Why this exists

Four questions, answered from primary sources with sample sizes attached: how we know the locus coeruleus tracks attention and executive function, why it expresses tyrosine hydroxylase, what neuromelanin actually is, and how the two connect through imaging.

Every citation carries a DOI or PMID retrieved from PubMed. Where a widely repeated claim turned out to be weaker than its reputation, that is stated rather than smoothed over, and there are five such cases.

The short version

The evidence that the LC matters for attention and executive function is genuinely strong, but the strongest tier is not imaging at all. It is post-mortem neuron counts against decades of cognitive testing, and it requires no imaging assumption whatsoever.

Two of the most-repeated claims about the LC do not hold up well. Pupil diameter is a poor readout of LC activity, despite being cited constantly as a proxy. And standard-resolution fMRI of the LC should generally not be believed, for reasons of physics rather than analysis.

Neuromelanin is not brain melanin. It is a catecholamine-derived, iron-binding, protein-and-lipid pigment inside an autophagic organelle, and the amount you accumulate is set by the mismatch between how much catecholamine sits in your cytosol and how well your vesicular transporter packages it away.

---

1. How we know the LC tracks attention and executive function

Tier 1: single-unit recording in behaving monkeys, plus causal work in rodents

The foundational study recorded 47 single-cell and 126 multicell LC units in four cynomolgus monkeys performing an oddball visual discrimination task. Every LC neuron examined was phasically and selectively activated by the target cue, and not by reward, lever release, fixation spot, or non-targets. Response latency averaged 90.7 ms, roughly 200 ms before the behavioral response, and LC latency correlated with behavioral latency (r = 0.30, p < 0.0001).

The finding that makes it about attention rather than sensation: response magnitude was attenuated during epochs of poor performance, and LC responses and performance decayed together over more than 90 minutes of task.

Later work in rhesus macaques showed LC activation timing tracks the behavioral response more closely than the stimulus, and occurs before both correct and incorrect identifications but not before non-task lever movements. That is the result that reframed the LC from a salience detector into a readout of a decision process.

Adaptive gain theory is the synthesis. Phasic LC activity facilitates the ensuing response and supports exploitation; tonic activity accompanies disengagement and exploration. Anterior cingulate and orbitofrontal cortex, both utility monitors, project directly to monkey LC and are proposed to drive the mode switch.

State honestly what that theory rests on: correlational primate electrophysiology plus computational modeling, not causal manipulation.

The causal test came later, in rats. Chemogenetically elevating LC tonic activity during a patch-foraging task reduced task participation, increased response times and omissions, and made animals leave patches earlier, best explained by increased decision noise rather than a value bias. Direct causal support for the tonic-equals-disengagement half of the theory, in a rodent, at a supraphysiological manipulation.

Tier 2: post-mortem neuron counts against ante-mortem cognitive trajectories

This is the most underrated tier and the one that most directly answers whether LC integrity predicts cognition in humans.

165 participants in the Rush Memory and Aging Project completed a mean of 5.8 years of annual testing on a 19-test battery, then came to autopsy. Aminergic neuron density was estimated in locus coeruleus, dorsal raphe, substantia nigra and ventral tegmental area.

Analyzed separately, higher density in every nucleus except VTA predicted slower cognitive decline. Modeled together, only LC density remained associated (estimate 0.003, SE 0.001, p < 0.001). Higher LC density also attenuated the association between Lewy bodies and cognitive decline, which is the specific pattern expected of a reserve variable rather than a disease marker.

The specificity cuts both ways, and the same group's negative result matters. In 124 participants without dementia, LC tyrosine-hydroxylase-immunoreactive density was not associated with late-life depressive symptoms, while VTA density was (estimate -0.014, SE 0.003, p < 0.001). So the post-mortem LC signal is relatively specific to cognitive trajectory, not to mood.

A later study bridged in vivo and neuropathology at scale: 174 people with LC MRI plus tau and amyloid PET, anchored against autopsy data from 1,524 ROS/MAP cases and 2,145 National Alzheimer's Coordinating Center cases with three separate LC measures. Lower LC integrity related to entorhinal tau in unimpaired individuals and, with amyloid present, to greater cortical tau spread and retrospective memory decline.

One detail that matters for the imaging section: LC pigmentation ratings correlated with LC neuronal density but not with LC tangle density. The signal is closer to "how many pigmented neurons are left" than to "how sick they are."

