First principles modelling of the ion binding capacity of finger millet

Mineral deficiencies (iron, calcium, zinc) remain widespread despite fortification efforts, in part due to limited bioavailability arising from interactions with intrinsic food components such as dietary fibre, polyphenols, phytic acid and proteins. Millets are nutrient-dense, climate-resilient staples in Asia and Africa and are targeted for biofortification; finger millet is notably rich in calcium and also contains substantial potassium and zinc. However, high levels of antinutritional factors and complex dietary fibres (notably arabinoxylans) raise questions about mineral binding and bioavailability. While phytate has been extensively studied, the specific chemistry of mineral binding to millet arabinoxylans is unclear, and only limited in vitro and human data exist for bioavailability/retention. This study addresses the research question: which structural motifs in finger millet arabinoxylan bind K⁺, Ca²⁺ and Zn²⁺, and what is the relative thermodynamic stability of such complexes? Using first-principles DFT, the authors model ion binding to proposed arabinoxylan structures to identify key binding pockets and evaluate the roles of charge density versus steric compatibility.

Prior work shows millets contain abundant dietary fibres including arabinoxylans, hemicelluloses and glucans, with bran often highly enriched in arabinose and xylose. Finger millet arabinoxylan structure has been characterized by NMR and FTIR as a β-D-(1,4)-xylan backbone with predominantly mono-substituted arabinose and uronic (glucuronic) acid side chains; pearl millet arabinoxylans contain ~6% uronic acids. Limited studies on mineral binding in millets indicated variable affinities, with calcium most likely to bind to bulk dietary fibre. In vitro assessments showed processing (e.g., fermentation, germination) can increase calcium bioavailability, though not universally. Dietary fibres and tannins reduce in vitro iron and zinc solubility in pearl millet by chelation. Human studies on calcium retention from finger millet are few (n=8 in each), showing ~20–26% retention, while meta-analyses suggest moderate in vitro bioavailability. DFT studies on other polysaccharides (alginate, chitosan, pectin) show cation binding via hydroxyl and carboxylate groups. However, ion–arabinoxylan interactions in millets had not been elucidated, motivating the present computational investigation.

Three simplified arabinoxylan models, based on Subba Rao & Muralikrishna’s proposed finger millet structure, were constructed and geometry-optimized: PolyXA (β-(1→4) xylan backbone with one arabinose at C3), PolyXGA (xylan with C3 arabinose and C2 glucuronic acid), and PolyXGG (xylan with two C2 glucuronic acids). Models were built in CrystalMaker and optimized using CASTEP plane-wave DFT. Key computational settings: GGA-PBE exchange-correlation, ultrasoft pseudopotentials, kinetic energy cutoff 900 eV, Monkhorst–Pack k-point grid 1×1×1, simulation box 35×20×15 Å, and convergence thresholds of 1×10⁻⁵ eV atom⁻¹ (energy), 0.001 Å (force), and 0.03 eV Å⁻¹ (displacement). Deprotonation states were chosen to achieve charge-saturated systems: glucuronic acid carboxyl groups deprotonated; selected hydroxyls deprotonated when required for charge balance (noting that OH deprotonation is unlikely in planta or in the GI tract, whereas glucuronic acids with pKa 2.93 are expected to be deprotonated). Mulliken population analysis classified bonds as ionic (<0.40 |e|) or covalent (>0.40 |e|). Chemical potentials (µ) for Ca, K, Zn were computed from their stable crystalline phases (Ca: FCC; K: BCC; Zn: HCP) and H from H₂. Formation energies were computed as: Ef = Efinal − (Einitial + µA·x − y µH), where x is the cation charge and y the number of hydrogens removed for charge balance. Binding geometries, coordination numbers, bond lengths, and populations were analyzed for K⁺, Ca²⁺, and Zn²⁺ with each model; backbone-only interactions were also evaluated (Supplementary).

