Song et al. 2024 — HPLC-ICP-MS method for Cr(VI) in foods and reduction mechanism

A fast HPLC-ICP-MS method was developed for Cr(VI) speciation in food matrices, achieving separation in 1.5 minutes using a weak anion-exchange column (Sepax Proteomix WAX-NP5) with 70 mM NH4NO3 mobile phase at pH 7.0. The limit of detection was 0.1 µg/kg, substantially better than most prior published HPLC-ICP-MS methods. When applied to seven commercially purchased food products from a Nanjing supermarket (milk powder, rice flour, whole wheat bread, yoghurt, white vinegar, orange juice, green tea), Cr(VI) was not detected in any sample. Spiking experiments revealed that Cr(VI) is unstable in food matrices and rapidly converts to Cr(III), with conversion accelerated by vitamin C, tea polyphenols, acidic conditions (pH 3.0), and elevated temperatures (80 °C). The authors conclude that natural reducing substances in foods likely prevent the occurrence of Cr(VI).

Key numbers

Method performance: LOD = 0.1 µg/kg Cr(VI) (3σ S/N). Analysis time 1.5 min per sample, versus 5–11 min for prior published HPLC-ICP-MS methods (Table 2 comparison). Linear range 0–50 µg/kg with calibration y = 587.32x − 24.39, R² = 0.9996. Precision RSD = 1.20% on five replicate measurements of a 5 µg/kg Cr(VI) standard.

Cr(VI) in the seven tested foods: not detected (< 0.1 µg/kg) in milk powder, rice flour, whole wheat bread, yoghurt, white vinegar, orange juice, and green tea (Figure 3 chromatograms; each food n = 3 triplicate measurements).

Recovery rates across three spiking approaches (spike levels 5, 25, 50 µg/kg; Table 3):

Cr(VI) reduction time-to-completion in spiked food components (1 g component per 1 L ultrapure water; 0.01 g for vitamin C; Cr(VI) spiked at 200 µg/kg for vitamin C and 25 µg/kg for the other five; Figure 5):

Reduction-rate acceleration at 80 °C versus 25 °C (Figure 6): vitamin C 1.93×, tea polyphenols 1.04×, whey proteins 3.92×, gelatin 2.07×, fructose 10.11×, cellulose 1.70×.

Reduction-rate acceleration at pH 3.0 versus pH 7.0 (proteins/sugars/cellulose only; vitamin C and tea polyphenols already complete at neutral pH): whey proteins 5.12×, gelatin 4.47×, fructose 6.81×, cellulose 4.5×.

Methods (brief)

Speciation analysis by HPLC-ICP-MS: Agilent 1260 HPLC system coupled to an Agilent 7700 ICP-MS via PEEK tubing into the ICP nebulizer. Separation column: Sepax Proteomix WAX-NP5 weak anion-exchange (50 × 4.6 mm, 5 µm). Mobile phase 70 mM NH4NO3 adjusted to pH 7.0, isocratic at 0.8 mL/min, 5 µL injection, ambient column temperature. ICP-MS at 1550 W RF, He carrier and compensatory gas at 1.0 L/min each, time-resolved peak-hopping acquisition monitoring m/z 52 (Cr). Cr(VI) was identified by retention time; Cr(III) chelated to EDTA served as the reference Cr(III) peak.

Sample extraction was modified from US EPA Method 3060A: 0.5 g sample combined with 0.5 mL phosphate buffer (0.5 M K2HPO4 + 0.5 M KH2PO4), 2.5 mL alkaline solution (0.5 M NaOH + 0.28 M Na2CO3), and 0.4 g MgCl2 in a 50 mL polyethylene tube; ultrasonic extraction 20 min in a KQ-400DE ultrasonic bath; centrifuged at 5120 × g for 5 min in an H1850 centrifuge; supernatant filtered through a 0.22 µm mixed cellulose ester membrane before injection. Calibration with Cr(III) (1000 mg/L) and Cr(VI) (100 mg/L) standards from the National Institute of Metrology (Beijing); food-grade reagents from Aladdin (Shanghai); analytical-grade salts from Macklin (Shanghai). LOD calculated as 3σ/(S/N). Triplicate runs per sample (n = 3) with mean ± SD; one-factor ANOVA + Tukey’s HSD via SPSS Statistics 25; chromatograms in Origin 2018. Published open access in MDPI Toxics under CC BY 4.0.

