Brain iron deficiency in RLS

Brain iron deficiency (BID) is one of the two core pathophysiological pillars of RLS named in this project's anchor facts (Master Plan Section 3), alongside altered dopamine signalling that BID is thought to drive. This synthesis covers the CSF, autopsy, MRI, transcranial-sonography, hepcidin/iron-regulatory, genetic, and animal-model evidence for BID, and what it implies for a non-drug, iron-first long-term management approach. Every claim below cites a study note (filename) or PMID from this knowledge base.

What is established (grade A/B)

Genuinely A/B-graded evidence specific to brain iron deficiency (as opposed to peripheral iron-deficiency-RLS epidemiology, which is a separate topic) is thin — most of the direct human mechanistic evidence below is grade C, and the one large in-vivo-MRI meta-analysis is a negative finding (see Contested, below). What can be stated at B-grade or higher:

• In a randomized, double-blind, placebo-controlled trial, IV ferric carboxymaltose (1500 mg) in RLS patients with iron deficiency anemia produced significantly greater symptom improvement than placebo at 6 weeks (IRLS change -13.6 vs -3.6, p = 0.011), and this clinical benefit was accompanied by MRI-measurable increases in brain iron (QSM/R2*) in the thalamus, substantia nigra and pulvinar, with the thalamic QSM change correlating with symptom improvement (p = 0.028) — grade B (single RCT) (2023-bae-fcm-qsm-rct.md, PMID 37437492).
• The AASM 2024/2025 clinical practice guideline, built on a GRADE systematic review, makes a good-practice statement that all patients with clinically significant RLS should have serum ferritin and transferrin saturation (TSAT) checked, and gives a strong recommendation (moderate certainty) for IV ferric carboxymaltose when ferritin is ≤75-100 ng/mL or TSAT <20% — the guideline document itself is graded D (guideline/expert consensus) under this project's rubric, but the specific FCM recommendation it contains is explicitly moderate-certainty evidence (2025-winkelman-aasm-guideline-rls.md, PMID 39324694).

Beyond these two anchors, the case that BID is causally central to RLS rests on the convergence of many independent grade-C lines of evidence (CSF, autopsy, genetics, pharmacologically-validated animal models) rather than on a single high-grade study — this convergence is itself the strongest argument for the model, but it should not be overstated as "proven" by any single high-grade trial.

