PCore: Pallidal Core BG Algorithm

[rcoreilly]

PCore: Pallidal Core BG Algorithm

This discussion provides updates to the pcore algorithm powering BG in Axon: https://github.com/emer/axon/blob/main/PCORE_BG.md in light of recent updates to understanding the BG in #294. Also, the original pcore model was based on very new research (at the time), which has now had time to be further explored and tested. There are two highest-level points:

1. The specialized pcore circuitry is likely only applicable to DLS (dorsolateral striatum, i.e., primary motor striatum), yet we're using it for VS (ventral striatum) as well. It would be better to establish a simpler model applicable just to VS, appropriate for its specific function: firing Go when a goal state in PFC is likely to lead to a good outcome (positive net utility). The pcore example test is currently using this as its evaluation criterion, but it doesn't really make sense given the next point.

2. Per Motor learning mechanisms #294, the functional role of DLS is not really action selection as classically defined, and as employed during the original development of the pcore model in Leabra (and by SuryanarayanaHellgrenKotaleskiGrillnerEtAl19 which strongly informed the original pcore model). Instead, it is much more about timing and sequencing, and likely about generating a large repertoire of context-dependent motor sequence elements that can be reinforced or punished according to outcomes. See Motor learning mechanisms #294 (reply in thread) for hypothesized learning algorithm. This mechanism provides a good functional role for various known elements of BG, but we need to revisit all of the pcore dynamics in light of this new framing.

[rcoreilly]

GuilhemsangMallet24

Guilhemsang, L., & Mallet, N. P. (2024). Arkypallidal neurons in basal ganglia circuits: Unveiling novel pallidostriatal loops? Current Opinion in Neurobiology, 84, 102814. https://doi.org/10.1016/j.conb.2023.102814

This is a key update review on the novel PCore circuitry. Actually not that great of a review, but still useful. Does not adequately address the PV vs. Npas1 distinction..

Overall todo: rename "TA" -> Arky -- TA = "type A" which is lame.

The main GPe neuron population, making up nearly 70-75% of all GPe cells, is referred to as Prototypic neurons. These neurons consistently send long-range axonal projections targeting the STN and all major downstream BG nuclei (Figure 1b). Notably, prototypic neurons often establish large and potent synapses directly onto the cell bodies of their target neurons. This strategic synaptic organization allows Prototypic neurons to not only regulate the firing ac- tivity levels in many target nuclei but also to control synchronization levels and impose phase-firing relationship during oscillatory activity [29].

Indeed, a subset of Prototypic neurons also projects axons back to the striatum (Figure 1b). Although this projection is relatively sparse, they selectively contact GABAergic interneurons and could thus indirectly affect the firing rate or synchrony level of SPNs networks [30,31]. The specific impact of the Prototypic pallidostriatal GABAergic inputs onto motor behaviours has never been investigated, but these GABA inputs could also contribute to the selection / execution of striatal motor programs [32]

• We have these prjns.

In contrast, the remaining 25-30% of GPe neurons, so- called Arkypallidal, defy the conventional nature of GPe neurons by exclusively projecting to the striatum (Figure 1b). These axonal projections are denseda single neuron can make up more than 10 000 synapses [23]dand they predominantly innervate the dorsolat- eral ‘motor’ region of the striatum as opposed to the medial ‘associative’ region [36].

• todo: remove TA from VS. Note: if we maintain Out / In distinction, then roughly 1/3 of each type -- consistent with model.

hese Arkypallidal projections appear to be topographically organized concerning the major striatal domains (i.e. rostral vs. caudal and dorsal vs. ventral). Whether they are also organized with respect to finer striatal functional do- mains, as defined by somatotopic map [37], remains unknown, although it seems unlikely given the broad striatal extent of these projections.

• broad pooled surround inhibition!

The striatal synaptic targets of Arkypallidal neurons encompass SPNs and all the major types of striatal interneurons (with a preference for cholinergic interneurons). Moreover, the synaptic terminals of Arkypallidal neurons exhibit different localization pat- terns when targeting SPNs compared to interneurons [23]. Specifically, Arkypallidal terminals established perisomatic contacts with striatal interneurons, whereas they target the distal dendrites of SPNs. Additionally, synapses onto SPNs often target dendritic shafts or the necks of dendritic spines, making these Arkypallidal synapses an effective mechanism for filtering excitatory inputs and controlling which SPNs become activated.

• todo: think about implications on CINs -- not currently implemented.

In normal animals, Arkypallidal projections equally affect D1-SPNs and D2-SPNs, although the evoked synaptic current appears slightly stronger in D2-SPNs [38�]. As a result, it seems that the inhibitory projections from Arkypallidal neurons lack specificity toward striatal cell type and are instead distributed across extensive striatal territories. This property aligns well with the proposed ‘stop’ signal function that these neurons could poten- tially perform, particularly when the cancellation of an imminent action is required [39�].

Previous in vivo electrophysiological recordings con- ducted in Parkinsonian animals with 6-OHDA lesions, have revealed that Prototypic and Arkypallidal neurons exhibit opposing phase firing during synchronized oscillatory activity events [23]. Furthermore, opto- genetic excitation of STN neurons had contrasting ef- fects on both GPe neurons: it induced rapid excitation in Prototypic neurons but a corresponding inhibitory response in Arkypallidal cells [29].

Collectively, these findings raise the intriguing possibility that the func- tional dichotomy between Prototypic and Arkypallidal neurons extends to disparities in their input organiza- tion. This hypothesis was recently examined for D2- SPNs and STN inputs using optogenetic stimulation in D2-Cre and Vglut2-Cre animals, respectively [45��,46].

• NoGo -> Proto (GPeIn) but not NoGo -> TA (in model)
• Reduces inhibition of STN; STN -> TA is strong (note: STNp only -- what about STNs?)
• STN activity = net TA inhibition from Proto -> -TA, which must be stronger than STN -> TA excitation..

For instance, D2-SPNs stimulation led to the inhibition of Prototypic neurons but excitation in Arkypallidal neurons (Figure 2a). Conversely, STN stimulation triggered excitation in Prototypic neurons but inhibition in Arkypallidal activity (Figure 2b). This opposition of response was attributed to the potent inhibitory impact of Prototypic local axon collaterals onto Arkypallidal neurons.

In addition to these two primary GPe inputs, the axons of D1-SPNs neurons form ‘bridging’ collaterals into the GPe [47]. Interest- ingly, optogenetic excitation of D1-SPNs in D1-Cre mice triggered a strong biased response towards Arky- pallidal neurons [45��], but see Ref. [34]. These syn- aptic organizations are further supported by in vitro studies showing that D1-SPNs preferentially influence Npas1þ GPe neurons (half of which are Arkypallidal) [27], while STN inputs evoke stronger synaptic currents into PV þ neurons (many of which are Proto- typic) than in Npas1þ cells [19]. Similar conclusions regarding GPe inputs organization were reached when dissecting the responses elicited by in vivo sensory stimulation, rather than bulk optogenetic excitation of specific inputs [48].

Investigation into how the activity change in Prototypic and Arkypallidal neurons, induced by optogenetic stimulation of D2-SPNs or STN, influenced locomotion behaviour revealed interesting findings. While both inputs are traditionally considered as motor-suppressing pathways, locomotion inhibition was exclusively observed following optogenetic stimulation of D2-SPNs neurons, not STN neurons [46]. This disparity in behavioural outcomes could be attributed to the distinct influences these stimulations exert on BG output ac- tivity. Indeed, the prevailing idea in the field is that movement inhibition induced by both pathways de- pends solely on their net effect on BG output. However, this conventional view is challenged by a recent study showing that locomotion inhibition following D2-SPNs stimulation is achieved through the local influence of D2-SPNs axon collaterals [49��]. Similarly, Arkypallidal opto-excitation reproduced the locomotor inhibitory effect observed with D2-SPNs stimulation, but without disturbing Prototypic neurons, thus leaving the remaining indirect pathway unperturbed [46].

This last work demonstrates that Arkypallidal neurons effectively form a pallidostriatal feedback loop capable of modu- lating action inhibition at the BG input stage: the striatum. Importantly, this Arkypallidal feedback might be directly controlled by D1-SPNs inputs [45��] or indirectly recruited by D2-SPNs through the inhibition of Prototypic neurons [46].

[rcoreilly]

PamukcuCuiXeniasEtAl20

Pamukcu, A., Cui, Q., Xenias, H. S., Berceau, B. L., Augustine, E. C., Fan, I., Chalasani, S., Hantman, A. W., Lerner, T. N., Boca, S. M., & Chan, C. S. (2020). Parvalbumin+ and Npas1+ Pallidal Neurons Have Distinct Circuit Topology and Function. Journal of Neuroscience, 40(41), 7855–7876. https://doi.org/10.1523/JNEUROSCI.0361-20.2020

Context: Npas1+ are 50% Arky and 50% something else -- what is that something else??

