebellum(83); the Zona incerta (catecholamine group A13), which has projections to the interposed cerebellar nuclei(84); and the locus ceruleus, which has been previously shown to release dopamine in the hippocampus, an area with less dopamine than cerebellar nuclei(49, 85–88).

We found that inhibition of D1R neurons results in decreased spatial navigation, social recognition, WM performance, and alterations in response inhibition, which is in agreement with findings of animals and humans with lateral cerebellar lesions(23–25, 89). These findings are relevant in the context of sensory prediction error functions for which the cerebellum is specialized. The cerebellum integrates predicted sensory outcomes of motor commands with sensory feedback to achieve optimal kinematic performance^{1, 68} by generating appropriately timed anticipatory signals for accurate feed-forward predictions, particularly in adaptation of involuntary reflexes during limb movements, eyeblink conditioning, and the vestibulo-ocular reflex(3, 90–94). Notably, aspiration lesions of the cerebellum result in attenuation of the acoustic startle reflex(95) and rotarod performance(96, 97), which we did not see in D1R:Hm4Di animals, lending weight to the idea that LCN is more “cognitive” in function than other cerebellar regions. Posterolateral cerebellar cortex and LCN are postulated to influence cognitive functions by generating time-based predictions of sensory information, and predicting and synchronizing motor or cognitive activity with selective sensory input using a forward timing model²⁰. Dysfunction in these regions would result in altered processing or weighting of sensory data relative to prior learning and memory, resulting in poor cognitive performance(98). Decreased performance of D1R:Hm4Di animals in the Barnes maze (predictive self-motion, visual spatial cues), PPI (predictive acoustic prepulses), and social tasks (predictive odorants, visual and auditory cues) would make sense in this context. For example, cerebellar cortical processing of self-motion information is required to maintain stable hippocampal place fields during successful spatial navigation(99).

It may seem inconsistent that D1R:Hm4Di animals showed alterations in the DRL-10 task, but did not show disruptions on two different measures of DA until it was extended to 16 seconds. However, DA and DRL may preferentially test different aspects of attention such as cueing/alerting, orientation, action selection/retrieval, timing, execution, and error-checking. For example, DA may favor associating a sensory cue (levers extending after delay) with learning a rule and action selection (‘press the other lever’) and recalling the previous lever location. Lateral cerebellum is specifically engaged during tasks requiring precise representation of temporal information in the subsecond to second range(27, 62, 68, 89, 100–103) (as in DRL), whereas performance on DA may not require direct attention to specific time intervals(104). Consistent with this, humans with cerebellar lesions show increased variability in performance on timing tasks, with overestimations at shorter intervals(89). Some humans with highly focal damage in lateral cerebellar cortex and nuclei have timing deficits in the seconds range, without effects on attention, memory or executive function(101).

We hypothesize that deficits in D1R:Hm4Di mice on the cognitive tasks we performed are a reflection of disturbances in sensory prediction errors that perturb basic predictive attentional processes, consistent with forward models of cerebellar function(20, 105). Thus, we propose that D1R LCN neurons are a locus of integration of predictive internal models involved in cognitive processes.

Supplementary Material

supplement

Supplemental Figure S1. Mapping of Tomato+ cells (from a D1R^Cre/+;TdTomato cross) confirms subregional localization of D1R+ neurons in LCN.

A, Illustration depicting the lateral (dentate) nucleus (DN/LAT) from the rostral to caudal extent that was analyzed for D1R neuron location (viral injections were made in the middle of the structure, rostral-to-caudal).

B, Illustration of the rostral region of LCN in coronal plane (top), representative image of D1R:GFP expression in rostral LCN overlaid with divisions of dorsal, ventral, medial and lateral zones (bottom). Dorsal and lateral orientations are denoted in inset.

C, Illustration of the caudal region of LCN in coronal plane (top), representative image of D1R:GFP expression in caudal LCN overlaid with divisions of dorsal, ventral, medial and lateral zones (bottom), Scale = 120 μm. In addition to the LCN, few cells were observed in the parvicellular region of the LCN and the interposed nuclei. (DN = Dentate (or Lateral) Nucleus; IntA = Interposed Nucleus Anterior part; IntP: Interposed Nucleus, posterior part; PC: Dentate (or Lateral) Nucleus, parvicellular part; Y: Nucleus Y of the vestibular complex).

D, Quantification of amount of D1R:Tmto positive cells, relative to DAPI positive cells, in coronal sections from rostral to caudal (t=5.8, df=60).

E, Quantification of distribution of D1R:Tmto positive cells in medial vs. lateral zones in a coronal section (t=2.76, df=60).

F, Quantification of distribution of D1R:Tmto positive cells in dorsal vs. ventral zones in a coronal section (t=2.94, df=60).

G, Quantification of amount of D1R:Tmto positive cells, relative to DAPI positive cells, in coronal sections from rostral to caudal. Results were acquired from 4 mice, 4 sections per mouse counted bilaterally and are represented as the average number of cells per section per side or the percentage of cells within each region per section per side. Illustrations in A–C are from Paxinos and Franklin, 2013. * P < 0.05 ** P < 0.01, Student’s t test, two-tailed.

