Regional Brain Iron, Copper, and Zinc in Alcohol Use Disorder: Sex-Specific Differences in Human Post-Mortem Tissue

Abstract

Background.

One of the risks associated with alcohol use disorder (AUD) is dementia. While the exact pathology is unknown, there is increasing evidence that alcohol consumption may dysregulate iron homeostasis in the brain. Unregulated iron can produce oxidative stress and cellular damage to neurons, including the iron-dependent form of cell death termed ferroptosis. Previous human work has relied on magnetic resonance imaging and quantitative susceptibility mapping, which estimate iron indirectly. We therefore measured iron directly, together with copper and zinc, in three brain regions of donors with and without a diagnosis of AUD.

Methods.

Hippocampus, caudate, and substantia nigra were obtained from the New South Wales Brain Tissue Resource Centre. Tissues came from 40 donors — 20 males and 20 females, half of each with a diagnosis of AUD. Samples were wet-ashed and analysed by total reflection X-ray fluorescence, which quantifies ten elements simultaneously. Data were analysed by two-way analysis of covariance (diagnosis, sex; age as covariate), with simple main effects tested as model contrasts. Because concentrations are expressed per unit wet mass, the seven elements unrelated to the study hypothesis were used as internal reference analytes to screen for common scaling effects.

Results.

In the caudate, iron was elevated in donors with AUD (p = 0.003), an effect confined to women (+47.2 μg g⁻¹, p = 0.0002; men +5.4 μg g⁻¹, p = 0.62). Caudate copper was elevated in both sexes (p = 0.016). In the hippocampus, iron was modestly lower in AUD (p = 0.032). In the substantia nigra, a diagnosis-by-sex interaction was observed for iron and zinc; however, the same interaction was present across the reference elements, and adjustment for tissue composition attenuated the iron interaction (p = 0.091) and abolished that for zinc (p = 0.331). Donor groups were well matched for age, post-mortem interval, brain pH, and RNA integrity, but cirrhosis was present only among donors with AUD, and the caudate copper effect did not persist when these donors were excluded.

Conclusions.

These data provide direct evidence of regional trace-metal differences associated with a diagnosis of AUD, most robustly an elevation of caudate iron in women. The nigral findings should be interpreted with caution, and the copper finding may reflect hepatic disease rather than alcohol exposure per se. Sex differences in metal handling in the striatum warrant replication in a larger, independently characterised cohort.

1. INTRODUCTION

Among the trace metals that have biological functions in eukaryotes, iron is arguably the most important of the transition metals. It plays essential roles in cellular respiration, enzyme activity, DNA synthesis, and haemoglobin function. In the brain, iron is essential for neurotransmitter production and function and for myelin production, and consequently for cognition, affect, and motor activity. Brain iron concentration is tightly constrained, and several proteins control its accumulation, removal, and action. Too little iron in the brain during development can produce cognitive deficits that are permanent (Lozoff et al., 2013), and in adulthood iron deficiency is associated with restless legs syndrome (Earley et al., 2014). Conversely, brain iron concentrations above the homeostatic range can cause damage through the production of free radicals and oxidative stress. Nearly all neurodegenerative diseases, including Parkinson’s disease, share iron dysregulation with locally elevated iron concentrations (Kulaszyńska et al., 2024; Ahern et al., 2025). We and others have shown that exposure to toxins and toxicants can dysregulate brain iron (Torres-Rojas et al., 2020; Xiu et al., 2025), and we have shown that 16 weeks of ethanol consumption in female mice increased hippocampal iron in all 12 BXD recombinant inbred strains examined, accompanied by strain-dependent reductions in hippocampal mass (Jones et al., 2021). Listabarth and colleagues (2020) have proposed that heavy alcohol consumption in humans could dysregulate iron and thereby contribute to dementia.

Three human studies have examined this question using magnetic resonance imaging with quantitative susceptibility mapping, an indirect estimate of tissue iron (Juhás et al., 2017; Topiwala et al., 2022; Adams et al., 2023). To our knowledge, no study has reported directly measured elemental concentrations in defined brain regions of donors with a diagnosis of AUD. We therefore obtained hippocampus, caudate, and substantia nigra from the New South Wales Brain Tissue Resource Centre (Sutherland et al., 2016) and quantified iron, copper, and zinc.