We and others established that the GPe contains two principal neuron classes distinguished by their expression of the calcium-binding protein parvalbumin (PV) or the transcription factor Npas1. They account for ;50% and 30% of the GPe neuron population, respectively (Flandin et al., 2010; Nóbrega Pereira et al., 2010; Abdi et al., 2015; Dodson et al., 2015; ...) PV+ neurons and Npas1 neurons have different firing characteristics and projection targets. Specifically, PV neurons exhibit higher firing rates and referentially target the subthalamic nucleus (STN) and substantia nigra pars reticulata (SNr), whereas Npas1+ neurons have lower firing rates and preferentially target the dorsal striatum (dStr) (refs).

In accordance with their distinct electrophysiological and circuit characteristics, we found that PV neurons and Npas1 neurons have distinct functional roles—PV neurons promote locomotion and Npas1 neurons suppress it. By examining the excitatory inputs to the GPe, we showed that STN input to the GPe is unique in its cell type specificity and preferentially targets PV1 neurons over Npas11 neurons.

• todo: STN -> GPeOut / In, NOT Arky!

By harnessing a chronic unilateral 6-hydroxydopamine (6-OHDA) lesion model of PD, we showed that the STN–PV1 input is selectively weakened in the chronic 6-OHDA-lesioned model of PD. In vivo optogenetic stimulation of PV1 neurons promoted locomotion and alleviated hypokinetic symptoms in 6-OHDA-lesioned mice.

Notably, EPSC amplitudes, as a measure of the STN input strength, were approximately five times larger in PV1 neurons compared with those in Npas1 neurons.

• Npas+ does get some STN input

In contrast to the STN input, PF Thal and PPN inputs did not produce detectable EPSCs in all GPe neurons.

While Npas1 neurons have extensive axonal arborization in the dStr, their downstream impact on the excitability of striatal projection neurons is relatively weak (Glajch et al., 2016). It is possible that additional signaling partners are involved.

In this study, we found that the STN provides stronger input to PV+ neurons than to Npas1 neurons. This finding is supported
by a recent in vivo electrophysiological study showing that STN strongly targets “prototypic” neurons that are predominantly
PV+ neurons (Ketzef and Silberberg, 2020).

• PV+ -> prototypical!

In addition, our finding contradicts computational studies that predicted a stronger connection from the STN to a subset of Npas1 neurons (Bogacz et al., 2016; Suryanarayana et al., 2019)

• key point! we are also subject to this issue.

Recent studies indicate the presence of heterogeneity in STN neurons (Sato et al., 2000b; Isoda and Hikosaka, 2008; Koshimizu et al., 2013; Xiao et al., 2015; Papathanou et al., 2019).

• todo: track down recents

[rcoreilly]

CourtneyPamukcuChan23

Courtney, C. D., Pamukcu, A., & Chan, C. S. (2023). Cell and circuit complexity of the external globus pallidus. Nature Neuroscience, 26(7), Article 7. https://doi.org/10.1038/s41593-023-01368-7

Finally found the review paper from this group!

Summary of cell types and primary connectivity, functions, with my own simplified terminology in absence of anything they provided:

• GPePr prototypic PV+ neurons, PV+KCNG4+, maybe 45% (50% total PV+): iSPN (D2, No) --o Pr --o SNr; also inhibit other GPe cells (self and ?, but not much GPeArky), and bidirectionally connect with STN.
• GPePf PV+ PF (parafasicular thalamus) projecting neurons, PV+LHX6-, maybe 5%: dSPN (D1, Go) --o Pr
• GPeAk arkypallidal NPAS1+FOXP2+, 18% (60% of 30% NPAS1+): dSPN (D1, Go) --o Arky --o dStr (not much iSPN, STN, or local input) -- basically a tonic brake on SPN that can be inhibited in a "Go" scenario, allowing broader activity (disinhibition).
• GPePt "patch-like" (striosomal) projection pattern, NPAS1+NKX2.1+, 12% (40% of 30% NPAS1+): dSPN (D1, Go) --o GPePatch --o SNc, TRN, Cortex, LHb; also recv from cortex. Maybe get most lateral input from GPePr?
• [remaining 20% are heterogeneous, 5% ChAT+ cholinergic, 15% PV-NPAS1-]
• Everyone gets PF and L5 cortical input (excitatory).

PV-expressing (PV+) neurons and NPAS1-expressing (NPAS1+) neurons constitute 50% and 30% of the adult GPe, respectively, and are considered as the two principal neuron classes (Fig. 2)

However, by systematically examining a number of transgenic lines, a recent examination definitively con- firmed that PV+ neurons account for approximately 50% of neurons in the GPe [28].

Aside from the two principal neuron classes, neurons lacking expression of either PV or NPAS1 amount to 20% of the GPe. This population includes choline acetyltransferase-expressing (ChAT+) neurons, which amount to 5% of the total population27,28,30 and project to the cortex30. The remaining GPe neurons (~15%) have not been rigorously characterized. This neuron population has a unique transcription fac- tor profile: they express LHX6 but not SOX6 (LHX6+SOX6−)28. Probably owing to the relatively low abundance of LHX6+SOX6− neurons, no data are available from the previous single-cell transcriptomic data- set4,36, and identifying reliable markers for these neurons has been challenging. Importantly, LHX6 expression should not be considered as a classifier as it is seen across multiple GPe neuron populations, including in PV+, NPAS1+ and PV−NPAS1− neurons. Furthermore, the reported expression of LHX6 within these subpopulations varies widely across laboratories; some studies show little overlap with other neuron groups, while others show up to ~95% overlap with other established cell type-specific markers4,11,14,16,25,26,28,29,34 (see ‘Neuron subclasses’ and Table 1).

By harnessing molecular tools to target specific GPe neuron populations, recent studies have identified distinct subclasses of PV+ and NPAS1+ neu- rons with unique properties (Table 1). Within the NPAS1+ class, 60% are FOXP2+ (NPAS1+FOXP2+, also known as ‘arkypallidal’ neurons; Box 1); these neurons represent the first unique GPe neuron sub- class described38. Compelling data show the distinct features of NPAS1+FOXP2+ neurons, such as their developmental origin and their electrophysiological, anatomical and molecular profiles26. Importantly, as FOXP2 expression is maintained in adults, the Foxp2Cre and Foxp2Flp mouse lines have been successfully used across laboratories to capture this population of neurons14,15,28,39.

The remaining neurons within the NPAS1+ class are both NKX2.1+ and LHX6+. Unlike NPAS1+FOXP2+ neurons, which project exclusively to the dStr, NPAS1+NKX2.1+ neurons project to the midbrain, the cor- tex and the reticular thalamus14,28,30 (see ‘Synaptic outputs and func- tions’ and Fig. 4). Accordingly, both NPAS1+FOXP2+ and NPAS1+NKX2.1+ neuron subtypes should be regarded as bona fide subclasses, as recent single-cell data4,36 further confirmed the unique transcriptomic proper- ties of these two NPAS1+ neuron subclasses. Notably, these studies led to the discovery that GPe neurons labeled using the Npr3Cre driver are molecularly and anatomically consistent with NPAS1+NKX2.1+ neurons14.

In addition to the two distinct NPAS1+ neuron subclasses, multiple lines of evidence point to at least two subtypes of PV+ neuron (Table 1). One of these populations is identified by their axonal projec- tion to the STN and the SNr. The marker KCNG4 demarcates a subset of these STN-projecting PV+ neurons. However, a marker encompass- ing all STN-projecting PV+ neurons has yet to be found14. In addition, parafascicular thalamus (Pf)-projecting PV+ neurons constitute a unique subtype of PV+ neurons within the GPe: they have distinct syn- aptic partners, intrinsic properties and behavioral roles16. These PV+ neuron subclasses are further distinguished by LHX6 expression, which converging evidence suggests is present in the STN-projecting popula- tion but is weak or absent in the Pf-projecting population16,29. While it is clear that STN-projecting (PV+KCNG4+) neurons and Pf-projecting (PV+LHX6−) PV+ neurons are unique neuron subclasses, it is unclear how they assimilate into existing fate-map and transcriptomic data (Developmental origins).

The dStr is, by far, the largest input to the GPe: the number of input neu- rons from the dStr constitutes ~80% of the total projection to the GPe and is at least an order of magnitude larger than that from other brain regions47,49. This observation is consistent with the earlier finding that GABAergic synapses amount to over 80% of all synapses in the GPe20.