Emerging C

• CSF ferritin is reduced and CSF transferrin elevated in idiopathic RLS despite normal serum ferritin/transferrin, the original biochemical evidence for brain-specific iron deficiency (2000-earley-csf-ferritin-transferrin.md, PMID 10762522). The CSF-ferritin deficit is most pronounced in early-onset (<45y) RLS and correlates with age of onset (r=0.64) (2005-earley-csf-ferritin-onset-phenotype.md, PMID 16268375).
• Autopsy studies of substantia nigra show a coordinated iron-management protein signature: decreased ferritin, DMT1, ferroportin and transferrin receptor (TfR) in neuromelanin cells, with paradoxically decreased (not compensatorily increased) TfR linked to reduced iron regulatory protein 1 (IRP1) activity (2004-connor-transferrin-receptor-neuromelanin.md, PMID 15136682). The same TfR-fails-to-rise, IRP-dysregulated pattern was later found at the choroid plexus and cortical microvessels — i.e., at the blood-brain interface itself (2011-connor-brain-iron-acquisition-bbb.md, PMID 21398376).
• Iron appears to be mis-compartmentalized rather than simply absent: mitochondrial ferritin is increased and cytosolic H-ferritin decreased in RLS substantia nigra neurons (2009-snyder-mitochondrial-ferritin-nigra.md, PMID 19816198), and iron deficiency triggers pseudo-hypoxic HIF-1α/HIF-2α/VEGF pathway activation in nigral neurons and cortical microvessels without evidence of true tissue hypoxia (2011-patton-hif-pathway-activation.md, PMID 21985026).
• Downstream, RLS autopsy tissue shows increased/activated tyrosine hydroxylase (more dopamine synthesis capacity) alongside decreased putamen D2 receptor density correlating with severity — reproduced in animal and cell iron-deficiency models (2009-connor-dopaminergic-profile-autopsy.md, PMID 19467991).
• Hepcidin (the master iron-release hormone) is dysregulated in a site-specific and non-linear way: CSF pro-hepcidin is reduced in early-onset RLS while tissue pro-hepcidin is elevated in nigra/putamen (2006-clardy-hepcidin-ferroportin-signaling.md, PMID 16759669); serum hepcidin is elevated even in RLS patients with entirely normal ferritin, with a complex U-shaped (not linear) relationship to severity (2018-dauvilliers-hepcidin-severity-rls.md, PMID 29418021).
• MRI relaxometry/QSM studies, taken individually, generally support regional (substantia nigra, thalamus, dentate nucleus, striatum) iron reduction that scales with severity or normalizes with treatment (2001-allen-mri-r2-brain-iron.md, PMID 11160969; 2016-li-qsm-7tesla-iron.md, PMID 27544840; 2023-kim-qsm-longitudinal-iron-therapy.md, PMID 37257418), including in secondary/hemodialysis-related RLS superimposed on globally elevated brain iron (2020-wang-qsm-hemodialysis-iron.md, PMID 32045852) and in the paradoxical case of systemic iron-overload (haemochromatosis) with focal brain iron deficiency (2005-haba-rubio-hemochromatosis-rls.md, PMID 15965214).
• Transcranial sonography (TCS) reliably detects substantia nigra hypoechogenicity (low nigral iron) in RLS with good diagnostic accuracy (AUC 0.91, sensitivity 82%, specificity 90% at a <0.2 cm² cutoff), in direct contrast to the hyperechogenicity of Parkinson's disease (2005-schmidauer-tcs-nigral-hypoechogenicity.md, PMID 16037973; 2007-godau-sn-hypoechogenicity-marker.md, PMID 17133515), validated in a 188-person cohort (2007-godau-sonographic-brainstem-cohort.md, PMID 18024170), and baseline TCS severity predicts the degree of symptom and dopaminergic-dose-reduction benefit from IV iron (2019-garcia-malo-tcs-iron-therapy-predictor.md, PMID 31875533).
• The MEIS1 RLS risk haplotype reduces MEIS1 mRNA/protein expression in brain tissue (2009-xiong-meis1-expression-brain.md, PMID 19126776) and this reduced expression bidirectionally alters iron-handling gene expression (ferritin, DMT1) in a worm ortholog/human-cell model, while iron deficiency itself further suppresses MEIS1 expression — a feed-forward genetic-metabolic loop (2011-catoire-meis1-iron-homeostasis.md, PMID 21710629).
• Peripheral iron delivery to the brain is a regulated, not passive, process: in a blood-brain-barrier cell-culture model, transferrin-bound iron transport tracked with plasma hemoglobin (erythropoietic demand) but not with serum ferritin (2020-connor-peripheral-iron-csf-communication.md, PMID 32295615) — direct mechanistic support for why ferritin under-detects brain iron deficiency (see Master Plan anchor facts).
• A 2026 mechanistic study identified a specific molecule behind the defective TfR1 response: CSF from RLS patients reduces iron uptake in iPSC-derived brain endothelial cells via elevated extracellular-vesicle miR-124-3p that binds and destabilizes TfR1 mRNA (2026-palsa-csf-bbb-endothelial-mir124.md, PMID 41947413).
• Animal models causally link diet-induced brain iron deficiency (often combined with A11 dopaminergic lesioning) to RLS-like hyperactivity that is reversible with D2/D3 agonists and worsened by D1 agonists (2007-qu-a11-lesion-iron-deprivation-mouse.md, PMID 17483695), and non-human-primate iron depletion reproduces a peripheral/CSF iron dissociation with increased striatal dopamine/serotonin turnover (2015-hyacinthe-primate-iron-depletion.md, PMID 25662508); a 2022 review argues these rodent models validly reproduce the human dopaminergic phenotype and have already identified adenosinergic/opioid pathways as additional non-dopaminergic targets (2022-earley-bid-rodent-model-review.md, PMID 35779614).
• Pre-existing notes from other topic sweeps in this knowledge base extend the mechanistic chain from BID to hyperarousal: BID in rodents downregulates adenosine A1 receptors (which normally restrain glutamate release) more consistently than it upregulates A2A receptors (2016-quiroz-adenosine-a1-receptor-downregulation.md); loss of A1R "unmasks" hypersensitive corticostriatal glutamatergic terminals that also facilitate striatal dopamine release, unifying the hyperdopaminergic and hyperglutamatergic models (2018-ferre-pivotal-adenosine-neurotransmission.md); this A1R-gated glutamate hypersensitivity was confirmed optogenetically and reversed by the A1R-boosting drug dipyridamole (2019-ferre-adenosine-corticostriatal-terminals.md); and BID in mice increases dopaminergic sensitivity of the opioid-receptor-co-expressing striatal-entopeduncular (GPi) pathway (2026-valle-leon-striatal-entopeduncular-opioid-dopamine-model.md), consistent with iron-deficient opioid-receptor-knockout mice independently developing an RLS-like anemic, circadian-hyperactive phenotype (2019-lyu-triple-opioid-receptor-knockout-mouse.md; 2020-lyu-mu-opioid-receptor-knockout-mouse.md).