Supplemental Figure S2. Immunohistochemical staining for the Dopamine D1 Receptor, in human striatum as a positive control.

A, DAB+ staining of D1R in the striatum (2X magnification).

B, DAB+ staining of D1R in the striatum (10X magnification).

Supplemental Figure S3. A–H, Immunohistochemical staining for the Dopamine D1 Receptor, the Dopamine D2 Receptor in the lateral nucleus of the cerebellum and striatum of mouse.

A, Co-staining of D1R (green) and D2R (red) in LCN cell body.

B, Co-staining of D1R (green) and D2R (red) in striatum as a positive control.

C, Staining of D1R in the green color channel in LCN cell body.

D, Staining of D1R in the green color channel in striatum.

E, Staining of D2R in the red color channel in LCN cell body.

F. Staining of D2R in the red color channel in striatum.

G. Staining of DAPI in the blue color channel in striatum.

Supplemental Figure S4. A–P, Immunohistochemical staining for the Dopamine D1 Receptor, the Dopamine D2 Receptor in the lateral nucleus of the cerebellum and striatum of D1R KO and WT littermate control mice.

A, Co-staining of D1R (green), D2R (red) and DAPI (blue) in coronal section of Striatum in WT littermate (positive) control.

B, Co-staining of D1R (green) D2R (red) and DAPI (blue) in coronal section of striatum in D1R KO mouse.

C, Co-staining of D1R (green), D2R (red) and DAPI (blue) in coronal section of cerebellar nuclei (LCN, interpositus nucleus) in WT littermate (positive) control.

D, Co-staining of D1R (green), D2R (red) and DAPI (blue) in coronal section of cerebellar nuclei (LCN, interpositus nucleus) in D1R KO mouse.

E, Staining of D1R in the green color channel in striatum of WT mouse.

F, Staining of D1R in the green color channel in striatum of D1R KO mouse.

G, Staining of D1R in the green color channel in LCN of WT mouse.

H, Staining of D1R in the green color channel in LCN of D1R KO mouse.

I, Staining of D2R in the red color channel in striatum of WT mouse.

J, Staining of D2R in the red color channel in striatum of D1R KO mouse.

K, Staining of D2R in the red color channel in LCN of WT mouse.

L, Staining of D2R in the red color channel in LCN of D1R KO mouse.

M, Staining of DAPI in the blue color channel in striatum of WT mouse.

N, Staining of DAPI in the blue color channel in striatum of D1R KO mouse.

O, Staining of DAPI in the blue color channel in LCN of WT mouse.

P, Staining of DAPI in the blue color channel in LCN of D1R KO mouse.

Supplemental Figure S5. SynaptoGFP expression in projections to extracerebellar regions in Vglut2-Cre mice injected with AAV-DIO-SynaptoGFP in LCN.

A, Expression in Ventro-medial thalamus (arrowhead).

B, Expression in Red nucleus (arrowhead) and Anterior Pretectum (arrow).

Supplemental Figure S6. DREADD Receptor expression and CNO application results in silencing of neurons in vitro and in vivo.

A, Illustration depicting the lateral dentate nucleus (DN/LAT) location of injection for DREADD Receptor viral construct.

B, Illustration of DREADD Receptor construct.

C, Example trace of suppression of action potential firing in a D1R+ neuron in slice following bath application of CNO, Scale: 20 mV, 10 s.

D, Location of tetrodes during in vivo recordings of D1R LCN neurons.

E, Average z-score of neurons recorded in vivo before and after CNO application.

F, Average firing rate for each cell type, ± SEM.

G, Average change in firing rate after application of CNO, relative to baseline, ± SEM.

Supplemental Figure S7. Some measures of Barnes Maze learning and Gross motor and auditory function is not regulated by D1R LCN neurons.

A, Performance of each group during Barnes maze training trials, as measured by distance traveled (cm). N = 17, D1R:Hm4Di mice and N = 23 littermate D1R:GFP control mice. Error bars are SEM (Two-way rmANOVA, reveals no significant differences between groups or interaction).

B, Performance of each group during Barnes maze training trials, as measured by center crossings. N = 17, D1R:Hm4Di mice and N = 23 littermate D1R:GFP control mice. Error bars are SEM (Two-way rmANOVA, reveals no significant differences between groups or interaction).

C, Performance of each group during training trials, as measured by latency to reach goal hole (s). N = 17, D1R:Hm4Di mice and N = 23 littermate D1R:GFP control mice. Error bars are SEM (Two-way rmANOVA reveals no significant differences between groups or interaction).

D, Schema for when CNO was injected IP prior to probe tests only above performance of each group during training trials, as measured by average duration in the goal quadrant (left) and as measured by nose pokes in target quadrant holes and goal hole (right). N = 16 D1R:Hm4Di mice, and N = 11 littermate D1R:GFP control mice. For goal quadrant duration, Error bars are SEM. (Two-way rmANOVA, 2 factors was significant: training (F_3,75 = 9.98, P < 0.0001), and interaction (F_1,25 = 2.89, P < 0.05) but not presence of Hm4Di. On post hoc analysis, difference between groups was significant on day 2 of training only (P < 0.05). For nose pokes in goal quadrant, Error bars are SEM; one factor was significant: Training (F_3,75 = 12.78, *P < 0.0001) ; The factors of interaction or presence of Hm4Di alone were not significant (Two-way rmANOVA).