Our primary, a priori hypothesis concerned iron. Because the analytical method quantifies several elements simultaneously, copper and zinc were also measured; analyses of these two metals are reported as exploratory. The remaining seven elements returned by the assay were not of biological interest here and are used only as internal reference analytes (Section 2.4).

2. MATERIALS AND METHODS

2.1. Donors and tissue

Tissue samples of hippocampus, substantia nigra, and caudate were obtained from frozen brains held by the New South Wales Brain Tissue Resource Centre (NSW BTRC), University of Sydney. Donors comprised 20 women and 20 men, half of each with a diagnosis of AUD established by the Centre’s standard clinical review procedures (Sutherland et al., 2016), and half with no recorded alcohol-related problems. Approximately 25 mg of each region was dissected and shipped frozen to the Memphis laboratory.

Donor characteristics are given in Table 1. Age, post-mortem interval, brain pH (measured in fresh-frozen cerebellum), and RNA integrity number were compared between diagnostic groups. Cause of death, liver pathology at autopsy, and smoking history were recorded categorically.

Table 1.

Donor characteristics. Values are mean (SD) or n.

	Female AUD	Female Control	Male AUD	Male Control	AUD vs Control
n	10	10	10	10	—
Age (y)	57.0 (14.2)	56.9 (15.3)	60.8 (14.7)	61.4 (14.9)	p = 0.96
Post-mortem interval (h)	32.2 (13.7)	32.0 (7.7)	28.7 (8.5)	33.6 (12.8)	p = 0.50
Brain pH	6.5 (0.3)	6.6 (0.4)	6.6 (0.4)	6.6 (0.3)	p = 0.66
RNA integrity number	7.2 (1.2)	7.0 (1.2)	7.4 (0.7)	7.5 (0.7)	p = 0.86
Cirrhosis at autopsy (n)	4	0	4	0	—
Current smoker (n)	6	2	6	3	p = 0.012
Age at onset of alcohol use (y)†	21.3 (4.8)	—	23.2 (11.4)	—	—
Duration of heavy alcohol use (y)†	35.9 (14.9)	—	34.2 (8.9)	—	—
Estimated alcohol intake (g/day)†	353 (443)	—	253 (181)	—	—
Abstinence before death (y)†‡	0.45 (0.76)	—	3.44 (10.0)	—	—

^†Alcohol consumption history is reported for AUD donors only; these variables were not systematically recorded for control donors. Values were obtained from next-of-kin interview and medical records held by the NSW Brain Tissue Resource Centre at the time of donation and should be regarded as approximations.

^‡One male donor with AUD had been abstinent for 32 years before death (AUD in sustained remission); excluding this donor, mean abstinence in male AUD donors was 0.04 years (SD 0.13). The median abstinence across all AUD donors was 0 years; 15 of 20 AUD donors had no recorded period of abstinence before death.

2.2. Tissue preparation and analysis

Tissues were wet-ashed in 400 μl ultra-pure nitric acid and digested for 48 h at room temperature. Samples were diluted with 400 μl distilled water, with gallium added as internal standard at a final concentration of 1 μg ml⁻¹. Ten μl of each sample was pipetted onto a quartz sample carrier and dried on a hotplate. Technical duplicates were read on a Bruker S2 PicoFox total reflection X-ray fluorescence spectrometer (Bruker Instruments, Billerica, MA) with sub-ppb sensitivity, each for 500 s. Concentrations are expressed as μg g⁻¹ wet tissue mass.

Elemental values were not returned for seven of the 120 samples (five caudate, one substantia nigra, one hippocampus. These samples were excluded listwise from analyses of the affected elements, giving analysed sample sizes of 39 (hippocampus), 39 (substantia nigra), and 35 (caudate).

2.3. Statistical analysis

Each region was analysed separately. For each metal, a two-way analysis of covariance was fitted with diagnosis and sex as between-subjects factors and age as a covariate, using Type II sums of squares. Where a diagnosis-by-sex interaction was detected, simple main effects of diagnosis within each sex were tested as contrasts on the fitted model, using the pooled error term and retaining the age adjustment; these were adjusted for multiplicity across the two contrasts by the Holm procedure. Effect sizes are reported as partial η² for factorial effects and as adjusted mean differences with 95% confidence intervals for simple effects. Exact p values are reported throughout.