Tier 3: neuromelanin-sensitive MRI and cognition in living people

The study most often cited for a clean dissociation is real, and its headline is accurate, but it needs three caveats carried with it.

Design: 3T neuromelanin sequence, contrast ratio against cerebral peduncles and pontine tegmentum. Recruited 135 Parkinson patients and 72 controls; the LC-cognition analysis ran on 61 patients and 25 controls.

Findings, all controlling for age, sex, education, and for the other nucleus:

Neither predicted memory, visuospatial function, or language.

The three caveats:

1. The dissociation did not survive in the fully overlapping subsample. Re-run in the 42 patients who completed everything, the nigra-reinforcement effect fell to p = 0.14, LC-attention to p = 0.07, LC-executive to p = 0.08. The authors say so explicitly and call the result preliminary.
2. Age was a stronger predictor of cognition than LC signal in every domain (memory beta = -0.47, p < 0.001).
3. In controls, neither nucleus predicted anything (all p > 0.14).

Surrounding literature is more reassuring. In 66 younger and 228 older adults, LC integrity was positively associated with learning and memory across multiple tasks, with age effects confined to rostral LC. In 229 older and 67 younger adults, LC contrast related to cortical thickness, prominently in frontoparietal regions.

Tiers 4 and 5: pupillometry and fMRI

Both are weaker than their reputation. See the dedicated section below.

Tier 6: pharmacology, which constrains the transmitter and not the nucleus

A clean double dissociation in healthy volunteers: atomoxetine, a selective noradrenaline reuptake inhibitor, improved stop-signal response inhibition and did nothing to probabilistic learning; citalopram impaired probabilistic learning with no effect on inhibition. A randomized double-blind crossover of single-dose atomoxetine in 26 people with cocaine use disorder and 28 controls replicated the speeding effect and found stopping improvement conditional on baseline performance.

What this proves: manipulating central noradrenergic tone changes response inhibition and speed.

What it does not prove: anything about the LC specifically. Four reasons:

• Atomoxetine blocks the norepinephrine transporter everywhere, and in prefrontal cortex that transporter also clears dopamine, so atomoxetine raises prefrontal dopamine too.
• Guanfacine and clonidine act on postsynaptic alpha-2A receptors in cortex and presynaptic autoreceptors in LC, with opposite net effects by dose, giving the classic inverted U.
• Propranolol's cognitive effects are dominated by emotional memory via amygdala, not executive function.
• Methylphenidate blocks both dopamine and norepinephrine transporters, so it cannot isolate the noradrenergic contribution at all.

Pharmacology constrains the transmitter. Only lesion, chemogenetic, optogenetic and recording work constrains the nucleus.

---

2. Why the LC is tyrosine hydroxylase expressing

The biochemistry, which makes this almost definitional

Tyrosine hydroxylase catalyzes tyrosine to L-DOPA. That is the first and rate-limiting step of every catecholamine pathway, and all three run through it:

tyrosine → (TH) → L-DOPA → (AADC) → dopamine → (DBH) → noradrenaline → (PNMT) → adrenaline

So TH is not a dopaminergic marker. It is the common ancestor of all three. Every dopaminergic, noradrenergic and adrenergic neuron expresses TH, because none can make its transmitter without it.

What makes the LC noradrenergic rather than dopaminergic is dopamine beta-hydroxylase, which unlike cytosolic TH and AADC is a vesicular enzyme. So the sequence inside an LC neuron is: make dopamine in the cytosol, load it into a vesicle by VMAT2, convert it to noradrenaline inside the vesicle.

Dopamine is therefore an obligate intermediate in noradrenaline synthesis. That single fact makes the co-release finding at the end of this section both possible and unsurprising.

A full noradrenergic identity requires TH, AADC, VMAT2, DBH, the norepinephrine transporter, and the transcription factors PHOX2A and PHOX2B, which specify the phenotype and directly transactivate the DBH promoter.

The histochemical history

The method came first. In 1962, Falck, Hillarp, Thieme and Torp showed that catecholamines and indolamines exposed to dry formaldehyde vapor condense into fluorescent derivatives, with intense fluorescence only when the amine carries hydroxyls at the 3 and 4 positions. For the first time you could see monoamines in situ rather than infer them from bulk assay.

Dahlström and Fuxe applied it systematically to rat brainstem in 1964 and published the map that still organizes the field. They assigned catecholamine cell groups the A designations running caudal to rostral, and serotonergic groups the B designations. The LC became A6, the largest and most compact noradrenergic group, supplying most of the brain's noradrenaline.