• Zn²⁺ binding: Zn²⁺–arabinoxylan complexes were thermodynamically unfavorable across all models: PolyXA Ef = +4.85 eV; PolyXGA Ef = +0.62 eV; PolyXGG Ef = +4.82 eV. In PolyXA and PolyXGG, Zn²⁺ did not bind; in PolyXGA, Zn²⁺ formed some bonds (including a covalent Zn–O4, Mulliken population ~0.49 |e|) but the complex remained unstable (positive Ef).
• Ca²⁺ and K⁺ binding stability: Both formed stable complexes, with stability enhanced by glucuronic acid residues. Formation energies: • PolyXA: K⁺ −0.19 eV; Ca²⁺ −1.06 eV. • PolyXGA: K⁺ −1.49 eV; Ca²⁺ −2.96 eV. • PolyXGG: K⁺ −3.82 eV; Ca²⁺ −2.99 eV. The PolyXGG motif (adjacent glucuronates) provided the most stable binding for both K⁺ and Ca²⁺.
• Binding pockets and sites: Carboxylate (RCOO⁻), carbonyl (RCO), and hydroxyl oxygens are key ligands. Adjacent glucuronic acid residues create a strong bidentate/bis-carboxylate pocket. In PolyXGG, Ca²⁺ coordinated to all four carboxylate oxygens; K⁺ coordinated more weakly, often to one O from each carboxylate plus nearby hydroxyl/ring O.
• Coordination and bond metrics: • PolyXA: K⁺ and Ca²⁺ were bi-coordinated (CN=2) to O atoms; average K–O ~2.505 Å; Ca–O ~2.024 Å; Mulliken populations low, indicative of ionic bonds (K–O ~0.01–0.02 |e|; Ca–O ~0.30 |e|). Zn²⁺ did not bind. • PolyXGA: K⁺ CN=1 (K–O ~2.634 Å, ionic ~0.01 |e|); Ca²⁺ CN=5 (avg Ca–O ~2.36 Å, ionic ~0.12 |e|); Zn²⁺ CN=4 (avg Zn–O ~2.04 Å; one covalent Zn–O4 ~0.49 |e|). Resonance in RCOO⁻ splits charge between the two O atoms, giving similar Ca–O distances. • PolyXGG: K⁺ displayed one strongly bound site (CN=4) and a nearby non-bonded K⁺ likely deterred by K–K repulsion; average K–O bonds stronger/shorter here than in PolyXGA. Ca²⁺ sat centrally, bonding to four carboxylate O with similar lengths (within 0.12 Å), consistent with a symmetric pocket and ionic character (~0.08 |e| average).
• Governing factors: Stability is not dictated solely by charge density (Zn²⁺ > Ca²⁺ > K⁺). Steric compatibility and ionic radius are critical: smaller Zn²⁺ requires larger torsional distortions to access multi-oxygen pockets, reducing stability; Ca²⁺ and K⁺ better match the pocket size formed by glucuronates. Glucuronic acid residues are key determinants of binding strength; adding them dramatically improves Ca²⁺/K⁺ complex stability versus the xylan backbone alone.
• Backbone-only models: Ca²⁺ and K⁺ bound ionically to RCO and hydroxyl groups with moderate stability (Ca²⁺ Ef ≈ −0.99 eV; K⁺ Ef ≈ −0.42 eV), while Zn²⁺ remained unfavorable.

The DFT results directly address the research question by identifying that finger millet arabinoxylan strongly and favorably binds Ca²⁺ and K⁺—particularly within pockets formed by adjacent glucuronic acid residues—while Zn²⁺ forms unstable complexes or fails to bind. This provides a mechanistic rationale for observed mineral distributions in finger millet grains (high Ca and K; low Zn). The predominance of ionic interactions with carboxylate, carbonyl, and hydroxyl oxygens, the enhanced stability with glucuronate-rich motifs (PolyXGG), and the central role of sterics and ionic radius explain why Zn²⁺ is disfavored despite higher charge density. From a nutrition perspective, strong Ca²⁺/K⁺ binding to arabinoxylans does not necessarily imply poor bioavailability, as gut microbiota degrade arabinoxylans, potentially releasing bound ions; indeed, in vitro and limited human data indicate moderate Ca bioavailability/retention, and fermentation/germination can increase bioaccessibility. The findings suggest Zn may be better delivered via fortification/supplementation outside the arabinoxylan matrix, whereas breeding or processing strategies that influence glucuronic acid content or fibre fermentability could impact Ca and K storage and release. The work highlights glucuronic acid residues as potential limiting factors for cation-binding capacity and underscores the need to characterize arabinoxylan structural variability across millet varieties and its effect on mineral interactions and bioavailability.

This study uses first-principles DFT to elucidate ion–arabinoxylan interactions in finger millet. It shows that Ca²⁺ and K⁺ form thermodynamically stable, largely ionic complexes, especially within binding pockets formed by two glucuronic acid residues (PolyXGG), whereas Zn²⁺ complexes are unstable across models due to steric incompatibility and required backbone distortions. Glucuronic acid residues play a pivotal role in stabilizing cation binding and may limit overall binding capacity. These mechanistic insights rationalize observed mineral profiles in millet and provide a basis for designing confirmatory experiments and clinical studies on bioavailability. Future work should: (i) validate in vivo mineral binding/release and absorption; (ii) explore broader ion interactions (e.g., Fe, Mg, P) and co-factors (phytate, phenolics, tannins); (iii) assess variability of arabinoxylan structures across millet varieties and their fermentability; and (iv) consider breeding or processing to modulate glucuronic acid content and fibre architecture to optimize mineral delivery.

• Modeling scope: Small, single-chain arabinoxylan models were used; real grain matrices likely contain multiple overlapping chains and cross-linked networks that may create additional or stronger chelation sites.
• Protonation states: Some hydroxyl deprotonations were applied for charge balance in certain models, which may not reflect physiological conditions; only glucuronic acid carboxylates are expected to be deprotonated in planta/GI environments.
• Ion coverage: Only K⁺, Ca²⁺, and Zn²⁺ were studied; interactions with other nutritionally relevant ions (e.g., Fe, Mg, P) were not modeled.
• Co-components: Potential interactions with phytate, phenolics, and tannins, which may modulate mineral binding/bioaccessibility, were not included.
• Structural flexibility: The xylan backbone was relatively inflexible in models; larger-scale conformational variability and solvent/environmental effects were not explicitly treated, which could alter binding geometries and stabilities.