Differential innervation of direct- and indirect-pathway striatal projection neurons

Abstract

The striatum integrates information from multiple brain regions to shape motor learning. The two major projection cell types in striatum target different downstream basal ganglia targets and have opposing effects on motivated behavior, yet differential innervation of these neuronal subtypes is not well understood. To examine whether input specificity provides a substrate for information segregation in these circuits, we used a monosynaptic rabies virus system to generate brain-wide maps of neurons that form synapses with direct- or indirect-pathway striatal projection neurons. We discovered that sensory cortical and limbic structures preferentially innervated the direct pathway, whereas motor cortex preferentially targeted the indirect pathway. Thalamostriatal input, dopaminergic input, as well as input from specific cortical layers, was similar onto both pathways. We also confirm synaptic innervation of striatal projection neurons by the raphe and pedunculopontine nuclei. Together, these findings provide a framework for guiding future studies of basal ganglia circuit function.

Figure 1. Experimental design

(A) Adult D1R-Cre or D2R-Cre mice are injected in the dorsal striatum with 180 nl of AAV9 expressing TVA and rabies glycoprotein in a Cre-dependent manner. (B) 21 days later, the same mice are injected with 180 nl of monosynaptic rabies virus that can only infect cells expressing TVA, and can only spread retrogradely from cells expressing rabies glycoprotein. (C) Direct inputs onto either direct pathway (D1R-expressing) or indirect pathway (D2R-expressing) MSNs are labeled one week after rabies injection.

Figure 2. Monosynaptic rabies virus injections into the striatum target the direct or indirect pathways

Abbreviations: DS: dorsal striatum, GP: globus pallidus, EP: entopeduncular nucleus, SN: substantia nigra. (A) Injection site and resulting fiber projections from starter cells in a D1R-Cre mouse. As expected, direct-pathway MSNs are labeled and heavily innervate EP and SNr, with some termini in GP. All scale bars indicate 250 μm. (B) The border between dorsal striatum and globus pallidus shows that direct infection is tightly constrained to the striatum. (C) Injection site and resulting fiber projections from starter cells in a D2R-Cre mouse. As expected, indirect-pathway MSNs heavily innervate GP, with little to no label in EP or SNr. (D) Higher magnification at the border of striatum and globus pallidus shows strong direct infection in the striatum, with axon termini in GPe. (E–F) Mean center and mean injection span are diagrammed with colored crosses in a sagittal section (E) and a coronal section (F). Red crosses indicate the mean extent of injections into D1R-Cre mice, and blue crosses indicate injections into D2R-Cre mice. (G) Mean position, extent, and variability of the primary injection site for D1R-Cre and D2R-Cre mice. AP: anterior-posterior (positive values indicate anterior to bregma), LM: lateral-medial (values indicate lateral distance from bregma), DV: dorsal-ventral (negative values indicate ventral from bregma). See also Figure S1.

Figure 3. Summary of monosynaptic inputs onto direct- and indirect-pathway MSNs

Direct pathway inputs are labeled in red, indirect pathway inputs labeled in blue. Only inputs that were detected in at least three animals are displayed. The majority of direct synaptic inputs arise from cortex and thalamus, with a smaller proportion of inputs from midbrain and hindbrain structures. Error bars indicate 1 SEM.

Figure 4. Biased cortical synaptic input to the direct and indirect pathways

(A–B) Cortical inputs from somatosensory cortex onto both direct and indirect pathway MSNs arise almost exclusively from superficial layer 5, though overall input density appeared to be higher onto direct pathway MSNs. Scale bar = 100μm and applies to panels A–F. (C–D) Motor cortex inputs onto both pathways arise primarily from superficial layer 5, but some superficial layer cells also contribute input. Motor cortex preferentially innervates the indirect pathway. (E–F) Inputs from the insular and orbital cortices (labeled PFC) arise from both superficial and deep layers, and appear to innervate both the direct and indirect pathway similarly. (G) Cortical inputs were segregated into four major input streams. Direct pathway inputs are colored red, whereas indirect pathway inputs are shaded in blue. Individual p values indicate two-tailed t-test comparison of direct vs. indirect pathway input for each stream. Error bars indicate 1 SEM. (H) Center of gravity analysis demonstrates that cortical input to the direct pathway is significantly caudal to that for the indirect pathway. Individual anterior-posterior centers of gravity are indicated by colored circles at the top of the graph. The two-dimensional chart shows center of gravity for cortical inputs to the direct and indirect pathways in both the anterior-posterior and dorsal-ventral dimensions. Bar width indicates ±1 SEM. The plane of slice is 2.04 mm lateral to the midline. The border between primary somatosensory and primary motor cortex at this plane of slice is indicated by the dashed line for reference. One mouse with significantly different center of gravity is excluded in the D1R-Cre cohort, and is indicated by the light circle. (I) Summary diagram showing the relative contribution of different cortical layers to striatal input for four different cortical areas in the plane of slice containing the cortical center of gravity (ordered from rostral to caudal). Within a given cortical region, each cortical layer provided similar proportions of synaptic input to the direct and indirect pathways, even though total proportion of inputs could be dramatically different (see figure 3). More rostral brain structures provided more superficial input than more caudal cortical areas. ND: not determined, representing the portion of labeled neurons whose layer identity could not be accurately determined, largely due to plane of slice providing ambiguous layer information. (J) Summary image diagrams both the relative strength of layer input from major cortical input structures (shaded in tan) as well as the relative amount of input streaming from each cortical area onto direct (red) and indirect (blue) motor pathway MSNs, as well as the downstream targets of these MSNs. See also Figures S2 and S3.