Two honest caveats. The Falck-Hillarp method does not reliably distinguish dopamine from noradrenaline in cell bodies; the noradrenergic assignment rested on fluorescence plus pharmacological manipulation plus regional biochemical assay. And the 1964 work is rat.

Modern confirmation

Protein. TH immunohistochemistry in human LC is routine and underlies the neuron counts above. The most direct link to imaging: post-mortem human brainstem scanned at 7T with 78 micrometer in-plane resolution, then stained, showed LC MRI contrast corresponding to the location of neuromelanin-containing cells, and those cells were the TH-positive noradrenaline neurons.

Transcript, in human, at single-cell and spatial resolution. The first transcriptome-wide characterization of human LC used spatially resolved transcriptomics plus single-nucleus RNA sequencing, 20,191 nuclei across three neurotypical donors after quality control. The noradrenergic cluster confirmed TH, DBH, the norepinephrine transporter and VMAT2, with 327 significantly differentially expressed genes. The authors explicitly checked dopamine marker genes to confirm the TH expression maps to noradrenaline-synthesizing rather than dopamine-synthesizing neurons, which is exactly the discrimination the biochemistry demands.

Two useful precisions from that work. VMAT2 was by far the most strongly differentially expressed of the four canonical markers, and TH the least. And noradrenergic nuclei showed an unexpectedly high mitochondrial read fraction, attributed to large fragile cell bodies and high metabolic load, which forced the authors to abandon standard quality filtering.

The nuance: what else is in there, and LC dopamine

Non-noradrenergic cells. The transcriptomic study found a distinct cluster of 186 serotonergic nuclei, but showed spatially that this population was distributed across LC and non-LC regions and was not enriched inside the annotated LC. So the dissected LC region contains serotonergic and inhibitory neurons; that study does not establish a serotonergic population inside the LC proper. As far as can be found, there is no good quantitative estimate of the fraction of human LC-proper neurons that are TH-negative. Treat any specific number with suspicion. Peptide co-expression, principally galanin and NPY, is well established and is co-transmission rather than a separate identity.

LC dopamine co-release is real, and it is one of the more interesting results of the last decade. Because dopamine is an obligate intermediate, a terminal that fails to complete the DBH step releases dopamine.

Two independent 2016 mouse studies:

The first showed LC firing is especially sensitive to environmental novelty, that LC TH-positive neurons project more profusely to hippocampus than VTA TH-positive neurons do, and that optogenetic activation of LC TH-positive neurons mimics novelty-driven memory enhancement while VTA inactivation does not abolish it. The decisive pharmacology: both the memory enhancement and the long-lasting CA1 potentiation were blocked by hippocampal D1/D5 antagonism and resistant to adrenoceptor blockade.

The second, independently, used optogenetics plus HPLC to show photostimulating LC axons increases dopamine release in dorsal hippocampus, enhancing selective attention and spatial object recognition via D1/D5 receptors. Their framing of the paradox is the clean one: dorsal hippocampus has a dense dopamine receptor network but a surprising scarcity of VTA dopamine axons.

Both are mouse. Whether human LC terminals co-release dopamine meaningfully is not established. But it is a live confound for anyone attributing a cognitive deficit to "the noradrenergic system," and careful authors flag it in their own designs.

---

3. What neuromelanin actually is

Chemistry

Calling it "brain melanin" is misleading. It is a catecholamine-derived, iron-binding, protein-and-lipid-containing pigment held inside a double-membrane autophagic organelle, and the organelle is part of the entity.

The most granular data available come from electron microscopy plus nano-SIMS plus X-ray microspectroscopy at 5 to 10 nanometer resolution on seven post-mortem human LC tissues. That work confirmed the three-compartment architecture, showed metal accumulation localized predominantly to the sulfur-rich pheomelanin-like compartment, and found sulfur and iron both trending upward with age.

How it forms, and why VMAT2 capacity is the controlling variable

Formation is auto-oxidation, not enzymatic synthesis by tyrosinase. Cytosolic dopamine or noradrenaline that escapes vesicular sequestration is oxidized, largely by iron-mediated catalysis, into membrane-impermeant quinones and semiquinones, which polymerize and become trapped with protein and lipid inside an autophagic vacuole.