Contested or negative findings

• The largest and most rigorous MRI study to date (72 RLS vs 72 controls) plus a formal meta-analysis of prior MRI literature found no significant pooled evidence of lower brain iron in RLS in any region, and in the authors' own cohort found increased, not decreased, R*/susceptibility in the caudate, putamen and red nucleus. Sensitivity analysis showed likely publication bias: only 5 of 8 previously meta-analyzed studies found significant substantia-nigra iron reduction, and a negative pooled effect for putamen/red nucleus/substantia nigra only appeared when this new dataset was excluded (2022-beliveau-mri-metaanalysis-revisited.md, PMID 35500370, grade A). This is the single most important caveat to the whole BID-by-MRI narrative.
• A methodological head-to-head in the same 77-person cohort showed that whether RLS patients appear to have lower nigral iron on MRI depends heavily on the relaxometry metric and region-of-interest method used — significant with a fixed-shape ROI, not significant (except in the red nucleus) with a semi-automated ROI (2015-moon-mri-relaxometry-methods.md, PMID 26257527). This suggests at least part of the field's inconsistency is measurement artifact, not necessarily an absent effect — but it also means individual positive MRI findings should be read cautiously.
• Serum ferritin, despite being the most widely used clinical iron marker and the primary threshold in guidelines, showed no correlation with RLS severity in a well-controlled cohort of drug-free, normal-ferritin RLS patients, while hepcidin did (non-linearly) (2018-dauvilliers-hepcidin-severity-rls.md, PMID 29418021) — a direct empirical demonstration of ferritin's limits as a brain-iron proxy, reinforcing rather than contradicting the broader BID model, but contesting ferritin's adequacy as its sole clinical marker.

Mechanistic picture in 5 bullet points

1. Transport bottleneck, not just supply shortage: iron acquisition across the choroid plexus and blood-brain-barrier microvasculature is impaired at the receptor/regulatory level (reduced transferrin receptor expression despite low cellular iron, defective IRP1 activity, and — per a 2026 mechanistic study — miR-124-3p-driven destabilization of transferrin receptor mRNA), so the brain can remain iron-deficient even when peripheral iron stores look adequate (2004-connor-transferrin-receptor-neuromelanin.md; 2011-connor-brain-iron-acquisition-bbb.md; 2020-connor-peripheral-iron-csf-communication.md; 2026-palsa-csf-bbb-endothelial-mir124.md).
2. Systemic hormonal gatekeeping compounds the problem: hepcidin, the hormone that restricts iron export from cells into plasma, is dysregulated in RLS in a site-specific, non-linear way and remains elevated even with normal ferritin, plausibly restricting iron flow toward the brain independent of total-body iron stores (2006-clardy-hepcidin-ferroportin-signaling.md; 2018-dauvilliers-hepcidin-severity-rls.md).
3. Cellular iron gets mis-routed and read as "hypoxia": within iron-short neurons, iron is shifted toward mitochondrial storage at the expense of cytosolic availability, and the resulting cellular iron shortage activates hypoxia-inducible-factor signalling (HIF-1α/2α, VEGF, nitrosative stress) even without true oxygen deprivation (2009-snyder-mitochondrial-ferritin-nigra.md; 2011-patton-hif-pathway-activation.md).
4. Iron deficiency reshapes neurotransmission along two convergent axes: it drives a presynaptic hyperdopaminergic state (increased tyrosine hydroxylase) with reduced postsynaptic D2 signalling, and separately downregulates adenosine A1 receptors, unmasking corticostriatal glutamatergic hypersensitivity that itself further facilitates dopamine release — linking BID to both the classic dopaminergic model and the newer hyperarousal/glutamatergic model of RLS (2009-connor-dopaminergic-profile-autopsy.md; pre-existing notes 2016-quiroz-adenosine-a1-receptor-downregulation.md, 2018-ferre-pivotal-adenosine-neurotransmission.md).
5. Genetic risk and acquired iron deficiency reinforce each other: the MEIS1 RLS risk haplotype lowers MEIS1 expression, which independently perturbs ferritin/DMT1 expression, while iron deficiency itself further suppresses MEIS1 — a feed-forward loop that may explain why some individuals develop clear brain iron deficiency despite normal diet/peripheral stores, and why others need more aggressive or repeated iron repletion (2009-xiong-meis1-expression-brain.md; 2011-catoire-meis1-iron-homeostasis.md).