Contrary to the simplified basal ganglia model, it has been known since the 1980s that axons from both direct-pathway spiny projec- tion neurons (dSPNs) and indirect-pathway spiny projection neurons (iSPNs) converge at the GPe. While iSPN axons terminate exclusively in the GPe, modern techniques have confirmed that dSPN axons col- lateralize in both the GPe and the SNr47,49–54; these studies solidify the fact that the GPe receives notable direct-pathway inputs and demon- strate the physiological and targeting properties of dSPN collaterals. Nevertheless, the indirect pathway remains the predominant input to the GPe. While previous studies showed that dSPNs provide roughly half as many boutons as do iSPNs in the GPe, more recent estimations of the number of boutons formed by iSPNs are three to eight times higher than that formed by dSPNs47,49. Both viral-tracing and patch-clamp studies demonstrate that iSPNs strongly target canonical STN-projecting PV+ (PV+KCNG4+) neurons, whereas dSPNs largely target NPAS1+ neurons, Pf-projecting PV+ (PV+LHX6−) neurons and ChAT+ neurons15,16,30,39,47,55,56.

GPe neurons provide lateral GABAergic inhibition to neighboring neurons through their local axon collaterals14,15,31,39,56–61. These collat- erals give rise to a large number of local synapses, in contrast to the relatively few synapses typically formed by inputs from individual SPNs59 (Fig. 3). They have not been studied in great detail until recently owing to the inherent difficulty in identifying and selectively activat- ing individual classes of GPe neurons and their collateral axons. These technical challenges have been overcome by combining transgenic and optogenetic approaches. Although subtle nuances exist across studies, three key findings have recently emerged. First, PV+ neurons provide the largest local input. Second, NPAS1+FOXP2+ neurons do not produce appreciable levels of local connections. Third, the local collaterals can strongly inhibit neighboring neurons and thus are integral to both local circuit dynamics and downstream network effects14,15,31,39,56,57. While further studies are needed to investigate the functional relevance of the organization of local collaterals, the existence of local inhibitory networks creates an anatomical substrate for feedback and functional opponency across neuron subtypes.

Three independent studies have convincingly demonstrated that STN inputs target PV+ neurons more strongly than NPAS1+ neurons13,15,39. While the neuronal makeup of the STN was generally thought to be homogeneous, recent studies show that STN neurons are more heterogeneous than previously expected6,9,58,62–64. Therefore, it will be important to deter- mine whether PV+-targeting neurons within the STN are distinct from those targeting NPAS1+ neurons.

Importantly, viral-tracing data suggest that the spatially dis- tributed cortical inputs together form the largest source of excitatory inputs to the GPe16,28,65–68, in contrast to the traditional model, which assumes dominant glutamatergic input from the STN. These cortical inputs arise primarily from pyramidal tract neurons9,28,67. However, the precise downstream targets require further characterization, as data from recent studies are somewhat conflicting. While a recent study suggests that secondary motor cortex input selectively targets NPAS1+FOXP2+ neurons65, data from others show that cortical innerva- tions do not target specific neuron subtypes9,16,28.

In addition to the STN and the cortex, the Pf and the pedunculo- pontine nucleus send glutamatergic projections to the GPe13,16,47,69–71, two sources that are omitted from the classic model. These inputs do not show a biased connectivity pattern across PV+ neuron and NPAS1+ neuron classes13. In sum, the most recent data indicate that the GPe integrates a wide array of synaptic inputs, including from sources beyond traditional basal ganglia nuclei.

The GPe receives substantial dopaminergic input: recent viral-tracing data show that this input arises primarily from the substantia nigra pars compacta, although we have yet to determine the precise neuronal subpopulation responsible3,16,46,47. While the signaling partners of this input remain to be explored, strong evidence indicates that PV+ neurons and their projections are a major target, as they robustly express the receptor for CRF46,48,80,81. Notably, the central amygdala forms one of the largest extrinsic inputs targeting the GPe47, and this projection has been recently implicated in fear learning80.

PV+ neurons form the principal inhibitory innervation to the STN25,27,29,32,33,35. Consistent with this, optogenetic and chemogenetic stimulation of pan-PV+ neurons or selectively of STN-projecting GPe 11,13,14,16 neuron subsets promotes movement.
In addition, this effect can be recapitulated by direct inhibition of STN neurons13. These find- ings agree with the well-established relationship between STN activity and movement suppression85–93. Along with the projection to the STN, canonical PV+ (PV+KCNG4+) neurons also collateralize within the SNr and the GPi14,16,94. In addition to locomotion, modulation of PV+ neuron collaterals in the SNr induces avoidance behavior48. These findings are consistent with the idea that both motor and non-motor signals are processed in parallel within the basal ganglia.

However, it is important to note that optogenetic inhibi- tion of PV+ neurons does not yield any motor response11, suggesting the possibility of compensatory network mechanisms, such as involvement of the local collaterals. Alternatively, the basal ganglia may control motor output by computation through a cascade of logic gates that operate on multiple rules. This idea is consistent with the coactivation of dSPNs and iSPNs during volitional movement95–97.

Additionally, a small subset of PV+ neurons target the Pf (Neuron subclasses). These non-canonical PV+ (PV+LHX6−) neurons are not involved in motor control but are associated with reversal learning16. Given that STN-projecting (PV+KCNG4+) and Pf-projecting (PV+LHX6−) PV+ neurons receive distinct sets of synaptic inputs, their respective roles in motor and non-motor functions reinforce the concept that GPe neuron subtypes enact their functions through their unique circuit architecture.

On the basis of relative spike timing, earlier literature inferred that NPAS1+FOXP2+ neurons mediate motor inhibition by broadcasting an inhibitory signal across the dStr98, their exclusive projection target. However, this causality had not been demonstrated. Importantly, recent optogenetic and chemogenetic approaches demonstrate that stimula- tion of either pan-NPAS1+ neurons or the NPAS1+FOXP2+ subset leads to motor inhibition during both spontaneous ambulation and a trained task12–15, further establishing the fact that GPe neurons directly modu- late motor behavior. Additionally, optogenetic inhibition of NPAS1+ neurons leads to motor promotion13, indicating that the observed phe- notype with excitatory actuators is not a spurious gain of function.

In contrast to NPAS1+FOXP2+ neurons, the projection and function of NPAS1+NKX2.1+ neurons are less well characterized. These neurons project to the midbrain, the cortex and the reticular thalamus14,28,30. However, it is unclear whether a single NPAS1+NKX2.1+ neuron projects to multiple brain regions or whether separate neurons are responsible for each of these projections.

The GPe projections to the cortex arise from NPAS1+NKX2.1+ neurons and ChAT+ neurons28,30. A mixture of neurons in the cortex has been shown to receive this GPe input30,106,107. While the functions of these projections await further characterization, recent data suggest that they are involved in regulating sleep84,108,109. Additionally, indirect evidence suggests that NPAS1+NKX2.1+ neurons may contribute to other non-motor functions. Specifically, NKX2.1+ neurons that are PV− and FOXP2− innervate lateral habenula-targeting neurons in the GPi; however, it is possible that these neurons are LHX6+SOX6− rather than NPAS1+. Nonetheless, these data further implicate GPe neurons in the limbic functions of the basal ganglia2,83.

CuiDuChangEtAl21

Cui, Q., Du, X., Chang, I. Y. M., Pamukcu, A., Lilascharoen, V., Berceau, B. L., García, D., Hong, D., Chon, U., Narayanan, A., Kim, Y., Lim, B. K., & Chan, C. S. (2021). Striatal Direct Pathway Targets Npas1+ Pallidal Neurons. Journal of Neuroscience, 41(18), 3966–3987. https://doi.org/10.1523/JNEUROSCI.2306-20.2021

[rcoreilly]

CourtneyPamukcuChan23 inaccuracies?

Is it true that GPePr mainly inhibits itself and Pt, but not Ak? CourtneyPamukcuChan23 suggests that GPeAk does not get much GPe lateral connectivity. True? No.

Is it true that STN mainly projects to Pr? No.

KetzefSilberberg21

Ketzef, M., & Silberberg, G. (2021). Differential Synaptic Input to External Globus Pallidus Neuronal Subpopulations In Vivo. Neuron, 109(3), 516-529.e4. https://doi.org/10.1016/j.neuron.2020.11.006

Anatomical data (Sadek et al., 2007) and ex vivo recordings in slices (Bugaysen et al., 2013) have revealed only sparse connectivity among GPe cells, but it has been predicted that a primary intra-pallidal synaptic pathway exists from prototypic to arkypallidal cells (Nevado- Holgado et al., 2014).

These results show that prototypic cells provide strong and reliable inhibition to arkypalli- dal cells and are also interconnected among themselves. It also suggests that, in addition to afferent excitation, arkypallidal cells are disinhibited during up-states because of the reduction in spiking of prototypic cells.

Thus, our findings, in vivo and ex vivo, show very low connectivity from arkypallidal to prototypic cells and among arkypallidal cells. These data, combined with the strong connectivity observed from NKX2.1-Cre cells (Figure 3), point towards a highly non- reciprocal connectivity between the prototypic and arkypallidal populations.

These results show that, although the STN provides excitatory synaptic input to both GPe populations, the effect of this input is cell-type-specific and shaped by intra-pallidal connectivity.