Iron was the a priori hypothesis and constitutes the confirmatory family; within that family of six tests, p values were controlled at a false discovery rate of 5% by the Benjamini–Hochberg procedure, and all four nominally significant iron effects retained significance (adjusted p = 0.008–0.048). Copper and zinc are reported as exploratory and are presented without correction, so labelled.

Analyses were performed in Python 3 using statsmodels 0.14.

2.4. Screening for common scaling effects

Because concentrations are expressed per unit wet mass, variation in the mass of the analysed aliquot, in tissue water content, or in the proportion of adjacent structures captured during dissection scales all elemental concentrations by a common factor. To evaluate this, the seven elements unrelated to the study hypothesis (P, S, Cl, K, Ca, Br, Rb) were used as internal reference analytes. A composition index was defined as the mean of log-transformed phosphorus and sulfur concentrations; these two elements were selected because both are structurally bound — in phospholipid and nucleic acid, and in protein sulfur amino acids, respectively — and are stable post mortem, whereas K, Cl, and Rb are free intracellular ions subject to redistribution after death. Each metal was re-analysed with the composition index as an additional covariate, and results are reported both unadjusted and composition-adjusted.

2.5. Sensitivity analyses

Two potential confounders were identified from Table 1 and examined directly. Cirrhosis was present in eight donors, all with AUD, and therefore cannot be adjusted for; analyses were repeated with these donors excluded. Smoking status differed between groups and was entered as an additional covariate in a repeated analysis.

3. RESULTS

3.1. Donor characteristics

Donor characteristics are given in Table 1. The diagnostic groups were closely matched for age (58.9 vs 59.2 y, p = 0.96), post-mortem interval (30.5 vs 32.8 h, p = 0.50), brain pH (6.53 vs 6.57, p = 0.66), and RNA integrity number (7.30 vs 7.25, p = 0.86). Two variables differed. Current or former smoking was more frequent among donors with AUD (p = 0.012). Cirrhosis was recorded at autopsy in eight donors with AUD (four women, four men) and in no control donor; control donors more often showed steatosis. Cause of death differed in composition between groups, as expected: cardiac causes predominated among controls (11 of 20), whereas toxicity, hepatic, infective, and respiratory causes predominated among donors with AUD.

3.2. Hippocampus

Iron was lower in donors with AUD than in controls (F(1,34) = 4.98, p = 0.032, partial η² = 0.13), by 9.6% in women and 20.7% in men. Neither the main effect of sex nor the diagnosis-by-sex interaction was significant (both p > 0.37), so the effect is best described as a difference present in both sexes; inspection of simple effects indicates it was numerically larger in men (−6.2 μg g⁻¹, p = 0.035, Holm-adjusted p = 0.069) than in women (−2.7 μg g⁻¹, p = 0.34). No effects were detected for copper or zinc (all p > 0.18). Group values are given in Table 2 and individual donor values in Figure 1.

Table 2.

Metal concentrations (μg g⁻¹ wet tissue) in hippocampus. Mean ± SEM (n).

	Female AUD	Female Control	Male AUD	Male Control
Fe	24.85 ± 2.04 (10)	27.51 ± 1.92 (10)	24.35 ± 1.62 (10)	30.71 ± 2.20 (9)
Cu	2.00 ± 0.25 (10)	2.23 ± 0.23 (10)	2.51 ± 0.34 (10)	2.42 ± 0.20 (9)
Zn	11.94 ± 1.64 (10)	10.24 ± 0.75 (10)	11.59 ± 1.02 (10)	13.73 ± 1.88 (9)

Figure 1.

Iron, copper, and zinc in hippocampus, substantia nigra, and caudate of donors with and without alcohol use disorder. Each symbol is one donor; open symbols denote donors with cirrhosis at death (n = 8, all in the AUD group). Horizontal bars are group means, whiskers the standard error of the mean. Asterisks denote simple main effects of diagnosis within sex, tested as contrasts on the fitted analysis of covariance model with age as covariate and a pooled error term: *p < 0.05, **p < 0.01, ***p < 0.001; n.s., not significant. Analyzed n = 8–10 per cell (hippocampus N = 39, substantia nigra N = 39, caudate N = 35).