Implications for a non-drug long-term plan for people currently on dopamine agonists or alpha-2-delta ligands

• Test iron status properly, not just once. Ferritin and transferrin saturation should be checked in the morning, iron-supplement-free for 24 hours, per the AASM good-practice statement (2025-winkelman-aasm-guideline-rls.md). Because serum ferritin systematically under-detects brain iron deficiency — erythropoietic (hemoglobin) demand appears to take precedence over brain iron delivery, and the BBB transport step itself can be defective independent of ferritin — a "normal" ferritin should not be used alone to rule out an ongoing brain iron contribution to symptoms; TSAT, and where available a history of iron-responsiveness, should be weighed alongside it (2020-connor-peripheral-iron-csf-communication.md; 2018-dauvilliers-hepcidin-severity-rls.md; 2026-palsa-csf-bbb-endothelial-mir124.md).
• Iron repletion is a mechanistically targeted, non-dopaminergic first step, not just a lab-value correction. IV ferric carboxymaltose has RCT-level B evidence of both symptom relief and measurable brain iron increase within 6 weeks in iron-deficient RLS patients (2023-bae-fcm-qsm-rct.md), with further iron-content changes and symptom correlation still evolving out to 24 weeks (2023-kim-qsm-longitudinal-iron-therapy.md) — consistent with the AASM's strong recommendation for IV FCM when ferritin ≤75-100 ng/mL or TSAT <20% (2025-winkelman-aasm-guideline-rls.md).
• Nigral-iron severity may help select who can taper. In one small, non-blinded cohort, patients with the most severe baseline substantia-nigra hypoechogenicity on TCS showed the largest post-IV-iron symptom improvement and the highest rate (57%) of achieving a ≥30% dopaminergic dose reduction, with 17% (3/30) fully discontinuing dopaminergic medication (2019-garcia-malo-tcs-iron-therapy-predictor.md). This is suggestive, not proof, of a reliable tapering biomarker (small, unblinded, single study) — and per this project's non-negotiable honesty rules, any dopamine-agonist tapering must be physician-supervised and gradual, never abrupt, given the risks of augmentation and dopamine-agonist withdrawal syndrome (DAWS).
• Alpha-2-delta ligands pair naturally with an iron-first strategy. Because gabapentin/pregabalin/gabapentin enacarbil act on a different (glutamatergic/calcium-channel) mechanism than dopamine and are AASM-preferred over dopamine agonists regardless of iron status, combining ongoing iron monitoring/repletion with an alpha-2-delta ligand (rather than a dopamine agonist) is an evidence-aligned way to manage residual symptoms without incurring augmentation risk while iron stores are being rebuilt (2025-winkelman-aasm-guideline-rls.md).
• Set expectations for ongoing monitoring, not a one-time fix. Because a genetic (MEIS1-linked) predisposition to impaired brain iron handling exists in a meaningful subset of patients, and because MRI evidence for "cure" of brain iron deficiency is itself contested, patients on a long-term drug-tapering plan should expect periodic re-testing of iron status and possible repeat iron infusions rather than assuming one round of repletion is permanent (2009-xiong-meis1-expression-brain.md; 2011-catoire-meis1-iron-homeostasis.md; 2022-beliveau-mri-metaanalysis-revisited.md).