• similar strong initial excitation from STN -- but Ak then gets inhibited by Pr, and turns off, releasing the break on Go.

To study the inputs from dMSNs to GPe cells, we virally ex- pressed ChR2 in the striatum of D1-Cre mice and positioned an optic fiber in the dorsolateral striatum (Figures 7A and 7B). Of 33 prototypic cells recorded, only 6 (18.18%) showed measur- able responses to photostimulation (Figures 7C–7E). In contrast, all arkypallidal cells were inhibited by dMSN photosti- mulation, as seen by short-latency hyperpolarization of their membrane potential and suppression of spiking (n = 5; Figures 7C and 7F–7I). The amplitude of inhibitory responses was larger in arkypallidal than in prototypic cells (P, �0.58 ± 0.24; A, �10.99 ± 2.62 mV; p < 0.001, Mann-Whitney test, n = 33 pro- totypic and 5 arkypallidal cells; Figure 7J), but no difference was found in onset delay (P, 8.6 ± 1.28; A, 8.08 ± 1.14 ms; p = 0.99, two-sample t test, n = 6 prototypic and 5 arkypallidal cells; Figure 7K). These results indicate a strong bias in dMSNs input to the GPe, with strong and prevalent inhibition of arky- pallidal cells and only sparse and weak inhibition of prototypic cells.

• dMSN -> GPeAK preferentially, consistent with CuiDuChangEtAl21

AristietaBarresiAzizpourLindiEtAl21

Aristieta, A., Barresi, M., Azizpour Lindi, S., Barrière, G., Courtand, G., de la Crompe, B., Guilhemsang, L., Gauthier, S., Fioramonti, S., Baufreton, J., & Mallet, N. P. (2021). A Disynaptic Circuit in the Globus Pallidus Controls Locomotion Inhibition. Current Biology, 31(4), 707-721.e7. https://doi.org/10.1016/j.cub.2020.11.019

Altogether, these results indicate that the response of arkypallidal neurons to D2-SPN inputs is principally driven by a disinhibition from prototypic neurons, with only a small contribution from the direct STN excitatory drive.

Such differential firing responses toward STN inputs is consistent with our previous findings in rats in which opto-excitation of STN neurons led to quick excitation in prototypic neurons but inhibition in arkypallidal neurons26. Thus, similar to D2-SPN inputs, activation of STN glu- tamatergic afferents impacts preferentially onto prototypic and not arkypallidal neurons.

• Ok, there is a conflict in these results! Mallet lab shows no direct STN -> Ak, while above shows it is there. In either case, the Pr --o Ak inhibition is robust, in contradiction to Courtney et al. That is a key element of original Pcore.

[rcoreilly]

CuiDuChangEtAl21: key data on dSPN -> AK preferential

Black: = naive; Red = 6-OHDA

• roughly 1 order of magnitude larger impact from dSPN -> AK (Npas1+) -- but defintely hits both

• generally symmetric with iSPN effects the other way.

[rcoreilly]

MizutaniTakahashiOkamotoEtAl17

Mizutani, K., Takahashi, S., Okamoto, S., Karube, F., & Fujiyama, F. (2017). Substance P effects exclusively on prototypic neurons in mouse globus pallidus. Brain Structure and Function, 222(9), 4089–4110. https://doi.org/10.1007/s00429-017-1453-8

Substance P (SP) released only by dSPNs, activates PV+ but not AK (FoxP2+)

CuiDuChangEtAl21 on this paper:

Because substance P is excitatory (but 2nd messenger mediated?) the pattern is synergistic!

[rcoreilly]

GPePr -> Striatum does exist, goes to FSI interneurons

SaundersHuangSabatini16

Saunders, A., Huang, K. W., & Sabatini, B. L. (2016). Globus pallidus externus neurons expressing parvalbumin interconnect the subthalamic nucleus and striatal interneurons. PLOS ONE, 11(2), e0149798. https://doi.org/10.1371/journal.pone.0149798

PV+ neurons also innervate the striatum. Ret- rograde labeling identifies ~17% of pallidostriatal neurons as PV+, at least a subset of which also innervate the STN and SNr.

Optogenetic experiments in acute brain slices demon- strate that the PV+ pallidostriatal axons make potent inhibitory synapses on low threshold spiking (LTS) and fast-spiking interneurons (FS) in the striatum, but rarely on spiny projec- tion neurons (SPNs). Thus PV+ GP neurons are synaptically positioned to directly coordi- nate activity between BG input nuclei, the striatum and STN, and thalamic-output from the SNr.

RM Costa also mentioned same thing

Will try to get ref.

[rcoreilly]

Summary of Bio

• GPePr prototypic PV+ neurons, PV+KCNG4+, maybe 45% (50% total PV+): iSPN (D2, No) --o Pr --o SNr; also inhibit other GPe cells (self and GPePt, but not much GPeArky), and bidirectionally connect with STN.
• GPePf PV+ PF (parafasicular thalamus) projecting neurons, PV+LHX6-, maybe 5%: dSPN (D1, Go) --o Pf --o PFthal
• GPeAk arkypallidal NPAS1+FOXP2+, 18% (60% of 30% NPAS1+): dSPN (D1, Go) --o Arky --o dStr (not much iSPN, STN, or local input) -- basically a tonic brake on SPN that can be inhibited in a "Go" scenario, allowing broader activity (disinhibition).
• GPePt "patch-like" (striosomal) projection pattern, NPAS1+NKX2.1+, 12% (40% of 30% NPAS1+): dSPN (D1, Go) --o GPePatch --o SNc, TRN, Cortex, LHb; also recv from cortex. Maybe get most lateral input from GPePr?
• [remaining 20% are heterogeneous, 5% ChAT+ cholinergic, 15% PV-NPAS1-]
• Everyone gets PF and L5 cortical input (excitatory).

Functional principles

• Classical "Go - No" opponent dynamic: dSPN --o SNr and iSPN --o GPePr --o SNr -- result is a balance between these two opposing pathways.
• hyperdirect STN "wait for it" (vs. "hold your horses"): initial cortical excitation of STN directly excites SNr, which thus further strengthens the "brakes" on overall BG output (todo: key to establish how strong this is!). STN also projects strongly to GPePr (largely selectively). To the extent that STN -> SNr is stronger than STN -> GPePr, this is a net inhibitory dynamic. However, GPePr inhibits STN, which can then release excitation onto SNr. Thus, GPePr is in the driver seat potentially. Need a lot more clarity here.
• GPeAK "release brakes" -- it is directly inhibiting striatum, if it is inhibited, then striatum is more excitable.

[rcoreilly]

VS Learning rules (MSN / SPN)

Key principles:

• With ACh & CS & vmPFC activity, Go & No should both always fire -- ACh modulation can be strong enough to prevent otherwise during goal maintenance. Everything otherwise is always being evaluated.

• Random partial connectivity drives subset of SPN for any given situation, drives etc: the SPN rep is more-or-less determined by the inputs, and the job of SPNs is to represent proper Go / No balance over those units. For now at least, there isn't any meaningful representational learning.

• To ensure continued ability to learn for any outcome, both Go and No must continue to fire at gating time. This probably means some non-zero lower limit on weights? For Go case, it can disinhibit No via GPeAk pathway, but if No wins, Go still has to have some activity.

• Sign flip logic: if Go wins & gating happens, then US DA represents outcome for selected goal, and normal learning goes through. If No wins and no gating, is there anything meaningful? Something else will have to be selected to get an outcome. That other thing will learn. The things you said No to before that could get some learning? But that really doesn't make a lot of sense.. Definitely don't start with this -- KISS.

• S = sending, R = recv (MSN), A = ACh, D = DA

• Tr = gating trace

Basic 3-factor with trace:

• Tr = S * R
• dW = A * D * Tr
• Tr -= decay * A * Tr

[rcoreilly]

DS Learning rules

DS is more complicated than VS:

• VS just needs to do one go/no decision at start, and learn about how it turned out at the end.
• DS needs to make a sequence of individual action decisions at each step, and then update all of those based on the outcome.
• It needs to get a "plus phase" for each action taken, which can drive learning relative to the BG input on that action.
• The motor plus phase originates in efferent copy signals coming back up from spinal cord, into DCN (deep cerebellar nuclei), which then project to VL (cerebellar thalamus, going into M1). How does this get into DS?
• Each "motor zone" (trunk, hands, eyes, etc) operates semi-independently, but should potentially get input about how the zone as a whole responded to conditionalize credit assignment. How does this relate to the plus-phase signal?
• The massive SNr -> PF -> MSN loops and CM -> MSN (generally "intralaminar thalamus" -> MSN) inputs together provide the same number of synapses onto MSNs as cortical inputs, providing the ingredients for both plus phase and zone-level signals. This is the biggest part that still needs to be resolved. Some data indicates PF targets mostly CINs, but the MSNs are receiving a lot of direct thalamic synapses regardless.
• There is a potential role for local DA release, likely mediated in part by CIN activity.

