Bishop & Robinson 2004 — Aβ-metal complexes are paradoxically neurotoxic and neuroprotective in rat cerebral cortex

Bishop and Robinson (School of Psychology, Psychiatry and Psychological Medicine, Monash University, Australia) co-injected human Aβ1-42 with iron, copper, or zinc into rat cerebral cortex at the concentrations these metals are reported to reach in Alzheimer-disease senile plaques (Fe 1 mM, Cu 0.4 mM, Zn 1 mM; Lovell et al. 1998 cited in [28]), and counted Fluoro-Jade-labelled dying neurones 24 hours later. The experiment was designed to reconcile a recurring conflict in the in vitro literature in which Aβ-metal complexes appear sometimes more neurotoxic than free metal and sometimes less. The headline result is paradoxical at the in vivo level too: Aβ-iron and Aβ-copper complexes killed substantially fewer cortical neurones than iron or copper injected alone, while Aβ-zinc trended (non-significantly) toward greater neurotoxicity than zinc alone. The authors interpret the findings as evidence that Aβ contributes to antioxidant defence of the brain by binding redox-active metal ions, and that the neurotoxicity reported for amyloid plaques in vivo may be attributable to the iron and zinc components of the plaques rather than to Aβ or copper.

Why this matters

• It is one of the few in vivo (rat intracortical injection) characterisations of Aβ-metal complex neurotoxicity at the metal concentrations actually reported in senile plaques, rather than the higher concentrations typical of in vitro neuronal-culture work.
• It is mechanistic background for the broader heavy-metal-peptide binding literature collected in the Black Market Peptide Metal Survey folder — specifically, it documents that peptide-metal coordination can reduce the free-metal toxicity of redox-active transition metals (Fe, Cu) in the mammalian brain, which is the same chemistry that motivates the use of peptide-based chelation as a remediation modality discussed in luo2024-peptides-heavy-metal-remediation.
• It clarifies that the Aβ + Zn case is mechanistically distinct from the Aβ + Fe and Aβ + Cu cases — zinc neurotoxicity does not arise from redox chemistry, so chelation by a redox-inert peptide does not attenuate it.
• It quantifies that copper acetate alone, at the 0.4-mM plaque concentration, produces a smaller absolute increase in neuronal death than iron at 1 mM (Figure 2: copper ~140 % of saline vs iron ~260 %), and that Aβ co-injection essentially eliminates the copper-only excess — bringing neuronal loss down to a level similar to Aβ alone (Aβ+copper 77 ± 13 % of saline vs Aβ-alone 79 ± 17 % of saline per ref [6]; copper-alone vs Aβ+copper t[13]=2.576, p=0.023). This is relevant context for any downstream synthesis on the relative toxicity of redox-active versus redox-modulated metal ions in the central nervous system.

Methods (brief)

Animals. Female Wistar rats, n=56, aged 12-13 weeks, sourced from Monash University Animal Services. Standard rat pellets and water ad libitum; 12-hour dark cycle. Protocols approved by the Monash University Psychology Animal Ethics Committee under National Health & Medical Research Council guidelines.

Substances and stocks. Substances were dissolved in sterile saline (0.9 % NaCl, pH 4.5). A 1.0-mM stock of human Aβ1-42 (US Peptide, Calif., Lot #12) was prepared and stored at 4 °C for at least 24 hours. Iron stock: 1.0-mM ferric ammonium citrate, pH 5.2 (ICN). Copper stock: 0.4-mM cupric acetate, pH 5.4 (Sigma). Zinc stock: 1.0-mM zinc acetate, pH 5.8 (Sigma). For Aβ + metal co-injection, a 2.0-mM Aβ stock was prepared, vortexed at 2400 rpm for 10 s, then mixed in equal parts with a 2x-concentrated metal solution and vortexed at 2400 rpm for 30 s; the mixture was stored 18-22 hours at 4 °C, then re-vortexed at 2400 rpm for 3 s immediately before injection. The Aβ preparation yields high-fibrillar Aβ whether prepared alone or with metals (per refs [5] and [6]).

Injection protocol. Rats were anaesthetised with 3 % halothane in carbogen. Each rat received three 1.0-µL injections of sterile saline in one hemisphere and three injections of the experimental substance in the contralateral hemisphere. At 24 hours, animals were overdosed with sodium pentobarbitone (200 mg/kg) and transcardially perfused with 4 % paraformaldehyde in phosphate buffer (PB; 0.1 M, pH 7.2). Brains were post-fixed in paraformaldehyde for 4 hours, cut into 100-µm coronal sections on a vibratome, immunolabelled for Aβ by free-floating immunocytochemistry using the 6F/3D clone (Dako), and dying neurones stained with Fluoro-Jade.