Open questions

• Does the degree of IV-iron-induced change in brain iron (by TCS or QSM) reliably predict who can sustainably reduce or stop dopamine agonists long-term? Current evidence is a single small, non-blinded, short-follow-up cohort (2019-garcia-malo-tcs-iron-therapy-predictor.md; 2023-kim-qsm-longitudinal-iron-therapy.md).
• Why is in-vivo MRI evidence for brain iron deficiency so inconsistent — including a large, well-powered study and meta-analysis finding no pooled deficit and even some regional increases — when CSF, autopsy, and TCS evidence point more consistently toward deficiency? Is this measurement insensitivity/heterogeneity, true biological heterogeneity across patients, or does it mean the deficit is more localized/dynamic than current imaging can reliably capture (2022-beliveau-mri-metaanalysis-revisited.md; 2015-moon-mri-relaxometry-methods.md)?
• Could hepcidin-modulating therapy, or agents that restore transferrin-receptor expression / block miR-124-3p at the blood-brain barrier, become viable adjunct or alternative treatments? No clinical trials of such approaches yet exist (2018-dauvilliers-hepcidin-severity-rls.md; 2026-palsa-csf-bbb-endothelial-mir124.md).
• Is the blood-brain-barrier/choroid-plexus iron-transport defect a primary (genetic/developmental) lesion or an acquired/inflammatory one, and is it reversible with sustained iron repletion or does it represent a fixed trait requiring lifelong monitoring (2011-connor-brain-iron-acquisition-bbb.md)?
• What is the best peripheral marker (or marker combination — ferritin, TSAT, hepcidin, hemoglobin) for guiding brain-targeted iron therapy, given that ferritin alone is demonstrably a poor proxy for brain iron delivery (2020-connor-peripheral-iron-csf-communication.md)?

Notes in this topic

New notes written for this sweep (all tagged brain-iron-deficiency):