[rcoreilly]

SuryanarayanaHellgrenKotaleskiGrillnerEtAl19

Suryanarayana, S. M., Hellgren Kotaleski, J., Grillner, S., & Gurney, K. N. (2019). Roles for globus pallidus externa revealed in a computational model of action selection in the basal ganglia. Neural Networks, 109, 113–136. https://doi.org/10.1016/j.neunet.2018.10.003

This is the paper from which the original pcore model was developed.

Relative to this diagram, the current update raises the following questions:

• GP-TA (= GPeAk): does not self-inhibit, does not receive much direct STN input; is only source of inhibition back to SPNs; is activated specifically from dSPN (Go), not iSPN. In pcore, GPeTA recv only inhibition from GPeIn (GPePr).
• Could not find much on the Inner vs. Outer distinction, from SadekMagillBolam07

LindahlKotaleski16

Lindahl, M., & Kotaleski, J. H. (2016). Untangling Basal Ganglia Network Dynamics and Function: Role of Dopamine Depletion and Inhibition Investigated in a Spiking Network Model. eNeuro, 3(6). https://doi.org/10.1523/ENEURO.0156-16.2016

• Has TI (Pr) and TA (Ak) GPe cells, spiking, simulates various dynamics.
• Lots of params and citations that could be examined for reference.

The effective connectivity used in this model compares well to the predicted connectivity from a rate model by Nevado-Holgado et al. (2014) that was fitted to part of the same data set. They similarly predicted that the connec- tivity between MSN–GPe TI is strong, whereas MSN–GPe TA projections are weak. It is also shown here that TI–TA should be stronger than TA–TA, thus implying that TI neurons have a strong control over TA neurons. Both these findings also seem to be supported by experiments (Mallet et al., 2012). However Nevado-Holgado also pre- dicted that STN connectivity is stronger to TA than to TI. This is something we did not find; actually the conduc- tance of STN–TA synapses in our model is �3 times weaker than STN–GPe TI connections. However, we still saw that STN influences the activity in GPe TA in practice, since the STN input constitutes a larger proportion of the total excitatory input to GPe TA compared with the total excitatory input to GPe TI.

It seems overall that STN -> TA (Ak) is indeed weak. This model is missing the dSPN -> Ak projection.

Also, another paper showed that GPeAk do indeed have weaker Na excitatory currents -- need to track that one down again.

Nevado-HolgadoMalletMagillEtAl14

Nevado-Holgado, A. J., Mallet, N., Magill, P. J., & Bogacz, R. (2014). Effective connectivity of the subthalamic nucleus–globus pallidus network during Parkinsonian oscillations. The Journal of Physiology, 592(7), 1429–1455. https://doi.org/10.1113/jphysiol.2013.259721

Estimated params:

With PF:

• rate code
• Has TA <-> TA (weak in est) and TA -> TI prjns -- in general it is mostly TI -> TA and TI <-> TI, and est params are close but not perfect for that (TI <-> TI is very weak in model, should be stronger).
• Also has STN -> TA which is weak per above comment in LindahlKotaleski16 -- est is strong.

Thus, the genetic algorithm estimated that GP-TA neurons have substantially lower autonomous firing rates than GP-TI neurons.

Anatomical data suggest that intralaminar thalamic projections to STN are modest, at least when compared to thalamostriatal projections (Smith et al. 2004). However, our models indicate that the weight of this connection to STN is large (even greater than that of the corticosubthalamic projection), suggesting a paradoxically strong impact.

[rcoreilly]

STN

PrasadWallen-Mackenzie24

Prasad, A. A., & Wallén-Mackenzie, Å. (2024). Architecture of the subthalamic nucleus. Communications Biology, 7(1), Article 1. https://doi.org/10.1038/s42003-023-05691-4

Up to the moment review!

MRI techniques show compartmentalization of the STN anatomical regions40. These have been referred to as motor, associative and limbic regions, forming the so called tripartite model of STN, correlating anatomy with function43,44. These STN compartments are also known in the literature as territories, domains, subdivisions, subareas. The STN compartments run through the rostrocaudal, mediolateral and dorsoventral axes42

The dorsolateral compartment is associated with sensorimotor function, the ventromedial (or central) compartment with cognitive/associative function, and the medial-most compartment with limbic function45. The limbic area of the STN is often referred to as the limbic tip, and it is directly associated with the para-STN towards which no specific anatomical border is apparent.

Dopamine: The STN receives sparse dopaminergic projections from the substantia nigra pars compacta (SNc) and ventral teg- mental area (VTA) as reported in monkeys95,96 and humans73. Dopamine receptor subtype 2 (D2R) is distributed in the human STN in decreasing gradient along a ventromedial-to-dorsal axis76,97. D2R is co-localised with adenosine receptor A2A within the dorsal and medial regions of the human STN97. In non-human primates, Dopamine receptor subtype 1 (D1R) and D2R are distributed on presynaptic axons and putative gluta- matergic and GABAergic terminals98. In rats, D1R has high expression99. Further, firing rates of STN neurons are not altered by D2R stimulation, yet increased by selective stimulation of DRD1100.

The STN mainly receives inhibitory projections from globus pallidus (GP) externa (GPe) and extends glutamatergic projections to substantia nigra pars reticulata (SNr) and GPi 138,139. In line with the relay of the indirect pathway, optogenetic inhibition of STN neurons promotes movement117,131,140,141 while optogenetic excitation of the STN inhibits movement, reducing specific locomotion tasks117,131.

These are all very old refs, e.g., 138 = VanDerKooyHattori80 is main anatomical one -- not much news here.

Together, these results demonstrate that STN neuronal activity is capable of bidirectional motor regulation.

VanDerKooyHattori80

Van Der Kooy, D., & Hattori, T. (1980). Single subthalamic nucleus neurons project to both the globus pallidus and substantia nigra in rat. Journal of Comparative Neurology, 192(4), 751–768. https://doi.org/10.1002/cne.901920409

The present study suggests that the efferent projections of the rat STN are organized in a relatively simple and homogeneous manner. Virtually every STN neuron sends an axon collateral to both the globus pallidus and substantia nigra.

SmithHazratiParent90

Smith, Y., Hazrati, L.-N., & Parent, A. (1990). Efferent projections of the subthalamic nucleus in the squirrel monkey as studied by the PHA-L anterograde tracing method. Journal of Comparative Neurology, 294(2), 306–323. https://doi.org/10.1002/cne.902940213

These find- ings indicate that besides its well-known connection with the pallidum, the subthalamic nucleus gives rise to widespread projections to other components of the basal ganglia in primates.

JeonLeeKwonEtAl22

Jeon, H., Lee, H., Kwon, D.-H., Kim, J., Tanaka-Yamamoto, K., Yook, J. S., Feng, L., Park, H. R., Lim, Y. H., Cho, Z.-H., Paek, S. H., & Kim, J. (2022). Topographic connectivity and cellular profiling reveal detailed input pathways and functionally distinct cell types in the subthalamic nucleus. Cell Reports, 38(9), 110439. https://doi.org/10.1016/j.celrep.2022.110439

Given our findings that the IP_STN, HP_STN, and HP_GPe are topographically organized with gradient convergence at the mesoscopic level, we next performed a detailed analysis of HP connectivity at the cellular level.

Neurons innervating the STN are predominantly pyramidal tract (PT) neurons, whereas those innervating the GPe are PT and intratelencephalic (IT) neurons, with a subset (�15%) that collaterally innervates the GPe and STN (HPGPe/STN). The axonal density of neurons innervating the HPGPe/STN in the STN is 3.7-fold higher than that in the GPe. Consistent with our mesoscopic retrograde tracer data, shown in Figure 3A, we found that cortical neurons innervating the HPGPe/STN send collateral innervations to multiple brain areas, such as the striatum (STR), SN, SC, and others, which suggests that these neurons may help orchestrate movement.

We divided cortical subregions into three groups: motor, limbic, and association (further divided into prefrontal and rest association) areas. HP projections from the association, mo- tor, and limbic areas were only somewhat segregated in the ventromedial, dorsolateral, and ventrolateral areas, respectively, appearing along the longitudinal axis of the STN (axis 1). Howev- er, as we observed in the mouse HPSTN, there was a high degree of overlap and no evident anatomical borders or clear subdivi- sions (Figures S5D–S5H). We also found the HPSTN, HPGPe, and collateral HPGPe/STN organization in the human brain to be consistent with the complex signal integration patterns in the mouse. In agreement with our mouse connectivity maps, our analysis of human 7 T MRI-based tractography data revealed considerable and gradual overlap along the longitudinal axis of the human STN. These results suggest that graded convergence of IPSTN and HPSTN is organized along the STN geometric axes and that GPe neurons innervating the IPSTN might be activated by the same HPSTN.