3.3. Substantia nigra

Neither main effect was significant for any metal. A diagnosis-by-sex interaction was observed for iron (F(1,34) = 10.62, p = 0.003, partial η² = 0.24) and for zinc (F(1,34) = 6.20, p = 0.018, partial η² = 0.15). Simple effects showed that nigral iron was higher in women with AUD than in control women (+48.0 μg g⁻¹, 95% CI 15.9 to 80.2, p = 0.005, Holm-adjusted p = 0.009). The corresponding decrease in men did not reach significance (−23.9 μg g⁻¹, 95% CI −55.2 to 7.3, p = 0.13). For zinc, the decrease in men was of borderline significance (−2.0 μg g⁻¹, p = 0.045) and did not survive adjustment (Holm-adjusted p = 0.090); the increase in women was not significant (p = 0.16). These findings are qualified by the analysis in Section 3.5. Group values are given in Table 3.

Table 3.

Metal concentrations (μg g⁻¹ wet tissue) in substantia nigra. Mean ± SEM (n).

	Female AUD	Female Control	Male AUD	Male Control
Fe	131.02 ± 6.01 (9)	83.16 ± 6.88 (10)	104.58 ± 13.61 (10)	128.47 ± 13.59 (10)
Cu	9.73 ± 0.79 (9)	9.82 ± 1.01 (10)	8.17 ± 0.68 (10)	9.45 ± 0.81 (10)
Zn	12.00 ± 0.54 (9)	10.47 ± 0.67 (10)	10.24 ± 0.77 (10)	12.26 ± 0.74 (10)

3.4. Caudate

Iron was elevated in donors with AUD (F(1,30) = 10.60, p = 0.003, partial η² = 0.26), with a diagnosis-by-sex interaction (F(1,30) = 6.96, p = 0.013, partial η² = 0.19). Simple effects localised this entirely to women: caudate iron was 47.2 μg g⁻¹ higher in women with AUD than in control women (95% CI 24.0 to 70.5, p = 0.0002, Holm-adjusted p = 0.0005), a difference of approximately 80%, whereas the difference in men was negligible (+5.4 μg g⁻¹, p = 0.62). Copper was elevated in donors with AUD as a main effect, in the same direction in both sexes and without interaction (F(1,30) = 6.49, p = 0.016, partial η² = 0.18; interaction p = 0.69). No effect was detected for zinc. Group values are given in Table 4, and the full inferential results for all three regions are given in Table 5.

Table 4.

Metal concentrations (μg g⁻¹ wet tissue) in caudate. Mean ± SEM (n).

	Female AUD	Female Control	Male AUD	Male Control
Fe	100.75 ± 7.96 (9)	55.41 ± 6.15 (8)	68.95 ± 7.79 (9)	62.67 ± 9.08 (9)
Cu	4.77 ± 0.55 (9)	3.49 ± 0.47 (8)	3.81 ± 0.62 (9)	2.64 ± 0.42 (9)
Zn	11.81 ± 0.88 (9)	11.05 ± 0.68 (8)	11.00 ± 0.98 (9)	12.06 ± 0.72 (9)

Table 5.

Analysis of covariance and simple main effects. Diagnosis and sex as between-subjects factors, age as covariate, Type II sums of squares. Simple effects are adjusted mean differences (AUD − Control) with 95% confidence intervals, tested as model contrasts using the pooled error term.