Quantitation. Injection sites stained with Fluoro-Jade were analysed under FITC fluorescence; bleeding or mechanically damaged sites were excluded. Every Fluoro-Jade-labelled neurone in each injection site was counted. To limit injection-tract-depth variability, sites not 1.3 ± 0.5 mm deep were excluded. Dying-neurone counts per millimetre of injection tract were averaged across multiple analysable sites within each animal, and the mean for each experimental substance was expressed as a percentage of the mean for contralateral saline sites (% Saline = [N_E(experimental) / N_E(saline)] × 100). Saline = 100 %; a substance with the same toxicity as saline = 100 %; one that kills twice as many neurones = 200 %; one that kills half as many = 50 %.

Statistics. Box-plot outlier exclusion left n=7 in the copper-only group and n=8 in all other groups (the copper experimental group lost one animal). Each experimental group was compared with its contralateral saline using a 2-tailed paired-samples t-test (α=0.05). Between-group comparisons used 2-tailed independent-samples t-tests (α=0.05).

The Aβ-alone, iron-alone, and Aβ + iron data (“*Data originally presented in [6]” — Bishop & Robinson 2003) are re-shown in Table 1 and Figure 2 for direct visual comparison with the new copper and zinc groups; the present paper’s original data are the copper-alone, Aβ + copper, zinc-alone, and Aβ + zinc groups.

Key numbers

Neuronal death by group (mean ± SE, expressed as % of contralateral saline):

Metal-ion concentrations and Aβ concentration injected (Table 1, p. 449):

(*Aβ-alone, iron-alone, and Aβ + iron data originally presented in ref [6] — Bishop & Robinson 2003.)

Plaque concentrations of Fe, Cu, Zn cited from Lovell et al. (1998, ref [28]): Fe 1 mM, Cu 0.4 mM, Zn 1 mM. These are the concentrations the present study uses for the injected stocks.

Total dying neurones counted: more than 103 000 neurones from a total of 56 brains (8 animals per group except copper-only with 7).

Aβ immunoreactivity in injection sites: Aβ injected alone (Figure 1A, 1D), with iron (per ref [6]), or with zinc (data not shown in this paper) produced large fibrillar Aβ deposits at the injection site. Aβ injected with copper (Figure 1B, 1E) also produced Aβ deposits. Copper injected alone (Figure 1C, 1F) produced no Aβ deposit but a diffuse light immunolabelling for Aβ peptide in the neuropil, consistent with copper not creating new Aβ.

Findings

1. Aβ-iron and Aβ-copper complexes are less neurotoxic than the free metals

Co-injection of Aβ1-42 with iron reduced iron neurotoxicity from ~260 % of saline to ~150 % of saline (p<0.05; data from ref [6] re-shown in Figure 2). Co-injection with copper brought copper neurotoxicity down from ~140 % of saline (Figure 2 visual estimate of copper-alone) to 77 ± 13 % of saline — a value the authors describe as “a similar low number of neurones” to the 79 ± 17 % of saline reported for pure Aβ in their prior work (ref [6]). The copper-alone vs Aβ+copper comparison is statistically significant: t[13]=2.576, p=0.023. The text reports no separate copper-alone-vs-saline test statistic; the copper-alone effect is described qualitatively as “killed more neurones than saline.” The reductions in iron and copper neurotoxicity by Aβ co-injection are both statistically significant.

The authors’ proposed mechanism is that Aβ chelates redox-active Fe(III) and Cu(II), reducing the rate of Fenton-type reactions that generate hydroxyl radicals, lipid peroxidation, and consequent neuronal injury. They note that copper catalyses the Fenton reaction “at an order of magnitude faster than iron” (per Halliwell & Gutteridge 1999, ref [18]), which is why even the much lower 0.4-mM copper concentration causes any neuronal death at all.

2. Aβ does not protect against zinc neurotoxicity

Zinc alone (1.0 mM) killed 126 ± 12 % MORE neurones than saline (i.e., neuronal loss = 226 ± 12 % of saline; t[7]=3.012, p=0.020). Co-injection of Aβ + zinc produced neuronal loss of 159 ± 36 % of saline, evaluated against three reference groups: vs saline t[7]=1.534, p=0.169 (NS); vs zinc-alone t[8.4]=0.864, p=0.412 (NS); vs pure Aβ t[14]=1.998, p=0.066 (NS, trend toward more neuronal death than pure Aβ). The Aβ + zinc condition therefore did not significantly differ from any of the three reference groups, but the trend was toward greater neurotoxicity than pure Aβ rather than toward reduction of zinc-alone neurotoxicity. The authors emphasise the t[14]=1.998 result is a non-significant trend, but interpret it as consistent with reports that high zinc concentrations enhance Aβ-induced neuronal death in vitro (refs [29], [31]).