The paper does not report a certified reference material for Cr(VI) accuracy verification; method accuracy was assessed by spike-recovery only.

Implications

Certification (HMT&C): The paper’s main contribution to threshold work is methodological. It demonstrates that Cr(VI) is rapidly reduced to Cr(III) by reducing agents present in most foods — vitamin C, tea polyphenols, proteins, gelatin, sugars, cellulose — with timescales ranging from 20 minutes (vitamin C) to 60 hours (cellulose, fructose), accelerated by heat and acidity. For HMT&C threshold work where Cr-VI is a target analyte (one of the 10 official HMT&C analytes per CLAUDE.md Part 14), this paper supports the view that protocols permitting prolonged contact time between sampling and extraction will systematically underestimate any Cr(VI) that was originally present, and that the fast (1.5 min) anion-exchange extraction used here represents a methodologically more defensible upper-bound. The negative result (Cr(VI) ND across all seven foods, LOD 0.1 µg/kg) is, in the authors’ own framing (introduction §1 citing Novotnik et al. [9] and Vacchina et al. [10]; conclusion §4), consistent with the position that Cr(VI) is generally absent from plant-based and animal-derived foods at typical detection limits.

Courses: Excellent teaching case for the distinction between total Cr and Cr-VI: foods can contain detectable total Cr while having effectively zero Cr-VI because the food matrix reduces Cr(VI) to Cr(III). Vitamin C and tea polyphenols are quantitatively the most potent reducing agents identified; protein and gelatin matrices reduce more slowly due to steric hindrance and less favorable electron-donating geometry; cellulose and fructose reduce slowest. Heat and acidic conditions during food processing further accelerate the Cr(VI) → Cr(III) conversion, particularly for protein, sugar, and cellulose matrices.

App: Reinforces the wiki rule that Cr-VI must be distinguished from total Cr in occurrence reporting. In this single seven-food study, Cr(VI) was not detected at LOD 0.1 µg/kg, and the spike-recovery experiments showed effectively complete loss of Cr(VI) within 3 h of contact for orange juice and green tea (and partial loss for the other five matrices). The paper’s authors conclude (§4) that Cr(VI) is “unlikely to be present in foods”; the wiki and app should mirror that single-paper finding without extrapolating beyond it, while continuing to treat total Cr as a separately quantifiable analyte.

Wiki pages this source may touch

• chromium
• chromium-hexavalent
• tea
• rice-flour
• orange-juice
• yogurt
• vinegar
• whole-wheat-bread
• milk-and-dairy