Region	Metal	Effect	F	df	p	partial η2
Hippocampus	Fe	Diagnosis	4.98	1, 34	0.032	0.128
		Sex	0.24	1, 34	0.631	0.007
		Diagnosis × Sex	0.81	1, 34	0.374	0.023
Substantia nigra	Fe	Diagnosis	0.99	1, 34	0.326	0.028
		Sex	0.87	1, 34	0.358	0.025
		Diagnosis × Sex	10.62	1, 34	0.003	0.238
Substantia nigra	Zn	Diagnosis × Sex	6.20	1, 34	0.018	0.154
Caudate	Fe	Diagnosis	10.60	1, 30	0.003	0.261
		Sex	2.05	1, 30	0.162	0.064
		Diagnosis × Sex	6.96	1, 30	0.013	0.188
Caudate	Cu	Diagnosis	6.49	1, 30	0.016	0.178
		Diagnosis × Sex	0.16	1, 30	0.693	0.005
Region	Metal	Simple effect	Difference	95% CI	p	Holm p
Hippocampus	Fe	Female	−2.66	−8.20 to 2.89	0.337	0.337
Hippocampus	Fe	Male	−6.19	−11.90 to −0.48	0.035	0.069
Substantia nigra	Fe	Female	+48.01	15.85 to 80.17	0.005	0.009
Substantia nigra	Fe	Male	−23.92	−55.15 to 7.30	0.129	0.129
Substantia nigra	Zn	Female	+ 1.43	−0.58 to 3.44	0.156	0.156
Substantia nigra	Zn	Male	−2.00	−3.95 to −0.05	0.045	0.090
Caudate	Fe	Female	+47.24	24.04 to 70.45	0.0002	0.0005
Caudate	Fe	Male	+5.44	−16.92 to 27.79	0.623	0.623
Caudate	Cu	Female	+ 1.48	0.00 to 2.96	0.049	0.099
Caudate	Cu	Male	+1.08	−0.34 to 2.50	0.132	0.132

Non-significant results for metals and effects not listed are given in the Supplementary Statistics Table.

3.5. Screening for common scaling effects

In the substantia nigra, the diagnosis-by-sex interaction was not confined to the metals of interest. The same interaction was present for phosphorus (p = 0.029), sulfur (p = 0.044), potassium (p = 0.026), and rubidium (p = 0.045), with calcium approaching significance (p = 0.064). Across all ten elements, women with AUD showed values approximately 10–17% above their controls and AUD men approximately 14–20% below theirs, a pattern characteristic of a common scaling factor rather than an element-specific effect (Supplementary Table 2). The composition index itself showed a significant diagnosis-by-sex interaction in the substantia nigra (F(1,34) = 4.30, p = 0.046).

After adjustment for the composition index, the nigral iron interaction was attenuated to a non-significant trend (p = 0.091) and the zinc interaction was abolished (p = 0.331). By contrast, the composition index showed no effect of diagnosis in either hippocampus (p = 0.94) or caudate (p = 0.73), no reference element showed an effect of diagnosis in the caudate, and the principal findings in those two regions were unchanged or strengthened by adjustment: hippocampal iron p = 0.032 to p = 0.019, caudate iron p = 0.003 to p = 0.022. Caudate copper was attenuated to a trend (p = 0.076).

3.6. Sensitivity analyses

Excluding the eight donors with cirrhosis left the hippocampal iron effect (p = 0.033), the caudate iron main effect (p = 0.044), and the nigral iron interaction (p = 0.010) essentially unchanged. The caudate copper effect did not persist (p = 0.019 to p = 0.217, n = 27). Adding smoking as a covariate did not materially alter any finding (hippocampal iron p = 0.050; caudate iron p = 0.012; caudate copper p = 0.009; nigral iron interaction p = 0.008).

4. DISCUSSION

To the best of our knowledge, this is the first study to report directly measured concentrations of iron, copper, and zinc in defined brain regions of donors with a diagnosis of AUD. Three previous studies estimated brain iron indirectly using magnetic resonance imaging with quantitative susceptibility mapping. Juhás and colleagues (2017) reported AUD-related increases in susceptibility in caudate, putamen/globus pallidus, and dentate nucleus. Topiwala and colleagues (2022), studying UK Biobank participants reporting varying alcohol intake rather than a clinical diagnosis, reported alcohol-related increases in serum ferritin and liver iron and, in the brain, decreased susceptibility in thalamus but increases in substantia nigra and putamen. Adams and colleagues (2023) reported higher serum ferritin in AUD and, regionally, increased susceptibility in the left globus pallidus; whole-brain susceptibility was associated with older age and with male sex but not with AUD.

Our direct measurements agree with this body of work in identifying the striatum as susceptible to AUD-related iron elevation, and with Topiwala and colleagues in implicating the substantia nigra. They differ in one important respect: the striatal effect we observe is confined to women. This is not simply a matter of statistical power — the male estimate is close to zero (+5.4 μg g⁻¹) rather than merely imprecise — and it is the clearest sex-specific finding in the study.