The decisive experiment worked in rat substantia nigra and PC12 cells, in a species that does not normally make neuromelanin at all:

• Induced it by L-DOPA exposure, which is rapidly converted to cytosolic dopamine.
• Showed by paramagnetic resonance that the induced pigment was identical to human neuromelanin, sitting in identical double-membrane autophagic vacuoles.
• Abolished synthesis by overexpressing VMAT2, which lowers cytosolic dopamine by increasing vesicular loading.
• Inhibited synthesis with an iron chelator.

That is a clean necessary-and-sufficient design, which is why "neuromelanin formation is set by the mismatch between cytosolic catecholamine load and VMAT2 capacity" is a supported claim rather than a hypothesis.

The human correlate: in post-mortem midbrain, VMAT2 immunostaining is inversely related to neuromelanin content, and inversely related to Parkinson vulnerability, across nigrosome-1, matrix and ventral tegmental subregions. Neurons that hold the least dopamine in vesicles accumulate the most pigment and die first.

Distribution and age

Both structures are literally named for the pigment. Substantia nigra, black substance, from dopamine-derived neuromelanin. Locus coeruleus, blue spot, from the bluish appearance of the noradrenaline-derived pigment. You can see both with the naked eye in a fresh human brainstem.

Pigment appears in the first years of life, accumulates roughly linearly through adulthood, and peaks in old age. That is the direct mechanistic link between aging and selective vulnerability: the two most heavily pigmented catecholaminergic populations are the two most targeted in Parkinson disease.

Protective or toxic: genuinely contested

The protective case. Neuromelanin is an effective metal chelator, and the iron complex sequesters redox-active iron that would otherwise drive Fenton chemistry. It permanently traps reactive quinones and catechol adducts in a membrane-impermeant polymer inside an organelle. The elemental mapping data support the sequestration picture directly.

The toxic case. Neuromelanin released from dying neurons activates microglia. Human neuromelanin injected into rat substantia nigra produced microglial activation comparable to bacterial endotoxin, plus significant dopaminergic cell loss at one week. Synthetic neuromelanin analogues do the same. The proposed loop is self-propelling: neuron dies, releases pigment, activates microglia, microglia kill neurons. Adding to this, human nigral and LC catecholaminergic neurons express MHC class I, which is induced by factors released from microglia activated by neuromelanin or alpha-synuclein, and can trigger cytotoxic T-cell-mediated killing.

The strongest single experiment beat the species problem. Overexpressing human tyrosinase in rat substantia nigra produced age-dependent human-like neuromelanin inside nigral dopamine neurons, up to levels reached in elderly humans. Above a threshold of intracellular accumulation the animals developed hypokinesia, Lewy-body-like inclusions and nigrostriatal degeneration. Enhancing lysosomal proteostasis reduced pigment and prevented the degeneration.

A reading that reconciles them: neuromelanin is protective as an intracellular sink up to a load threshold, and pathogenic both above that threshold and once released extracellularly. That is what the tyrosinase data most directly support. Caveat: tyrosinase bypasses the normal auto-oxidation route entirely, so it models having the pigment rather than the process that makes it.

The species problem

Rodents do not make appreciable neuromelanin. Stated flatly in the primary literature. The consequences are serious:

• No rodent model spontaneously reproduces the age-dependent pigment accumulation that is the single largest risk factor for Parkinson disease.
• Neuromelanin-sensitive MRI cannot be developed or validated in the standard rodent workhorses.
• Every rodent LC or nigral result comes from an unpigmented nucleus, and the pigment is not a bystander, since it holds most of the neuron's iron.
• Both workarounds, L-DOPA loading and tyrosinase overexpression, induce pigment through non-physiological routes.

Non-human primates do pigment, which is one reason the monkey electrophysiology transfers better than the rodent optogenetics.

---

4. Neuromelanin-sensitive MRI

The physics, stated carefully, because the standard account may be wrong

The usual explanation: neuromelanin chelates iron, the complex is paramagnetic, paramagnetic centers shorten T1, so LC and substantia nigra appear hyperintense on T1-weighted sequences.

That account is probably incomplete, and possibly wrong about the dominant mechanism.

Work developing a magnetization transfer sequence at 3T and 7T asked directly whether the contrast arises from magnetization transfer effects or from iron-driven relaxation shortening, and concluded it relates to magnetization transfer effects. A follow-up phantom study using natural and synthetic melanins, with relaxation mapping, two-pool modeling and Bloch simulations, concluded the contrast likely results from a lower macromolecular fraction in the LC than in surrounding tissue, not primarily from paramagnetic T1 shortening. It also found T1 lengthening in the LC in older individuals.