Mechanistically, zinc is redox-inert, so Fenton-type chelation chemistry does not apply. The authors hypothesise alternative pathways: zinc inhibition of glutathione peroxidase (ref [42]), decreased Na+/K+-ATPase activity (ref [29]), or enhanced kainate neurotoxicity (ref [41]) — none of which are blocked by complexation with a non-redox peptide.

3. Aβ alone is essentially non-toxic at the plaque concentration

Injections of pure fibrillar Aβ1-42 did not kill significantly more neurones than saline (per ref [6]; consistent with the wider in vivo literature, refs [5], [6], [16], [19]). The authors note that this is in contrast to the in vitro literature, where Aβ is reported to be neurotoxic to cultured neurones (refs [34], [38]), and they attribute the discrepancy to the absence of astrocytic glutathione-peroxidase defence in monocultured neurones (refs [11], [14], [26]).

4. The “amyloid paradox” interpretation

The authors conclude that Aβ in vivo may have a normal antioxidant function — binding free Fe, Cu, and possibly other redox-active metals in the interstitial fluid and limiting their toxicity. In the AD brain, plaques may be neurotoxic because of the iron and zinc bound within them, not because of the Aβ peptide itself, and the overall toxicity of a plaque is still less than the toxicity that the free redox-active metals would produce in an unbound state. The “amyloid paradox” is that Aβ-metal complexes are both neurotoxic (relative to Aβ alone) and neuroprotective (relative to free metal alone) at the same time.

Implications

• Mitigation / chelation evidence. This paper is a clean in vivo demonstration that a peptide can substantially reduce the neurotoxicity of redox-active transition metals at biologically realistic concentrations, by chelation chemistry rather than by enzymatic activity. It supports the mechanistic narrative on remediation-evidence that peptide chelation is a legitimate modality for reducing metal toxicity in mammalian systems, even though the practical remediation literature focuses on environmental rather than neurological targets.
• Certification / HMTc threshold setting. No direct relevance. The metals studied (Fe, Cu, Zn) are not in the HMTc 10-analyte vocabulary, and the paper measures no contamination in any food, beverage, or personal-care matrix. The paper contributes nothing to threshold-setting work for Pb, Cd, iAs, tAs, MeHg, tHg, Ni, Al, Cr-VI, or Sn.
• App. No routing to product or ingredient pages. The paper does not measure occurrence in any consumed product.
• Courses. Useful as a textbook example in a neurochemistry or biochemistry-of-metal-toxicity module: it illustrates that the same peptide-metal complex can be more or less toxic than its parts depending on whether the metal is redox-active, and that the in vitro/in vivo discrepancy for Aβ is plausibly explained by the absence of astrocytic antioxidant defence in monoculture systems. It is also useful as background for discussing why “chelation reduces toxicity” is a useful first approximation but not universally true (the Zn case).
• Microbiome. No relevance. The paper does not engage gut microbiota or any microbiome axis.

Limitations

• This is an acute (24-hour) injection model, not a chronic or developmental exposure model. The authors discuss this implicitly when they note that AD-relevant plaque biology develops over years, while the present study examines neuronal injury at one acute time point.
• Sample size is modest (n=7-8 per group). Several of the key comparisons (Aβ + zinc vs zinc alone at p=0.066, Aβ vs saline NS) are borderline, and the study is not powered to detect small effects.
• Aβ stock preparation is identified as fibrillar (yield reported per the authors’ prior protocol in refs [5], [6]) but the present paper does not quantify the proportion of fibrillar vs oligomeric Aβ in each preparation, nor does it report whether the metal complexation affected the fibril morphology. This limits mechanistic interpretation of why Aβ + Cu produces deposits visually similar to Aβ alone but Aβ + Zn (data not shown in this paper) produces a different morphology that the text does not characterise.
• Metal concentrations are chosen to match the (single) Lovell et al. 1998 ref [28] estimate of plaque metal content. Plaque metal content estimates have wide uncertainty in the literature; the chosen concentrations are reasonable but not the only defensible choice, and the conclusions about “in vivo realism” depend on whether the 1-mM Fe, 0.4-mM Cu, 1-mM Zn estimate generalises to all AD plaques.
• The Aβ-alone, iron-alone, and Aβ + iron data are re-shown from the authors’ prior paper (ref [6], Bishop & Robinson 2003), not independently re-collected. The internal consistency of Figure 2 across the present and prior paper is therefore a matter of the original measurements, not an independent replication. The new data are the copper-alone, Aβ + copper, zinc-alone, and Aβ + zinc groups.
• Free zinc dose response is not characterised; only one zinc concentration (1.0 mM) was used. The narrative on Aβ + zinc being “non-protective” cannot be extrapolated to lower zinc concentrations where redox-independent toxicity may be subthreshold.