4.1. Why might the caudate be affected, and why in women?

The nigrostriatal system is of particular interest. Goutaudier and colleagues (2023) showed in rats that alcohol consumption reduced dopamine activity in anterior dorsolateral striatum, dorsomedial striatum, and nucleus accumbens, accompanied by increased alcohol consumption, suggesting a feed-forward relationship between nigrostriatal hypodopaminergia and alcohol intake. Iron and dopamine can combine to produce toxic redox species, and this is pronounced in the dopamine-rich nigrostriatal system (Hare and Double, 2016). An elevation of caudate iron of the magnitude observed here is therefore mechanistically plausible.

Women are at greater risk than men for a range of alcohol-related disorders, including hepatic, cardiovascular, pancreatic, neurological, and psychiatric conditions (White, 2020). Two pharmacokinetic factors contribute: alcohol distributes into total body water, which is proportionally smaller in women, and women produce less gastric and hepatic alcohol dehydrogenase (Chrostek et al., 2003), so that equivalent intake produces higher and more sustained blood alcohol concentrations. Whether this is sufficient to account for a female-specific striatal iron elevation is unknown. An alternative and testable possibility is that it reflects differences in systemic iron status: pre- and peri-menopausal iron loss followed by post-menopausal accumulation gives women a different lifetime iron trajectory from men, and the donors here span the menopausal transition. We did not have access to peripheral iron indices, and this remains speculative.

4.2. The hippocampal finding runs opposite to the striatal one and to our mouse data

Hippocampal iron was lower in donors with AUD, by 9.6% in women and 20.7% in men. This is the opposite direction to the striatal result reported here, to the imaging literature, and to our own previous work in mice. The discrepancy deserves to be stated precisely rather than attributed to species difference.

Two features of the human result argue that it is not a technical artefact. First, it is element-specific: phosphorus and sulfur, the structural reference elements, were unchanged in the hippocampus — ratios of 0.99 in both sexes — so the reduction in iron is not a consequence of altered tissue composition or of the wet-mass denominator (Section 3.5). Second, it survived both exclusion of donors with cirrhosis and adjustment for smoking (Section 3.6).

The comparison with our mouse data is therefore a real one, and it is unfavorable to a simple account. In that study (Jones et al., 2021), female mice from 12 BXD recombinant inbred strains underwent 16 weeks of ethanol consumption by the drinking-in-the-dark method; hippocampal iron increased in every strain, by 3 to 68%, while copper and zinc decreased by up to 42% and 35% respectively. Because those animals were female, the closest available comparison is with the women in the present sample, who show a non-significant 9.6% decrease. The significant strain-by-treatment interaction in the mouse study establishes that the magnitude of the response is genotype-dependent, and comparable heterogeneity is to be expected in an outbred human population; but the murine effect was uniform in sign across all twelve strains, and heterogeneity of magnitude cannot generate a reversal of direction. We tested the specific prediction that heterogeneous response would inflate the variance of hippocampal iron among donors with AUD, and it did not (SD ratio 0.89, Levene p = 0.61).

We therefore consider two substantive possibilities.

The first concerns what was measured. The mouse study reported increased iron concentration and, in the same animals, strain-dependent reduction in hippocampal mass. Concentration and content are distinct quantities, and a reduction in mass with conserved total iron raises concentration mechanically: a 20% loss of mass with unchanged content yields a 25% rise in concentration, within the range reported. Whether hippocampal iron content as well as concentration increased in those animals has not been determined, and we regard that as the first analysis required to interpret the present discrepancy. If content did not rise, the two studies may describe a common process — loss of hippocampal tissue — expressed through denominators that moved in opposite directions, since the mouse hippocampus was dissected and weighed whole whereas the human sample was a small aliquot of a structure whose volume was not measured.

The second concerns exposure. Sixteen weeks of binge-pattern consumption in young adult mice maintained on standard chow and killed in good health isolates alcohol from its usual human companions. The present donors had decades of exposure, a mean age of 57 to 61 years, cirrhosis in eight of twenty AUD cases, frequent tobacco use, probable nutritional compromise, and death from disease. We suggest, as a hypothesis for test rather than as an interpretation of these data, that the cerebral iron response to alcohol may be non-monotonic in exposure: accumulation at moderate or intermediate exposure, as in our mice and in the moderate drinkers studied by Topiwala and colleagues (2022), and depletion in advanced disease, where malabsorption and undernutrition dominate. Such a model would also accommodate the regional dissociation reported here, if striatal iron is held in long-lived ferritin and neuromelanin pools that resist depletion while hippocampal iron more closely tracks current systemic supply.