This matters practically: widely used fast spin echo sequences are getting incidental magnetization transfer contrast from their refocusing train, which is why they work at all, and why sequences with different magnetization transfer loading are not interchangeable across sites.

Recent protocol optimization gives useful concrete numbers: 3D gradient echo magnetization transfer at 3T reaching 0.67 by 0.73 by 2 mm, contrast-to-noise 8.27, relative contrast around 14 to 17 percent, inter-rater reliability 88.5 percent, LC visible across 6 to 12 mm rostro-caudally, and measured LC diameter 1.94 mm left and 1.67 mm right.

Those dimensions are the practical reason everything here is hard. At 0.7 mm in-plane you have two or three voxels across the structure. At typical fMRI resolution you have less than one.

The signal is not a count of neuromelanin

What is measured is a contrast ratio between an LC region of interest and a reference region, sensitive to reference choice, sequence, field strength, coil, resolution, motion and segmentation method. It is a relative index, not a concentration.

Validation against histology exists and is good. Post-mortem human brainstem at 7T showed the contrast corresponds to the location of neuromelanin-containing cells, which are the noradrenaline neurons. For substantia nigra, signal intensity correlated with regional neuromelanin concentration even in the absence of neurodegeneration, which is the prerequisite for using it as a proxy for dopamine function rather than only for cell loss.

The honest boundary: these establish that the spatial location corresponds to pigmented cells, and that intensity tracks concentration regionally. They do not establish that an individual's contrast ratio quantifies that person's pigment content. And the post-mortem finding that pigmentation ratings tracked neuronal density but not tangle density says the signal is closer to how many pigmented neurons remain than to how sick they are.

What the LC signal tracks

The 7T rostro-caudal gradient work

Design: 71 people with Parkinson disease and 40 matched controls had 7T magnetization-transfer-weighted MRI. A subgroup of 30 patients and 27 controls underwent 7T fMRI across two sessions, with patients scanned on and off dopaminergic medication in counterbalanced order.

Findings:

• Reduced neuromelanin signal in caudal LC in Parkinson disease (p = 0.010).
• Caudal LC disintegration correlated with orthostatic hypotension (p = 0.009) and cognitive impairment (p = 0.036).
• Reduced caudal LC fMRI activation to arousing visual and auditory stimuli (p = 0.012).
• The blunted activation reached significance only in the on-medication state; off-medication showed a similar but non-significant trend (p = 0.105).

Two precisions on phrasing. "Blunted caudal LC activation persisting on dopaminergic medication" is accurate to the authors' interpretation, but the underlying pattern is significant on and not significant off, which is not the same as demonstrated in both states, and with 30 versus 27 subjects that difference is well within sampling noise. The cognitive correlation at p = 0.036 is a single uncorrected p value in a paper testing several structure-symptom relationships.

Note also that the gradient direction conflicts across literatures: aging work finds rostral LC carries the age effect and rostral integrity predicts memory, this work finds caudal LC carries Parkinson pathology, and the depression work found opposite-signed cognitive associations for rostral right versus caudal left. Not necessarily contradictory, since aging and Parkinson disease need not attack the same segment, but nobody should present a unified rostro-caudal story yet.

Alzheimer disease, and the challenge to Braak

Braak's claim. In 2,332 non-selected brains aged 1 to 100, only 10 were immunonegative for abnormal tau. Fifty-eight cases had subcortical tau predominantly in the locus coeruleus with no abnormal cortical tau at all. Cortical involvement appeared first in transentorhinal region. The first amyloid plaques occurred in neocortex after the onset of brainstem tauopathy. Braak and Del Tredici argued this implies neuron-to-neuron, prion-like propagation from LC to transentorhinal cortex.

The quantitative follow-up supports early LC involvement. Unbiased stereology in 48 well-characterized subjects found that at Braak stage 0, 7.9 percent of LC neurons already harbored tau inclusions, against 2.6 percent in dorsal raphe. LC counts roughly doubled from stage 0 to stage I (p = 0.02).

The serious challenge. A cellular biosensor assay quantified tau seeding activity, as opposed to phospho-tau immunostaining, across four brain regions in each of 247 individuals spanning the full tangle stage range. The finding: the earliest and most robust seeding activity was in transentorhinal and entorhinal cortex, and the LC did not uniformly show seeding until later stages. Seeding was also detected in superior temporal gyrus and primary visual cortex before tangles or immunostaining were detectable there.