Verification notes

• 2026-05-20 merge-enhance pass against the full PDF (pages 1–13, including Tables 1–3 and Figures 1–6): corrected raw_handle: manual-fetch-kimi (a parent-folder placeholder) → MFK_development-of-a-fast-method-using-hplc-icp-ms-and; fixed truncated raw_path to include the full PDF filename “Reduction Mechanism of Cr(VI) in Foods.pdf”; expanded metals: [Cr] → [Cr-VI, Cr] since the paper’s primary analyte is Cr(VI) speciation and Cr(III) is discussed as the reduction product; rewrote ingredients: removing invented [[ingredients/milk]] (no such ingredient page; substituted [[ingredients/milk-and-dairy]]) and substituting [[ingredients/whole-wheat-bread]] for the overspecified [[ingredients/wheat]] (the paper tested whole wheat bread, not raw wheat); added the missing tested ingredients (yogurt, orange-juice, vinegar); rewrote products: removing invented slugs (milk-powder, yoghurt, fruit-juice — none exist in the current taxonomy) and the wrongly-applied plant-milks-non-soy-non-rice (the paper’s “milk powder” is cow-milk-derived, not plant milk); kept the canonical product slugs that match the tested foods (vinegar, fruit-juices-non-apple for orange juice, tea for green tea, bread-and-baked-goods for whole wheat bread); rewrote matrices: using the canonical vocabulary plus a few necessary extensions; renamed legacy section heading ## Wiki pages updated on ingest → ## Wiki pages this source may touch per Part 5b (the model does not maintain routing by hand); removed the stray [[testing/arsenic-speciation]] cross-reference (an As-speciation page is not analogous to a Cr-VI method paper for routing purposes); removed the speculative “Microbiome” subsection from Implications (the paper does not address gut metabolism directly); expanded Methods to include the EPA 3060A reference, full sample-prep stoichiometry, ICP-MS gas flows, instrument/centrifuge model numbers, and statistical software details; noted absence of certified reference material for Cr(VI) accuracy verification.
• Numerical-fidelity spot-checks against source: LOD 0.1 µg/kg confirmed (page 6, Section 3.2); analysis time 1.5 min confirmed (abstract, conclusions, Table 2); linear range 0–50 µg/kg and calibration equation y = 587.32x − 24.39 with R² = 0.9996 confirmed (page 6); recovery range 91.70–111.85% confirmed for spiking approach 1 across Table 3; orange juice and green tea ND in approach 3 confirmed (Table 3 cells for 5, 25, 50 µg/kg spike levels — all “not detected”); food-component reduction times confirmed against Figure 5 caption text on page 9; 80 °C and pH 3 acceleration factors confirmed against page 10 narrative; “Cr(III) … 100 to 300 times less toxic than Cr(VI)” interpretive framing not imported (paper page 1 actually states “Cr(VI) is 100 to 300 times more toxic than Cr(III)”) and that toxicology ratio is not surfaced in this page because it is upstream of any HMT&C use.
• Part 12 (brand firewall): the paper reports food samples generically by category (“milk powder”, “rice flour”, etc.) purchased from “a local supermarket (Nanjing, China)” without naming brands. No brand attribution attached to contamination values. Methods name instrument and reagent vendors (Agilent, Sepax, Milli-Q, Macklin, Aladdin, KQ-400DE / Kunshan Ultrasonic Instrument, H1850 / Xiangyi, Jinteng filter membranes, IBM SPSS, OriginLab) — all permitted under Exception 2 (scientific-method vendor names support reproducibility). No firewall issue.
• Part 2 (wiki/HMT&C firewall): the Implications “Certification” subsection describes the paper’s methodological contribution to Cr-VI threshold work without proposing a numerical threshold; no HMT&C threshold values are stated. The wiki page reports what this single paper found and does not synthesize with prior literature beyond the paper’s own framing.
• 2026-05-20 fresh-context audit subagent verdict: REVISE. Three ⚠️ findings, no ❌ findings. Numerical fidelity confirmed clean (Check 1) — 80 °C and pH-3 acceleration factor pairings independently verified against page 10 of the source. Findings applied: (Check 2 matrices) the canonical system-prompt.md matrices vocabulary contains rice-flour, cereal-grain, cow-milk, and tea; the three extensions in this page are yoghurt, fruit-juice, and vinegar, explicitly noted here as proposed additions to that vocabulary because the paper’s tested matrices include yoghurt, an orange-juice product, and a vinegar product that do not have closer canonical equivalents. (Check 5b Certification) the “consistent with prior literature” sentence was tightened to attribute the framing to the paper’s own introduction (citing Novotnik [9] and Vacchina [10]) and conclusion (§4), so the wiki is paraphrasing the paper rather than performing cross-source synthesis. (Check 5c App) the App subsection generalization (“most processed foods… typical storage conditions”) was hedged to a single-paper finding referencing the seven foods, the LOD, the 3 h orange-juice / green-tea loss, and the paper’s own conclusion language. Checks 3 and 4 were ✅ clean.

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.