Our serial multiplex smFISH data provide direct evidence that PV+ STN cells are glutamatergic with topographic segregation in the STN (Hontanilla et al., 1998; Walle ́n-Mackenzie et al., 2020; Wu et al., 2018) and suggest that PV cannot reliably serve as a marker for GABAergic interneurons, together with previous findings (Jinno and Kosaka, 2004; Shang et al., 2015; Wallace et al., 2017).

By clearing parasagittal sections (2.5 mm in thickness) that embrace the entire STN without making physical cuts, we 3D-mapped total PV+ neurons in the STN and found topographically segregated distribution in the dorsolateral and middle STN, consistent with serial smFISH data (Figure 6B; Video S2).

PV+ and PV� neurons displayed similar spontaneous firing and other basic properties, including input resistance (Table S4). However, only PV+ neurons showed phasic burst firing, which we refer to as firing pattern, composed of fast firing (an interspike interval [ISI] of less than 10 ms) and a pause; even current injection could not evoke bursts from PV� neurons (Figures 6C–6F and S7B– S7E; Table S5; STAR Methods). Burst and tonic firing patterns were observed at an equivalent membrane potential range (�58 to �45 mV) (Figure S7E), suggesting that burst firing was not due to differences in the membrane potential per se, even though firing patterns in STN neurons can be affected by membrane potentials (Kass and Mintz, 2006).

Our results provide clear evidence that there are at least two functional types of glutamatergic neurons in the STN and that glutamatergic PV+ neurons contribute to burst firing in the dorsolateral and middle STN. This is compatible with previous studies of the spatial distribution of firing patterns, with intrao- perative microelectrode recording within the human STN showing the occurrence of burst firing in in the dorsal STN and tonic firing toward the ventral STN (Kaku et al., 2019).

• burst PV+ specifically in DLS / motor. VS not

Furthermore, our results indicated that all four cell types in the STN and GPe (STN PV+, PV�, GPe arkypallidal, and prototypic neurons) communicate with each other and receive synaptic information from the cortex. Thus, in terms of cortex-GPe-STN circuitry, our data indicate that ‘‘everyone talks and everyone listens.’’

Next, in terms of connectivity patterns, the two types of STN neurons integrate excitatory and inhibitory signals (E/I) from the IPSTN and HPSTN. At the cellular level, the Gria1 to Gabra1 expression ratio was approximately similar in dorsolateral PV+ neurons, but this expression ratio was significantly higher in ventromedial PV� neurons (Figure 7I). Our results indicate that two types of glutamatergic PV+/� STN neurons, which integrate information from the IPSTN and HPSTN and fire in different patterns, are segregated along a topograph- ical distribution and express different ratios of genes encoding E/I receptors.

Gomez-OcadizSilberberg23

Gómez-Ocádiz, R., & Silberberg, G. (2023). Corticostriatal pathways for bilateral sensorimotor functions. Current Opinion in Neurobiology, 83, 102781. https://doi.org/10.1016/j.conb.2023.102781

• key paper summarizing PT vs IT projections and implications for local DA release by PT (ipsilateral) -> ACh pathways.

The striatum receives monosynaptic inputs from multiple brain regions including the cortex, the thalamus, the brainstem, and other BG nuclei [3]. Whereas most of these input pathways are ipsilateral, one striking exception is the corticostriatal IT pathway, which targets both striatal hemispheres. Pyramidal cells constituting the IT pathway consist of several molecularly-defined subtypes, some of which project only to the contralateral cortical hemisphere, while others project to both cortex and striatum [14,15]. Contralateral corticostriatal input from primary sensory cortices is relatively sparse compared to motor and prefrontal regions [16, 17, 18]. A recent study in mice showed that the amount of contralateral IT projections increases along the posterior-anterior cortical axis, with prefrontal cortical regions exhibiting the strongest and most symmetrical contralateral projections [19]. The other corticostriatal pathway is mediated by PT pyramidal neurons, which target the ipsilateral striatum as well as other subcortical targets including nuclei in the BG, thalamus, brainstem, and spinal cord (Figure 1). The PT population is also diverse, displaying different connectivity patterns with subcortical structures [15,20]. Of note, the “hyperdirect” pathway between the cortex and the subthalamic nucleus [21], and cortical projections to the substantia nigra pars reticulata [22], are mediated by ipsilateral corticostriatal projections that are part of the PT pathway [23,24].

Whereas both dMSNs and iMSNs receive cortical inputs, anatomical studies suggested a bias in innervation, with dMSNs being preferentially targeted by IT neurons and iMSNs by PT neurons [32,33]. Later functional studies using optogenetic activation of corticostriatal terminals showed comparable responses to IT and PT pathway stimulation in both MSN types [34], with relative strengths of corticostriatal inputs to dMSNs and iMSNs depending on the presynaptic cortical region and laterality [29].

A special case is the population of ChINs, constituting the only non-GABAergic striatal neuron type. ChINs receive cortical inputs primarily from the PT pathway, with very sparse inputs from the IT pathway [37], as seen also in the lack of inputs from the contralateral cortical hemisphere [29] (Figure 2). This difference between PT and IT innervation of ChINs directly affects the integration of cortical input by MSNs, resulting in biphasic excitatory responses to synchronous activation of the PT, but not IT, corticostriatal pathway [37]. Synchronized activation of ChINs induces also inhibitory responses in MSNs via polysynaptic interactions [38].

Notably, some of the signals conveyed through DA release, such as information about action value and RPE, are broadcast bilaterally, while some others, such as the strength of the action-outcome association, display striking lateralization. Interestingly, lateralization of DA dynamics can occur in specific striatal territories, as has been observed in the dorsal striatum but not in the ventral striatum during execution of a visual decision-making task [46].

An intriguing possibility is that the lateralization observed in DA signals conveyed to the striatum could emerge in the local microcircuit. As discussed in the preceding section, ChINs receive inputs primarily from the ipsilateral cortex via the PT pathway. This unilateral control of ChINs could represent a means to shape the activity of other neuromodulators in a single striatal hemisphere.

Recent studies have explored the local interactions of DA with other neuromodulators in the striatum, especially ACh. Ex vivo slice experiments have provided evidence that DA and ACh regulate each other's release in the striatum: ACh, released by ChINs, can trigger DA release through direct axonal depolarization mediated by nicotinic receptors [47], while DA in turn can inhibit ChINs and the release of ACh via D2 receptors [48, 49, 50]. The release of DA mediated by synchronous activation of ChINs can take place in response to activation of cortical terminals [51]. These interactions suggest that DA release in the striatum can be controlled in at least two different ways [52]: propagation of action potentials from DA cell bodies, and activation of nicotinic receptors in DA axons by local ChINs. In addition, DA terminals have been shown to co-release 5-HT [53], GABA, and glutamate [54,55]. These specific features suggest that corticostriatal inputs to ChINs, primarily from the PT pathway, are in a position to orchestrate the neuromodulatory landscape in the ipsilateral striatum (Figure 2).

These results provide evidence against neuromodulatory coordination between DA and ACh in dorsal striatum arising from direct intrastriatal interactions between DA and ACh in vivo. It is suggested that ACh fluctuations are instead controlled by extrastriatal glutamatergic afferents. Another recent study using fiber photometry to monitor the release of ACh and DA in the ventrolateral striatum during a reward-based decision-making task showed that these two neuromodulators’ release shows mostly anticorrelated transients, and that their dynamics are influenced by multiple behavioral variables [57]. However, DA release dynamics and reward prediction error (RPE) encoding were not disrupted by perturbations of ACh release by ChINs. Nonetheless, selective ablation of D2 receptors in ChINs showed that loss of DA regulation of ChIN activity impairs decision-making. Taken together, these studies argue against the physiological relevance of local DA-ACh interactions in the striatum in vivo. However, another new study monitoring DA release in the nucleus accumbens of awake, unrestrained rats while simultaneously measuring or manipulating ChIN activity [58] reported that ChIN stimulation evoked DA release in vivo. In addition, while rats engaged in an operant conditioning task [52], it was shown that fast ramps in ChIN activity and DA release co-occur during approach behaviors to either start a trial or to collect rewards. Notably, the increase in DA release co-occurring with ChIN activity ramps was independent of firing activity of DA neurons in the VTA. Furthermore, pharmacological blockade of nicotinic receptors disrupted performance in the task. These results support the role of ChINs as regulators of DA release during motivated behavior.

The divergent observations from the studies discussed above call for further work aimed to clarify if the interaction between DA and ACh in vivo varies across specific territories of the local striatal microcircuit and across behavioral contexts.

NambuChiken24

Nambu, A., & Chiken, S. (2024). External segment of the globus pallidus in health and disease: Its interactions with the striatum and subthalamic nucleus. Neurobiology of Disease, 190, 106362. https://doi.org/10.1016/j.nbd.2023.106362

• Good overall convergence with my diagram, for AK and Pr subtypes, consistent with CourtneyPamukcuChan23 generally.

Critically, this shows the biphasic activation pattern in STN, and corresponding one in GPe -- need to look for explanation.