• 2000-earley-csf-ferritin-transferrin.md — First CSF evidence of low ferritin/high transferrin in RLS despite normal serum values (PMID 10762522).
• 2005-earley-csf-ferritin-onset-phenotype.md — CSF ferritin deficit concentrated in early-onset RLS, correlating with age of onset (PMID 16268375).
• 2004-connor-transferrin-receptor-neuromelanin.md — Autopsy nigral cells show paradoxically low transferrin receptor and IRP1 dysfunction (PMID 15136682).
• 2006-clardy-hepcidin-ferroportin-signaling.md — CSF pro-hepcidin reduced (early-onset) while tissue pro-hepcidin elevated; ferroportin localized to choroid plexus/ependyma (PMID 16759669).
• 2009-connor-dopaminergic-profile-autopsy.md — Autopsy nigra/putamen show increased tyrosine hydroxylase and decreased D2 receptor density correlating with severity (PMID 19467991).
• 2009-snyder-mitochondrial-ferritin-nigra.md — Increased mitochondrial ferritin, decreased cytosolic ferritin in RLS nigral neurons (PMID 19816198).
• 2011-connor-brain-iron-acquisition-bbb.md — Landmark study locating an iron-acquisition defect at the choroid plexus/BBB microvasculature (PMID 21398376).
• 2011-patton-hif-pathway-activation.md — Iron-deficiency-linked pseudo-hypoxic HIF pathway activation in nigra and microvessels (PMID 21985026).
• 2001-allen-mri-r2-brain-iron.md — First in-vivo MRI (R2') evidence of low nigral/putamen iron scaling with severity (PMID 11160969).
• 2005-haba-rubio-hemochromatosis-rls.md — Paradoxical focal brain iron deficiency by MRI in systemic iron-overload (haemochromatosis) patients with severe RLS (PMID 15965214).
• 2015-moon-mri-relaxometry-methods.md — Shows MRI iron findings in RLS depend heavily on relaxometry metric/ROI method used (PMID 26257527).
• 2022-beliveau-mri-metaanalysis-revisited.md — Large cohort + meta-analysis finds no pooled MRI evidence of brain iron deficiency in RLS; flags publication bias (PMID 35500370).
• 2023-bae-fcm-qsm-rct.md — RCT: IV ferric carboxymaltose improves RLS symptoms and increases brain iron (QSM/R2*) at 6 weeks (PMID 37437492).
• 2016-li-qsm-7tesla-iron.md — 7T QSM shows reduced thalamic/dentate iron in RLS; nigral iron correlates with periodic limb movements (PMID 27544840).
• 2020-wang-qsm-hemodialysis-iron.md — QSM in hemodialysis patients shows RLS-specific regional iron deficits against a background of iron accumulation (PMID 32045852).
• 2023-kim-qsm-longitudinal-iron-therapy.md — Longitudinal QSM after IV iron shows region-specific iron change correlating with symptom improvement over 24 weeks (PMID 37257418).
• 2005-schmidauer-tcs-nigral-hypoechogenicity.md — Foundational transcranial sonography study: nigral hypoechogenicity in RLS vs hyperechogenicity in Parkinson's disease (PMID 16037973).
• 2007-godau-sn-hypoechogenicity-marker.md — Defines and validates a TCS cutoff for nigral hypoechogenicity as an RLS marker (AUC 0.91) (PMID 17133515).
• 2007-godau-sonographic-brainstem-cohort.md — 188-person TCS cohort validating nigral hypoechogenicity plus raphe/red-nucleus abnormalities (PMID 18024170).
• 2019-garcia-malo-tcs-iron-therapy-predictor.md — Baseline TCS severity predicts symptom and dopaminergic-dose-reduction response to IV iron (PMID 31875533).
• 2009-xiong-meis1-expression-brain.md — RLS-risk MEIS1 haplotype reduces MEIS1 mRNA/protein in brain tissue (PMID 19126776).
• 2011-catoire-meis1-iron-homeostasis.md — Establishes a bidirectional MEIS1-iron regulatory link (worm/human-cell models) (PMID 21710629).
• 2018-dauvilliers-hepcidin-severity-rls.md — Elevated serum hepcidin despite normal ferritin; ferritin does not correlate with severity (PMID 29418021).
• 2020-connor-peripheral-iron-csf-communication.md — In-vitro BBB model: iron transport tracks hemoglobin, not ferritin — explains why ferritin under-detects brain iron deficiency (PMID 32295615).
• 2025-winkelman-aasm-guideline-rls.md — AASM 2024/2025 guideline: ferritin/TSAT thresholds for iron therapy; strong recommendation for IV FCM; against standard dopamine-agonist use (PMID 39324694).
• 2007-qu-a11-lesion-iron-deprivation-mouse.md — Mouse model: dietary iron deprivation + A11 lesion causes RLS-like hyperactivity, reversed by D2/D3 agonist (PMID 17483695).
• 2015-hyacinthe-primate-iron-depletion.md — Macaque blood-withdrawal model reproduces peripheral/CSF iron dissociation and increased striatal dopamine turnover (PMID 25662508).
• 2022-earley-bid-rodent-model-review.md — Review validating brain-iron-deficiency rodent models against human autopsy findings; flags adenosinergic/opioid targets (PMID 35779614).
• 2026-palsa-csf-bbb-endothelial-mir124.md — 2026 mechanistic study: RLS CSF impairs iron uptake in BBB endothelial cells via miR-124-3p-mediated TfR1 destabilization (PMID 41947413).

Pre-existing notes (from other topic sweeps) referenced here for the mechanistic picture:

• 2014-earley-iron-dopamine-metabolic-theory-review.md — Review synthesizing CSF/autopsy/imaging evidence for the iron-to-dopamine causal chain.
• 2016-quiroz-adenosine-a1-receptor-downregulation.md — Brain iron deficiency downregulates adenosine A1 receptors in rodents.
• 2018-ferre-pivotal-adenosine-neurotransmission.md — Proposes BID-driven A1R loss unifies the hyperdopaminergic and hyperglutamatergic models of RLS.
• 2019-ferre-adenosine-corticostriatal-terminals.md — Optogenetic confirmation that A1R loss unmasks corticostriatal glutamate hypersensitivity, reversible by dipyridamole.
• 2019-lyu-triple-opioid-receptor-knockout-mouse.md — Opioid-receptor-knockout mice develop an RLS-like anemic, circadian-hyperactive phenotype.
• 2020-lyu-mu-opioid-receptor-knockout-mouse.md — Mu-opioid-receptor knockout mice show anemia, altered iron, and circadian dopaminergic/serotonergic disruption.
• 2026-valle-leon-striatal-entopeduncular-opioid-dopamine-model.md — Brain iron deficiency sensitizes the opioid/dopamine-co-expressing striatal-GPi pathway in mice.