How to hold both. Pretangle material genuinely appears in LC first; that is replicated and not in dispute. What is disputed is whether that material is the transmissible, self-propagating species that drives the disease, or an early but relatively inert marker of vulnerability in a cell type that is metabolically stressed for other reasons: very long unmyelinated axons, high oxidative load, catecholamine chemistry, pigment accumulation. The seeding data argue the latter.

Anyone presenting "Alzheimer's starts in the locus coeruleus" as settled is overstating it.

Harmonization, the practical ceiling on clinical use

A 42-author consensus statement lays out what would be needed to make LC imaging a biomarker and frames it as not yet there. The concrete problems:

• Sequences are not comparable. Different acquisition families produce contrast through partly different mechanisms. Absolute contrast ratios do not transfer between protocols.
• Reference region choice is unstandardized and directly scales the result.
• Segmentation was rater-dependent until recently, though automation is improving (reliability 0.91 against manual across 190 individuals, scan-rescan 0.82).
• Test-retest is decent but not clinical grade. With a heavily optimized protocol, strict head immobilization, flow compensation and arterial suppression, contrast stability varied 4 to 11 percent between scans. Against disease effects of comparable magnitude, that is not a comfortable margin for individual decisions.
• Physiological noise. The LC abuts the fourth ventricle. Cerebrospinal fluid pulsation plus cardiac and respiratory motion corrupt the region.

---

What is weaker than its reputation

1. Pupil diameter as an LC readout. This is the biggest overclaim in the field.

The claim is everywhere. Here is what was actually shown.

Recording from LC, inferior and superior colliculus, and anterior and posterior cingulate cortex in rhesus macaques while measuring pupil, with microstimulation at each site, produced two results. The field quotes only the first. Yes: LC activation reliably anticipates pupil changes, down to single spikes from single units. But this relationship is not specific to the LC. Similar relationships, with different timings and reliabilities, appear in inferior colliculus, superior colliculus, and both cingulate regions. The authors' own conclusion is that pupil changes may reflect LC-mediated coordination across a distributed network, which is weaker and more interesting than "pupil equals LC."

A mouse study with optogenetically identified LC units and simultaneous pupillometry is titled, unambiguously, that pupil diameter is not an accurate real-time readout of locus coeruleus activity. It established a graded monotonic relationship, dilation increasing with LC spike count, but with variability such that pupil predicts only a small fraction of LC activity moment to moment, and with large session-to-session fluctuations in response to identical optical stimulation.

The fair summary from the same group: pupil size reflects at least three separable systems, cortical modulation of the pretectal olivary nucleus for the light reflex, the superior colliculus for orienting and salience, and the LC for arousal. Attributing a pupil response to LC without controlling luminance, orienting and task salience is unwarranted.

Practical position: pupil is a decent index of arousal and effort, which is a real and useful construct. It is a poor index of LC spiking, and it is not anatomically specific to LC.

2. fMRI of the LC

A technical comment in the same journal showed that the responses reported as "human locus coeruleus" in two prior high-profile reports did not correspond to the anatomical location of the LC, and presented cautionary data on signal quality obtainable with standard acquisition. That comment has never been adequately answered.

The physical problem is not subtle. An LC roughly 1.7 to 1.9 mm across inside a 3 by 3 by 3 mm functional voxel means the voxel is overwhelmingly not LC. Add cerebrospinal fluid pulsation from the adjacent fourth ventricle, brainstem motion, and susceptibility at tissue-air-bone interfaces.

Modern work does better, but only with heavy machinery: high-resolution acquisition, specialized co-registration, physiological noise correction, and in the best current example, 7T with a neuromelanin-defined anatomical prior, two sessions per subject and strong arousing stimuli.

Practical position: LC fMRI at 3T with standard parameters and no neuromelanin prior should not be believed. At 7T with a neuromelanin-defined region of interest and physiological correction, it is credible but low-powered.

3. The nigra-versus-LC cognitive dissociation

Did not survive in the fully overlapping subsample of 42 patients, and the authors say so. Well designed and important, explicitly labeled preliminary, awaiting independent replication.

4. "Alzheimer's starts in the locus coeruleus"

Two strong methods disagree about whether LC pretangle tau is the disease-driving species or an early marker of vulnerability. Unresolved.