Based on the above findings, we can assume that HFD-P and LFD-B neurons observed in the monkey GPe in vivo presumably correspond to Type II and Type I GPe neurons in guinea pigs, and to protoGPe and arkyGPe neurons in rodents, respectively, although the evidence of molecular markers needs to be provided yet.

Besides, inhibitory, unidirectional local axon collaterals from protoGPe neurons to arkyGPe neurons were found within the rodent GPe (Fig. 1A) (Aristieta et al., 2021; Fujiyama et al., 2016; Ketzef and Silberberg, 2021; Mallet et al., 2012).

So far, it is not clear whether above-mentioned two types of rodent GPe neuronal circuits are applicable to primates including monkeys. In the major HFD-P GPe neurons in monkeys (Hasegawa et al., 2022; Kita et al., 2004) and humans (Nishibayashi et al., 2011), presumably corresponding to rodent protoGPe neurons, cortical electrical stimulation induced a triphasic response composed of early excitation, inhibition, and late excitation, each of which is mediated by the cortico-STN-GPe, cortico-iMSN-GPe, and cortico-iMSN-GPe-STN-GPe pathways, respectively (Fig. 1A, B). On the other hand, in the minor LFD-B GPe neurons in monkeys, presumably corresponding to rodent arkyGPe neurons, cortical electrical stimulation induced complex response patterns composed of excitation and inhibition (Hasegawa et al., 2022), which may be mediated by combinations of the cortico-arkyGPe, cortico-dMSN-arkyGPe, and cortico-iMSN/STN-protoGPe-arkyGPe pathways (Fig. 1A, B). More studies are needed to clarify what kind of neuronal
signaling is conveyed through these pathways, and how it affects arkyGPe activity.

In addition, signals through the cortico-STN-GPi/SNr hyperdirect pathway stop movement initiation by exciting GPi/SNr neurons (Fig. 1A) (Schmidt et al., 2013). On the other hand, arkyGPe neurons may also serve as another stop mechanism in rodents. ArkyGPe neurons were activated during stop trials and could play a role in cancelling ongoing movements through the inhibition of MSNs (Fig. 1A) (Gu et al., 2020; Mallet et al., 2016). Indeed, chemogenetic or optogenetic activation of Npas1+ arkyGPe neurons suppressed locomotion through arkyGPe-MSNs projections (Glajch et al., 2016; Pamukcu et al., 2020). Activation of arkyGPe neurons during stop trials may be mediated by the iMSN-protoGPe-arkyGPe pathway (Fig. 1A) because optogenetic activation of rodent iMSNs inhibited protoGPe neurons, disinhibited arkyGPe neurons through inhibitory protoGPe-arkyGPe projections, and induced locomotor inhibition (Aristieta et al., 2021). In contrast, optogenetic activation of STN neurons excited protoGPe neurons, inhibited arkyGPe neurons, and had minimal effects on locomotion (Aristieta et al., 2021).

In addition, the STN showed stop-related activity, and signals through the cortico-STN-GPi/SNr hyperdirect pathway excite GPi/SNr neurons (Fig. 1A), and stop movement initiation in monkeys as well (Pasquereau and Turner, 2017).

Cortically induced late excitation is mediated by the cortico-iMSN-protoGPe-STN-protoGPe pathway, supporting enhanced iMSN-protoGPe indirect pathway neurotransmission in the PD state. Dopamine replacement therapy reversed the enhancement of late excitation in monkeys (Chiken et al., 2021).

Dopamine depletion changed the preferred firing phases of protoGPe activity by ~180◦ in anesthetized rats, and made protoGPe neurons to preferentially discharge during the inactive (surface-negative) phase of the cortical slow oscillations (Abdi et al., 2015; Mallet et al., 2008). The coupling of protoGPe firings to the inactive phase may be mediated by the enhanced cortico-iMSN-protoGPe-STN-protoGPe pathway with long latency (Fig. 1A, C). In contrast, the preferred firing phases of arkyGPe activity remained unchanged in the PD state in rats (Abdi et al., 2015; Mallet et al., 2008).

[rcoreilly]

FangCreed24

Fang, L. Z., & Creed, M. C. (2024). Updating the striatal–pallidal wiring diagram. Nature Neuroscience, 27(1), 15–27. https://doi.org/10.1038/s41593-023-01518-x

• Nice recent review mostly consistent with the above. Emphasizes TavernaIlijicSurmeier08 for direct D2 -> D1 competition within striatum -- todo: explore that.

• Also critically provides recent updates on VP (ventral pallidum): strong evidence for opposing Glu vs. GABA projections to LHb: don't need the disynaptic "nogo" in PVLV model for net excitation of LHb -> inhibition of VTA!

• And good evidence for GPeAk in ventral side.

At the molecular level, VP PV+ neurons analogously release GABA within the VTA, and receive monosynaptic inputs from the NAc, midbrain and STN132, akin to their prototypical counterparts in the GPe. It is not established whether low-firing LHX6+ cells constitute a subclass of prototypical VP neurons and to what extent these cells may overlap with PV+VP neurons. At the anatomical level, unlike the GPe, prototypi- cal VP populations project both internally within the basal ganglia to the ventromedial STN and GABAergic neurons of the VTA (‘relay’ neurons), and outside the basal ganglia to the LHb and mediodorsal thalamus (‘output’ neurons). Neurochemically, all GPe prototypical neurons are GABAergic (Fig. 4a), whereas prototypical VP neurons are both glutamatergic (VPGlu; ~18%) and GABAergic (VPGABA; ~49%), with minimal overlap between these populations (0.9–3%)133,134 (Fig. 6).

This neurochemical identity carries important functional implica- tions, as VPGlu and VPGABA neurons synapse in overlapping territories in the LHb and midbrain, oppositely influencing activity in these struc- tures133,134. Specifically, stimulation of VPGlu neurons drives activity in the LHb, rostromedial tegmental nucleus (RMTg) and VTAGABA neurons, and inhibits dopaminergic VTA (VTADA) neurons through polysyn- aptic inhibition134. By contrast, stimulation of prototypical VPGABA neurons inhibits LHb and VTAGABA neurons, and disinhibits VTADA cell firing through the same polysynaptic network133. Consequently, VPGlu and VPGABA projection neurons exert opposing roles on appetitive behavior. VPGlu neurons are activated in response to reward omission and punishment and are inhibited by reward and reward-predictive cues135. Activating VPGlu is acutely aversive, and their activity is neces- sary to update reward-seeking behavior in response to conflict-induced negative outcomes, which depends on LHb and RMTg activity134,135.

Conversely, stimulation of VPGABA supports behavioral reinforcement, and VPGABA activity is necessary for the expression of reward seek- ing133,136,137. Although endogenous activity of VPGABA neurons is more heterogeneous than that of VPGlu neurons, a large proportion (~82%) exhibit activity patterns which are diametrically opposed to those of VPGlu neurons, as they are inhibited by punishment and activated by reward and reward-predictive cues135.

These opposing effects of VPGlu and VPGABA neurons on behavior have been ascribed to opposing influences on downstream struc- tures; however, their antagonistic effects are also probably ampli- fied by local connectivity within the VP. Specifically, VPGABA neurons synapse onto VPGlu neurons, and would attenuate the suppression of appetitive behavior imposed by VPGlu neurons. Likewise, VPGlu neurons innervate and excite neighboring VPGlu neurons133, syn- chronizing and amplifying downstream actions. It is unclear how the opposing patterns of endogenous activity in prototypical VPGlu and VPGABA neurons arises, as both populations receive input from largely overlapping upstream structures, including the NAc, medial wall cortex, basolateral amygdala, ventromedial STN and VTA133,134,138. However, VPGlu neurons do receive proportionately more input from accumbal D1-SPNs than VPGABA neurons138. Thus, under conditions in which D1-SPNs are active, VPGlu neurons would be inhibited relative to neighboring VPGABA neurons, providing one mechanism through which accumbal D1-SPN activity disinhibits appetitive behavior at the level of the VP.

As mentioned above, parallel GABAergic and glutamatergic pal- lidal output neurons that characterize the VP are not observed in the GPe; rather they are reminiscent of output pathways of the GPi/EPN. Within the EPN, three primary populations of output neurons have been described from unbiased clustering of single-cell transcriptional profiles: two separate PV+ populations that release either glutamate or GABA, and a third population co-releasing GABA and glutamate that also express the peptide somatostatin139. In both the EPN and VP, the population of neurons projecting to LHb co-releases GABA and glutamate, and an exclusively GABAergic population preferentially innervates the midbrain SNr and VTAGABA neurons, respectively132,139. As in the VP, glutamatergic EPN cells are nearly exclusively innervated by D1-SPNs, and not D2-SPNs139. Thus, some unique populations of prototypical VP neurons may be considered analogous to GABAergic, glutamatergic or co-releasing output neurons of the EPN. Molecular profiling of the VP and careful anatomical validation will be required to resolve this question.