Wiki pages this source may touch

• copper
• iron
• zinc
• remediation-evidence

Verification notes

Existing-page check. DOI grep (10.1111/j.1750-3639.2004.tb00089), raw_handle grep (MFK_24-amyloid-metal-complexes-can-be-neurotoxic-and-n), and cite-key glob (bishop2004-*) over wiki/sources/ on 2026-06-08 returned no matches. This is a NEW source page — no prior version to merge-enhance.

Evidence tier. A (primary peer-reviewed in vivo experimental study). The paper reports original animal-injection data with explicit sample sizes, t-statistics, and p-values. A-tier reservation in the wiki is for primary peer-reviewed studies and authoritative agency monographs; this is the former.

Metals frontmatter. Empty. The paper studies Fe, Cu, Zn at AD-plaque-relevant concentrations. None of these are in the HMTc 10-analyte vocabulary (Pb, tAs, Cd, MeHg, tHg, iAs, Ni, Al, Cr-VI, Sn). Per the convention established on luo2024-peptides-heavy-metal-remediation and shalev2022-peptide-metal-nmr-review, only HMTc-priority analytes are recorded in metals: frontmatter; out-of-scope metals are routed via the ## Wiki pages this source may touch section using the page-slug form [[metals/copper]], [[metals/iron]], [[metals/zinc]].

Ingredients, products, matrices, jurisdictions frontmatter. All empty. The source is an in vivo neurochemistry study using rat cortical injections; it measures nothing in any food, beverage, personal-care, or environmental matrix. No jurisdiction applies (the work was done in Australia at Monash University, but the experimental scope is not regulatory).

Sample size. 56 rats total, 7-8 per group after box-plot outlier exclusion. Recorded as sample_n: 56 (total animals) with population descriptor in sample_population:.

Brand firewall (Part 12). No consumer brands appear in the source. Scientific-method vendor names are present and retained per the 2026-05-17 Exception 2 locked rule: ferric ammonium citrate (ICN), cupric acetate and zinc acetate (Sigma), Aβ1-42 (US Peptide, Calif., Lot #12), 6F/3D anti-Aβ clone (Dako). These are reagent vendors associated with analytical method, not contamination values; they remain in the Methods section as written.

HMTc firewall (Part 2). The paper contains no HMTc-threshold language, no “consistent with the consensus that…” synthesis claims, and no consumer-audience risk advisories. The Implications section above is bounded to documenting what the paper contributes; no threshold values are proposed.

Reviewer’s note on scope fit. This paper is in scope per the 2026-06-02 commit 3f47f95 — scope: mitigation/remediation is in-scope, not a skip. It is a peptide-metal mechanism paper from the Black Market Peptide Metal Survey folder (Kimi Agent collection on heavy-metal peptide binding), aligned with the routing of the other peptide-metal entries already ingested (luo2024, shalev2022, marques2025, seregin2023, grill1989, spallacci2025, zhang2025, wang2022, sears2013). Like those entries, it does not advance any HMTc-analyte threshold but does support the wiki’s mechanistic coverage of peptide-metal chelation in mammalian systems.

Date arithmetic. Published in Brain Pathology 2004, Volume 14, Issue 4, pp. 448-452 (October 2004 issue per Crossref). DOI 10.1111/j.1750-3639.2004.tb00089.x. License is unknown — Brain Pathology was a Wiley journal in 2004 with a subscription model; no open-access designation appears on the PDF, hence license: unknown rather than CC-BY-*.

Statistical reporting in the paper. The text gives mean ± SE (standard error), not SD. The figure caption for Figure 2 states “neuronal loss (mean ± SE)“. This is preserved in the Key numbers table above.

Copper-group sample size correction. The text states “Eight animals were injected with each experimental substance. Box-plot analysis was used to find extreme outliers, which were excluded from further statistical analysis. This resulted in the copper experimental group having n=7; all other groups had n=8.” Recorded accurately in the Methods and Key numbers sections.