E, Memory recall during the probe trial on day 5, i.e., when CNO was given only on the probe trial, and not during training of the Barnes maze (as in A–C), as measured by nose pokes in target quadrant. (N = 16, D1R:HM4 mice and N = 11 D1R:GFP mice, no significant differences between groups were found, Student’s t test, two-tailed).

F, Memory recall during the probe trial on day 5 of a different version of the Barnes maze, i.e., when CNO was given only on the probe trial, and not during training (as in A–C) as measured by nose pokes in goal. (N = 16, D1R:Hm4Di mice and N = 11 D1R:GFP mice, *P < 0.05, n.s. = not significant, Student’s t test, two-tailed).

G, Motor coordination as assessed by latency to fall during accelerating rotarod across days is not different between groups, as assessed by Two-way rmANOVA (N = 22, D1R:GFP mice, and N= 20, D1R:HM4 mice).

H, Gait analysis (N = 16, D1R:HM4 mice and N = 11 D1R:GFP mice).

I, Acoustic Startle Curve assessed by increasing startle amplitude (N = 10, D1R:GFP mice, and N= 10, D1R:HM4 mice).

J, Social approach as measured by time spent in sniffing zone with a novel mouse or novel object. Both groups preferred social interaction (F_1,72 = 48.52, P < 0.0001). (N = 22, D1R:GFP mice, and N= 17, D1R:Hm4Di mice, Two-way ANOVA). No significant differences were found for presence of Hm4Di or interaction (presence of Hm4Di X zone).

K, Social preference as measured by time spent in sniffing zone with a novel mouse or familiar mouse. Factor for interaction (presence of Hm4Di X zone, F_1,72 = 4.59, *P < 0.05), but not presence of Hm4Di or zone were significant (N = 22, D1R:GFP mice, and N= 17, D1R:Hm4Di mice, no significant differences between groups were found on post-hoc analyses).

L–M, Exploration of the three-chambered arena (as measured by chamber crossings) is not affected by inactivation of D1R neurons during the social approach task (novel object vs. novel animal), L, or during the social interaction task (novel animal vs. familiar animal), M (N = 17, D1R:Hm4Di mice and N = 22 D1R:GFP control mice).

Supplemental Figure S8. Distribution of latencies of lever presses after reward (A) and total presses (B) in the DRL-10 week 6, week 7 (C–D), and week 8 (E–F) for D1R:GFP control female mice (black circles) and D1R:Hm4Di female mice (red diamonds). Group means ± SEM are presented for latencies binned in 2s intervals. Dotted line represents minimum response latency that is rewarded for each DRL paradigm. No differences were seen in the mean peak of response distribution of each group for each respective week of training and DRL paradigm. Statistics for presses after reward (2-way RM ANOVA): Presence of Hm4Di x training interaction: DRL-10, Week 6: F_29,348 = 1.85, P < 0.01; Presence of Hm4Di x training interaction: DRL-10, week 7: F_29,348 = 1.73, P < 0.05; Presence of Hm4Di: F_1,12 = 5.83, P < 0.05; Presence of Hm4Di x time interaction:DRL-10, week 8: F_29,348 = 1.52, P < 0.05 (N = 6, D1R:Hm4Di mice and N = 5, D1R:GFP mice). Statistics for total presses: (2-way RM ANOVA, Presence of Hm4Di x time interaction: DRL-10, Week 6: F_29,261 = 2.82, P < 0.0001). No significant differences were seen related to presence of Hm4Di or presence of Hm4Di x time interaction in week 7. Presence of Hm4Di x time interaction:DRL-10 week 8: F_29,348 = 1.80, P < 0.01; Presence of Hm4Di: F_1,12 = 5.80, P < 0.05 (N = 9, D1R:Hm4Di mice and N = 5, D1R:GFP mice).

Supplemental Figure S9. A, Cumulative lever presses on Day 1 (bold dotted lines; D1R:Hm4Di = dark gray, D1R:GFP = light gray) and Day 4 (solid lines; D1R:Hm4Di = dark gray, D1R:GFP = light gray).

B, Progressive ratio breakpoint as measured by the total number of lever presses to acquire a single reward. (A: GFP, N = 5, Hm4Di, N = 6; B and C: GFP, N = 12, HM4, N = 10).

C, Performance on the 10 s DRL task as measured by mean number of pellets rewarded over time in males. No significant differences by presence of Hm4Di or Hm4Di x time interaction on 2-way RM ANOVA.

D, Performance on the 10 s DRL task as measured by mean number of pellets rewarded over time in females. Performance was decreased for Hm4Di group, 2-way RM ANOVA, Presence of Hm4Di: F_1,12 = 4.93, P < 0.05; Hm4Di x time interaction: F_39,468 = 2.36, P < 0.0001; N = 9, D1R:Hm4Di mice and N = 5, D1R:GFP mice.