Arkypallidal neurons in both the GPe and VP send dense, filigree-like processes through large territories of the striatum, and their activity peaks during conditions requiring rapid suppression of action selec- tion63,64. Although arkypallidal neurons constitute only ~18% of pallidal neurons, their ability to inhibit the striatum is robust, with individual neurons forming over 10,000 striatal synaptic contacts over terminal fields in excess of 1 mm in rats62. Consequently, >80% of SPNs and >25% of striatal interneurons receive direct synaptic input from arkypal- lidal cells64,140. Arkypallidal neurons in the GP co-release GABA and enkephalin62,63,141, although this co-release has not been confirmed in the VP. This confers arkypallidal neurons with the ability to inhibit striatal neurons through fast ionotropic transmission (GABAA), slower G-protein coupled receptor transmission (GABAB), or mu and delta opioid receptors142. The conditions under which arkypallidal neurons release these distinct neurotransmitters is not known, although burst firing is typically required for enkephalin release143. Indeed, although arkypallidal neurons exhibit slow firing relative to other pallidal popu- lations, they can fire at up to ~75 Hz in the VP64 and over 100 Hz in the dorsal pallidus127, which is predicted to be sufficient for peptide release.

To date, no monosynaptic excitatory inputs to arkypallidal neu- rons in the GPe or VP have been conclusively identified, although input from thalamic or cortical nuclei has been proposed to account for the rapid onset of arkypallidal activity in response to task-relevant cues63. Characterization of cell-type-specific inputs will determine whether arkypallidal firing is inherited from specific upstream structures or arises within the arkypallidal population through disinhibition, neu- romodulation or integration of multiple excitatory inputs.

TavernaIlijicSurmeier08

Taverna, S., Ilijic, E., & Surmeier, D. J. (2008). Recurrent collateral connections of striatal medium spiny neurons are disrupted in models of Parkinson’s disease. Journal of Neuroscience, 28(21), 5504–5512. http://www.jneurosci.org/cgi/content/abstract/28/21/5504

• Key point: D2 -> D1 inhibition much more prevalent than D1 -> D2 -- direct competition!

Papers on VP Glu vs. GABA

FagetZellSouterEtAl18

Faget, L., Zell, V., Souter, E., McPherson, A., Ressler, R., Gutierrez-Reed, N., Yoo, J. H., Dulcis, D., & Hnasko, T. S. (2018). Opponent control of behavioral reinforcement by inhibitory and excitatory projections from the ventral pallidum. Nature Communications, 9(1), 849. https://doi.org/10.1038/s41467-018-03125-y

TooleyMarconiAlipioEtAl18 (Creed lab)

Tooley, J., Marconi, L., Alipio, J. B., Matikainen-Ankney, B., Georgiou, P., Kravitz, A. V., & Creed, M. C. (2018). Glutamatergic Ventral Pallidal Neurons Modulate Activity of the Habenula–Tegmental Circuitry and Constrain Reward Seeking. Biological Psychiatry, 83(12), 1012–1023. https://doi.org/10.1016/j.biopsych.2018.01.003

Stephenson-JonesBravo-RiveraAhrensEtAl20

Stephenson-Jones, M., Bravo-Rivera, C., Ahrens, S., Furlan, A., Xiao, X., Fernandes-Henriques, C., & Li, B. (2020). Opposing Contributions of GABAergic and Glutamatergic Ventral Pallidal Neurons to Motivational Behaviors. Neuron, 105(5), 921-933.e5. https://doi.org/10.1016/j.neuron.2019.12.006

FarrellEstebanFagetEtAl21

Farrell, M. R., Esteban, J. S. D., Faget, L., Floresco, S. B., Hnasko, T. S., & Mahler, S. V. (2021). Ventral Pallidum GABA Neurons Mediate Motivation Underlying Risky Choice. Journal of Neuroscience, 41(20), 4500–4513. https://doi.org/10.1523/JNEUROSCI.2039-20.2021

HeinsbroekBobadillaDereschewitzEtAl20

Heinsbroek, J. A., Bobadilla, A.-C., Dereschewitz, E., Assali, A., Chalhoub, R. M., Cowan, C. W., & Kalivas, P. W. (2020). Opposing Regulation of Cocaine Seeking by Glutamate and GABA Neurons in the Ventral Pallidum. Cell Reports, 30(6), 2018-2027.e3. https://doi.org/10.1016/j.celrep.2020.01.023

[rcoreilly]

Mechanisms of DA modulation of SPN excitability

FangCreed24

Taken together, phasic dopamine can alter excitatory drive onto an SPN ensemble to bias activation of D1-SPNs over D2-SPNs. This occurs through increased D1-SPN excitability (Fig. 3b) and decreased lateral inhibition of D2-SPNs onto D1-SPNs (Fig. 3c), ultimately favoring activation of D1-SPNs to D2-SPNs, and to execution of actions gated by that SPN ensemble.

However, the effect of dopamine on SPN excitability has been controversial, as signaling through the dopamine D1 receptor (D1R) has been reported to both potentiate89,90 and inhibit91 SPN Kir2 channels, which is the primary determinant of SPN membrane voltage in the hyperpolarized resting state92. Recent work sought to resolve this controversy by performing perforated patch-clamp recording of D1-SPNs to prevent dialysis of second-messenger signaling molecules while optogenetically emu- lating physiologically relevant dopamine release93. Under these con- ditions, phasic dopamine only negligibly altered resting membrane voltage, arguing against D1-mediated modulation of Kir2 channels. Instead, phasic dopamine robustly increased firing frequency and decreased latency to fire in response to current injections, suggesting facilitated upstate transitions (Fig. 3b). This effect was dependent on D1-mediated negative modulation of outward potassium currents mediated by Kv1.2 channels94,95 (Fig. 3b). By facilitating D1-SPN firing frequency, dopamine would facilitate spike-timing-dependent plastic- ity of excitatory input through back-propagation of action potentials and relief of magnesium block from NMDA receptors in D1-SPNs96. Over slower timescales, D1R signaling through canonical PKA-dependent cascades can potentiate excitatory transmission onto D1-SPNs by promoting the insertion and stabilization of AMPA receptors into the postsynaptic membrane97,98 (Fig. 3b).

93 = LahiriBevan20

LahiriBevan20

Lahiri, A. K., & Bevan, M. D. (2020). Dopaminergic Transmission Rapidly and Persistently Enhances Excitability of D1 Receptor-Expressing Striatal Projection Neurons. Neuron, 106(2), 277-290.e6. https://doi.org/10.1016/j.neuron.2020.01.028

Together, these findings indicate neither appre- ciable negative nor positive modulation of Kir2 channels. In contrast, D1-SPN responses to suprathreshold current steps during the same trials revealed strong dopaminergic modulation of firing frequency and latency to fire (Figures 5E–5G).

Outward currents generated by Kv and Ca2+-dependent K+ channels are also thought to play an important role in regulating the frequency and pattern of D1-SPN activity. For example, (1) slowly inactivating Kv1.2-containing channels limit the overall level of up-state depolarization and contribute to action potential afterhyperpolarization (AHP) (Nisenbaum et al., 1994; Shen et al., 2004), and (2) big- and small-conductance K+ channels (BK and SK, respectively), through their selective coupling to Ca2+ entry via Cav2.1 and Cav2.2 channels, contribute to the fast and me- dium-duration components of the AHP (fAHP and mAHP, respectively) (Pe ́ rez-Garci et al., 2003; Pineda et al., 1992; Vilchis et al., 2000). Because D1R activation can reduce Kv1.2-medi- ated currents as well as both Cav2.1 and Cav2.2 currents (Surmeier et al., 1995; Surmeier and Kitai, 1993), native dopami- nergic modulation could potentially elevate firing through atten- uation of the AHP and depolarization of the interspike interval. We therefore examined D1-SPN fAHP and mAHP peak voltages before and after optogenetic stimulation of DA axons. Using the long up-state protocol, AHPs were measured during the 2 s immediately preceding DA axon stimulation (pre-stimulus) and during the final 2 s of the current step (post-stimulus). We observed a reduction in both the fAHP and mAHP on optical- stimulation but not current-only trials (Figures 6C–6E and S5E).

PragerDormanHobelEtAl20

Prager, E. M., Dorman, D. B., Hobel, Z. B., Malgady, J. M., Blackwell, K. T., & Plotkin, J. L. (2020). Dopamine oppositely modulates state transitions in striosome and matrix direct pathway striatal spiny neurons. Neuron, 108(6), 1091-1102.e5. https://doi.org/10.1016/j.neuron.2020.09.028

Prager et al. show that dopamine promotes the maintenance of dendritically evoked ‘‘up-states’’ in mouse direct pathway matrix SPNs but opposes it in striosomes. This requires postsynaptic D1 receptors and involves differential engagement of L-type Ca2+ channels.