Figure 5. Thalamus similarly innervates direct- and indirect-pathway MSNs

(A) Sagittal slice diagram depicts the site of densest thalamic input. The dashed box corresponds to the region of interest displayed in (D) and (E). (B,D,F) Thalamic inputs onto direct pathway MSNs labeled at 1/6 sampling density. The majority of thalamic inputs arise from thalamic nuclei MD, MDL and PF, with other inputs from AD, AM, and the central nuclei. (C,E,G) Thalamic inputs onto indirect pathway MSNs labeled at 1/6 sampling density. The majority of thalamic inputs arise from nuclei MD, MDL, and PF, with nuclei VM, AM, and VL providing smaller proportions of input. Abbreviations: AD: anterior dorsal nucleus, AM: anterior medial nucleus, AV: anterior ventral nucleus, CL: central lateral nucleus, CM: central medial nucleus, Gus: gustatory nucleus, IAD: interior anterior dorsal nucleus, LDDM: lateral dorsal nucleus, dorsomedial portion, LPMR: lateral posterior nucleus, mediorostral portion, LHb: lateral habenula, MD: medial dorsal nucleus, MDC: medial dorsal nucleus, central portion, MDL: medial dorsal nucleus, lateral portion, MHb: medial habenula, PC: paracentral nucleus, PT: paratenial nucleus, Re: reuniens nucleus, Rt: reticular thalamic nucleus, Sub: submedius nucleus, VA: ventral anterior nucleus, VL: ventral lateral nucleus, VM: ventral medial nucleus.

Figure 6. Central and basolateral amygdala preferentially innervate the direct pathway

(A) Whole slice images were manually registered via scaled rotation at 1/2 sampling density. Region of interest is indicated with a square and is plotted in greater detail in (B–E). (B–E) Coronal sections through the amygdala are depicted at four different anterior-posterior positions. Direct pathway inputs depicted in red, indirect pathway inputs in blue. Abbreviations: BLA: basolateral nucleus of the amygdala, anterior portion., BMA: basomedial nucleus of the amygdala, anterior portion, BMP: basomedial nucleus of the amygdala, posterior portion, CeC: central nucleus of the amygdala, capsular portion, CeL: central nucleus of the amygdala, lateral portion, CeM: central nucleus of the amygdala, medial portion, DEn: dorsal endopiriform nucleus, La: lateral nucleus of the amygdala, Pir: piriform cortex, VEn: ventral endopiriform nucleus. Although stereotaxic coordinates of all cells are accurate, heterogeneity in distribution of amygdala nuclei in individual mice may not overlap completely with atlas borders.

Figure 7. Monosynaptic rabies virus only labels a small proportion of total dopaminergic input to the dorsal striatum

(A) Substantia nigra in a D1R-Cre mouse. A large number of axon fibers from direct pathway starter cells are visible in SNr, but few retrogradely labeled cells are detectable in SNc. (B) Substantia nigra in a D2R-Cre mouse. Since indirect pathway MSNs do not project to the substantia nigra, few if any fibers are detectable in SNr. Again, few retrogradely labeled neurons are detectable in SNc. (C) Diagram of rabies virus spread from targeted striatal MSNs to their presynaptic partners. Genetically-targeted, monosynaptic RV efficiently spreads at many types of synapses (see Figure 3), but less than 1% of labeled inputs arise from SNc. (D–E) Substantia nigra in C57 control mice injected in the striatum with a retrograde tracer rabies virus expressing mCherry. When rabies virus is taken up nonspecifically at axon terminals in striatum, many dopaminergic cells are labeled in SNc. (F) Diagram of rabies virus as a traditional retrograde tracer, which is taken up at axon terminals near the viral injection site. When injected into dorsal striatum, nearly 8% of retrogradely labeled cells are found in SNc, indicating that rabies virus is efficiently taken up at dopaminergic axon terminals. When compared to (C), these observations suggest that either the extracellular space or the dendritic composition in starter cells prevents monosynaptic rabies virus from spreading to the majority of apposed dopaminergic terminals. Scale bar = 100 μm for all panels.