We do not regard the direction of the hippocampal effect as explained. We identify it as a priority for replication with paired measurement of regional volume and of peripheral iron status.

4.3. The nigral findings require caution

The apparent diagnosis-by-sex interaction for nigral iron and zinc was accompanied by the same interaction across elements that have no relationship to alcohol exposure, including phosphorus, sulfur, potassium, and rubidium. This pattern is what a common scaling factor produces, and adjustment for tissue composition attenuated the iron interaction to a trend and removed that for zinc. The most likely source is dissection: the substantia nigra is small and irregularly bounded, and the proportion of adjacent midbrain captured in a 25 mg sample is difficult to control. We report the nigral results because the female iron elevation is substantial and because the region is of clear theoretical interest, but we do not regard them as established, and we would caution against building a mechanistic account on them until they are replicated with quantified dissection.

We note that this analysis was possible only because the assay returns several elements simultaneously. The same screen applied to the caudate found nothing, which is the reason we regard the caudate result as secure.

4.4. Copper and zinc

Copper is an essential micronutrient for brain development and function, participating in mitochondrial activity, catecholamine synthesis, and myelination (An et al., 2022). Like iron it exists in two oxidation states, and copper dysregulation is associated with impaired cognition, neurodegenerative disease, and psychiatric disorder (Gou et al., 2023). We observed elevated caudate copper in donors with AUD of both sexes. However, this effect did not persist when donors with cirrhosis were excluded. Secondary copper accumulation is well recognised in chronic liver disease, and because cirrhosis occurred exclusively among our AUD donors, we cannot distinguish an effect of alcohol exposure from an effect of hepatic disease. We report the finding but attribute it provisionally to liver pathology.

Zinc participates in neuroimmune signalling, is a structural component of numerous transcription factors, and acts as an antioxidant; both deficiency and overload are associated with cognitive impairment and neurodegeneration (Skalny et al., 2018; Granzotto et al., 2020; Li et al., 2022). AUD has been shown to interfere with intestinal zinc absorption (Pavuluri et al., 2022). We found no robust regional zinc difference: the nigral interaction did not survive the compositional screen, and no effect was detected in hippocampus or caudate.

4.5. Limitations

Several limitations qualify these findings.

The study is cross-sectional and observational. It identifies associations between a diagnosis of AUD and regional metal concentrations; it cannot establish that alcohol exposure caused them, and we have deliberately avoided causal language in reporting the results.

The sample is modest. Ten donors per cell provides reasonable power for main effects but limited power for interactions, which require substantially larger samples for equivalent sensitivity. The interaction effects reported here should be treated as requiring replication, and the confidence intervals in Table 5 convey the imprecision directly.

Donors were of mean age 57–61 years and represent established, late-stage disease. The design cannot describe how metal dysregulation develops over the course of AUD.

Confounding cannot be excluded. Although the groups were well matched on post-mortem variables, cirrhosis was present only in the AUD group and cause of death differed systematically. We have reported sensitivity analyses for the confounders we could measure, but nutritional status, thiamine status, and lifetime exposure were not available to us.

Finally, concentrations are expressed per unit wet mass. The compositional screen described above addresses this directly for the substantia nigra, but dry-weight normalisation would be preferable in future work.

5. CONCLUSION

We report directly measured iron, copper, and zinc concentrations in three brain regions of donors with and without a diagnosis of AUD. The most robust finding is a substantial elevation of caudate iron in women with AUD, which survives adjustment for tissue composition, exclusion of donors with cirrhosis, and adjustment for smoking. Hippocampal iron was modestly reduced. Nigral findings were confounded by a compositional difference between groups and should be regarded as provisional, and the caudate copper elevation may reflect hepatic disease. These results support continued investigation of trace-metal dysregulation in AUD, and specifically of sex differences in striatal iron handling, in larger and independently characterised cohorts.

Supplementary Material

1

Supplementary Table S1. Analysis of covariance, all metals and effects

3

Supplementary Table S2. Whole-panel elemental screen for a common scaling factor

DATA AVAILABILITY STATEMENT

All data are available from the corresponding author on request.