5. "Neuromelanin is protective" and "neuromelanin is toxic"

Both are presented as settled by their respective advocates. Neither is. The tyrosinase-overexpression rat is the strongest single experiment and supports a threshold model rather than either pure position, while inducing pigment through a non-physiological route in a species that normally has none.

6. Neuromelanin MRI as "measuring neuromelanin"

Post-mortem correlations are real and good, but the contrast mechanism is probably dominated by macromolecular fraction rather than paramagnetic relaxation, the signal is a protocol-dependent ratio rather than a concentration, and 4 to 11 percent between-scan variability under optimal conditions plus zero cross-site harmonization means individual-level clinical interpretation is not currently defensible.

---

Why this matters for the menopause work

Three connections back to the wider project.

The LC is TH-dependent, so it is inside the tyrosine hydroxylase story whether or not you consider it part of "the dopamine system." Anything that affects TH biology, including cofactor supply, iron status, and feedback inhibition, touches this nucleus.

It is imageable today, non-invasively, on a standard scanner, unlike dopamine synthesis capacity, which requires PET.

And the study that would matter most has not been done. No LC neuromelanin study stratified by menopause stage or hormone therapy use exists. Given that the LC tracks attention and executive function, that those are precisely the domains women describe as brain fog, and that a preprint has linked APOE genotype to LC molecular state and reduced neuromelanin content, that absence is the most obviously missing study in this whole area.

---

Key sources

Full per-question source lists, 52 peer-reviewed citations with DOIs and PMIDs, are held in the project record. The load-bearing ones:

1. Aston-Jones G, Rajkowski J, Kubiak P, Alexinsky T (1994). Locus coeruleus neurons in monkey are selectively activated by attended cues in a vigilance task. J Neurosci 14(7):4467-80. 10.1523/JNEUROSCI.14-07-04467.1994. PMID 8027789.
2. Aston-Jones G, Cohen JD (2005). An integrative theory of locus coeruleus-norepinephrine function. Annu Rev Neurosci 28:403-50. 10.1146/annurev.neuro.28.061604.135709. PMID 16022602.
3. Kane GA, Vazey EM, Wilson RC, et al. (2017). Increased locus coeruleus tonic activity causes disengagement from a patch-foraging task. Cogn Affect Behav Neurosci 17(6):1073-83. 10.3758/s13415-017-0531-y. PMID 28900892.
4. Wilson RS, Nag S, Boyle PA, et al. (2013). Neural reserve, neuronal density in the locus ceruleus, and cognitive decline. Neurology 80(13):1202-8. 10.1212/WNL.0b013e3182897103. PMID 23486878.
5. Jacobs HIL, Becker JA, Kwong K, et al. (2021). In vivo and neuropathology data support locus coeruleus integrity as indicator of Alzheimer's disease pathology and cognitive decline. Sci Transl Med 13(612):eabj2511. 10.1126/scitranslmed.abj2511. PMID 34550726.
6. Sun S, Madge V, Djordjevic J, et al. (2025). Selective effects of substantia nigra and locus coeruleus degeneration on cognition in Parkinson's disease. Mov Disord 40(5):844-54. 10.1002/mds.30148. PMID 39945211.
7. Dahl MJ, Mather M, Duzel S, et al. (2019). Rostral locus coeruleus integrity is associated with better memory performance in older adults. Nat Hum Behav 3(11):1203-14. 10.1038/s41562-019-0715-2. PMID 31501542.
8. Dahlstrom A, Fuxe K (1964). Evidence for the existence of monoamine-containing neurons in the central nervous system. Acta Physiol Scand Suppl 232:1-55. PMID 14229500.
9. Keren NI, Taheri S, Vazey EM, et al. (2015). Histologic validation of locus coeruleus MRI contrast in post-mortem tissue. Neuroimage 113:235-45. 10.1016/j.neuroimage.2015.03.020. PMID 25791783.
10. Weber LM, Divecha HR, Tran MN, et al. (2024). The gene expression landscape of the human locus coeruleus revealed by single-nucleus and spatially-resolved transcriptomics. eLife 12:e84628. 10.7554/eLife.84628. PMID 38266073.
11. Takeuchi T, Duszkiewicz AJ, Sonneborn A, et al. (2016). Locus coeruleus and dopaminergic consolidation of everyday memory. Nature 537(7620):357-62. 10.1038/nature19325. PMID 27602521.
12. Kempadoo KA, Mosharov EV, Choi SJ, Sulzer D, Kandel ER (2016). Dopamine release from the locus coeruleus to the dorsal hippocampus promotes spatial learning and memory. PNAS 113(51):14835-40. 10.1073/pnas.1616515114. PMID 27930324.
13. Sulzer D, Bogulavsky J, Larsen KE, et al. (2000). Neuromelanin biosynthesis is driven by excess cytosolic catecholamines not accumulated by synaptic vesicles. PNAS 97(22):11869-74. 10.1073/pnas.97.22.11869. PMID 11050221.
14. Liang C-L, Nelson O, Yazdani U, et al. (2004). Inverse relationship between the contents of neuromelanin pigment and the vesicular monoamine transporter-2. J Comp Neurol 473(1):97-106. 10.1002/cne.20098. PMID 15067721.
15. Zucca FA, Segura-Aguilar J, Ferrari E, et al. (2017). Interactions of iron, dopamine and neuromelanin pathways in brain aging and Parkinson's disease. Prog Neurobiol 155:96-119. 10.1016/j.pneurobio.2015.09.012. PMID 26455458.
16. Carballo-Carbajal I, Laguna A, Romero-Gimenez J, et al. (2019). Brain tyrosinase overexpression implicates age-dependent neuromelanin production in Parkinson's disease pathogenesis. Nat Commun 10:973. 10.1038/s41467-019-08858-y. PMID 30846695.
17. Zecca L, Wilms H, Geick S, et al. (2008). Human neuromelanin induces neuroinflammation and neurodegeneration in the rat substantia nigra. Acta Neuropathol 116(1):47-55. 10.1007/s00401-008-0361-7. PMID 18343932.
18. Priovoulos N, van Boxel SCJ, Jacobs HIL, et al. (2020). Unraveling the contributions to the neuromelanin-MRI contrast. Brain Struct Funct 225(9):2757-74. 10.1007/s00429-020-02153-z. PMID 33090274.
19. Cassidy CM, Zucca FA, Girgis RR, et al. (2019). Neuromelanin-sensitive MRI as a noninvasive proxy measure of dopamine function in the human brain. PNAS 116(11):5108-17. 10.1073/pnas.1807983116. PMID 30796187.
20. Lund AE, Madelung CF, Madsen KH, et al. (2026). Blunted response of caudal locus coeruleus to arousing stimuli in Parkinson's disease. Brain. 10.1093/brain/awag106. PMID 41873142.
21. Braak H, Thal DR, Ghebremedhin E, Del Tredici K (2011). Stages of the pathologic process in Alzheimer disease: age categories from 1 to 100 years. J Neuropathol Exp Neurol 70(11):960-9. 10.1097/NEN.0b013e318232a379. PMID 22002422.
22. Ehrenberg AJ, Nguy AK, Theofilas P, et al. (2017). Quantifying the accretion of hyperphosphorylated tau in the locus coeruleus and dorsal raphe nucleus. Neuropathol Appl Neurobiol 43(5):393-408. 10.1111/nan.12387. PMID 28117917.
23. Kaufman SK, Del Tredici K, Thomas TL, Braak H, Diamond MI (2018). Tau seeding activity begins in the transentorhinal/entorhinal regions and anticipates phospho-tau pathology in Alzheimer's disease and PART. Acta Neuropathol 136(1):57-67. 10.1007/s00401-018-1855-6. PMID 29752551.
24. Betts MJ, Kirilina E, Otaduy MCG, et al. (2019). Locus coeruleus imaging as a biomarker for noradrenergic dysfunction in neurodegenerative diseases. Brain 142(9):2558-71. 10.1093/brain/awz193. PMID 31327002.
25. Joshi S, Li Y, Kalwani RM, Gold JI (2016). Relationships between pupil diameter and neuronal activity in the locus coeruleus, colliculi, and cingulate cortex. Neuron 89(1):221-34. 10.1016/j.neuron.2015.11.028. PMID 26711118.
26. Megemont M, McBurney-Lin J, Yang H (2022). Pupil diameter is not an accurate real-time readout of locus coeruleus activity. eLife 11:e70510. 10.7554/eLife.70510. PMID 35107419.
27. Astafiev SV, Snyder AZ, Shulman GL, Corbetta M (2010). Comment on two reports of human locus coeruleus fMRI. Science 328(5976):309. 10.1126/science.1177948. PMID 20395497.