Audit subagent (2026-06-08) corrections applied — verdict QUARANTINE downgraded to REVISE-applied. A fresh-context audit subagent flagged six ❌ numerical-fidelity errors in the original draft of this page; all were independently verified against PDF p. 450 and corrected in this revision:

• t-statistic for copper-only vs Aβ+copper had wrong df. Original draft wrote t[7]=2.576, p=0.023; the PDF reads t[13] = 2.576, p = 0.023. Corrected to t[13]. (df=13 matches independent-samples test of n=7 + n=8 − 2 = 13.)
• Attribution of the copper t-test was wrong. Original draft framed t[13]=2.576, p=0.023 as a copper-vs-saline test (“significantly above saline”). The PDF text reads: “Injections of copper acetate killed more neurones than saline, and this additional neuronal loss was completely eliminated when copper was co-injected with Aβ (t[13] = 2.576, p = 0.023).” The parenthetical attaches to the elimination claim — the test is copper-only vs Aβ+copper, not copper vs saline. The original copper-vs-saline relationship is reported qualitatively only (“killed more neurones than saline”) with no explicit test statistic; this is now noted in the table.
• “79 ± 17 %” was misattributed to copper-alone. The PDF p. 450 explicitly says “Previously we reported that the number of dying neurones around injection sites of pure Aβ was lower (79 ± 17%) than the number around saline injection sites (6).” 79 ± 17 % is the Aβ-alone value from the authors’ 2003 paper (ref [6]), not the copper-alone value. The copper-alone row now correctly shows no explicit numeric mean (the paper gives none for copper-only), and the Aβ-alone row carries the 79 ± 17 % value from ref [6].
• “77 ± 13 %” for Aβ + copper was misread as “above saline”. The PDF reads: “co-injection of copper with Aβ killed a similar low number of neurones (77 ± 13%)” — the framing is “similar low number” to the 79 ± 17 % Aβ-alone value below saline. So 77 ± 13 % is “of saline” (below saline at ~77 %), not “above saline.” Original draft’s parenthetical “(i.e., ~110 % of saline as shown in Figure 2)” was wrong; corrected to “77 ± 13 % of saline” matching the source.
• “126 ± 12 %” for zinc-alone was misread as “of saline” rather than “MORE than saline”. The PDF reads: “Injections of zinc killed 126 ± 12 % more neurones than injections of saline.” So zinc-alone neuronal loss = 226 ± 12 % of saline (100 % baseline + 126 % more). Original draft’s parenthetical “(i.e., ~130 % of saline)” was wrong; corrected to “226 ± 12 % of saline” with the calculation shown.
• Aβ + zinc t-test attributions were swapped. Original draft wrote t[8.4]=0.864, p=0.412 vs saline (NS) and t[14]=1.998, p=0.066 vs zinc-alone (NS trend). The PDF p. 450 paragraph reads: “(t[8.4] = 0.864, p = 0.412)” attached to “where zinc was co-injected with Aβ was higher than when zinc was injected alone” → vs zinc-alone, not vs saline. The “(t[14] = 1.998, p = 0.066)” is attached to “Co-injection of zinc with Aβ also caused more neuronal death than injections of pure Aβ” → vs pure Aβ, not vs zinc-alone. The vs-saline test is “(t[7] = 1.534, p = 0.169)” — a third statistic the original draft omitted entirely. All three Aβ+Zn t-tests are now correctly attributed in the Key numbers table and in Findings §2.
• “Why this matters” bullet 4 referenced copper-alone as “~40 % above saline” with a “statistically marginal” framing — both derived from the misattributed 79 ± 17 % value. Rewritten to use the Figure 2 visual estimate (~140 % of saline) and the corrected copper-vs-Aβ+copper t-test framing.
• ⚠️ The audit also flagged [[mitigation/remediation-evidence]] as outside the taxonomy snapshot’s four-category vocabulary. Verified that wiki/mitigation/remediation-evidence.md exists; the mitigation/ subtree is a legitimate routing destination per the 2026-06-02 commit 3f47f95 — scope: mitigation/remediation is in-scope, not a skip and is the same routing slug used on luo2024-peptides-heavy-metal-remediation. No correction needed; the slug is valid wiki content outside the snapshot’s four-category drafting vocabulary.

The audit subagent recommended quarantining the page and rewriting Key numbers from scratch. Instead, the corrections were applied in place — the underlying paper-faithful re-reads were straightforward once the misattributions were identified, and the rest of the page (frontmatter, methods, slug routing, firewall posture, limitations) all passed audit clean. Page is downgraded from QUARANTINE to REVISE-applied.

Page history

The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.