Simultaneous iron, zinc, sulfur and phosphorus speciation analysis of barley grain tissues using SEC-ICP-MS and IP-ICP-MS

Abstract

The increasing prevalence of iron (Fe) and zinc (Zn) deficiencies in human populations worldwide has stressed the need for more information about the distribution and chemical speciation of these elements in cereal products. In order to investigate these aspects, barley grains were fractionated into awns, embryo, bran and endosperm and analysed for Fe and Zn. Simultaneously, phosphorus (P) and sulfur (S) were determined since these elements are major constituents of phytic acid and proteins, respectively, compounds which are potentially involved in Fe and Zn binding. A novel analytical method was developed in which oxygen was added to the octopole reaction cell of the ICP-MS. This approach greatly improved the sensitivity of sulfur, measured as ⁴⁸SO⁺. Simultaneously, Fe was measured as ⁷²FeO⁺, P as ⁴⁷PO⁺, and Zn as ⁶⁶Zn⁺, enabling sensitive and simultaneous analysis of these four elements. The highest concentrations of Zn, Fe, S and P were found in the bran and embryo fractions. Further analysis of the embryo using SEC-ICP-MS revealed that the speciation of Fe and Zn differed. The majority of Fe co-eluted with P as a species with the apparent mass of 12.3 kDa, whereas the majority of Zn co-eluted with S as a 3 kDa species, devoid of any co-eluting P. Subsequent ion pairing chromatography of the Fe/P peak showed that phytic acid (myo-inositol-1,2,3,4,5,6-hexakisphosphate: IP₆) was the main Fe binding ligand, with the stoichiometry Fe₄(IP₆)₁₈. When incubating the embryo tissue with phytase, the enzyme responsible for degradation of phytic acid, the extraction efficiency of both Fe and P was doubled, whereas that of Zn and S was unaffected. Protein degradation on the other hand, using protease XIV, boosted the extraction of Zn and S, but not that of Fe and P. It is concluded that Fe and Zn have a different speciation in cereal grain tissues; Zn appears to be mainly bound to peptides, while Fe is mainly associated with phytic acid.

---

Introduction

Iron (Fe) deficiency is the most common nutritional disorder in the world, being most prevalent in developing countries where cereals constitute the major part of the diet. On a global scale, Fe deficiency affects almost 4 billion people of which 25–50% have such low Fe status that they suffer from anaemia.¹ Zinc (Zn) deficiency is also a major problem in human nutrition, but compared with Fe deficiency, the global occurrence is less clear. A survey across 19 Chinese provinces showed that 60% of all children between 0 and 10 years of age suffered from Zn deficiency.² A number of more recent Chinese surveys have confirmed this staggering prevalence of Zn deficiency in young children.³ Thus, Fe and Zn deficiency seriously reduce human health and prevent the economic development of many poor communities across the world.

The bioavailability of Fe and Zn in cereal grains is generally low and believed to be controlled by phytic acid (myo-inositol-1,2,3,4,5,6-hexakisphosphate; IP₆), which is the main storage form of phosphorus (P) in cereals.⁴IP₆ as well as some of the less phosphorylated forms (IP_1–5) have a high affinity for various cations including Fe, Zn, Mn, Ni, Cd, Cu, Mg, Ca, Na and K,^5–8 forming both water soluble and insoluble chemical species. Imaging techniques such as Secondary Ion Mass Spectrometry (SIMS) and Energy Dispersive X-ray analysis (EDX) have been used in combination with Electron Microscopy to confirm that electron-dense IP₆ molecules co-localize with a range of cations and proteins in globoids located in the protein storage vacuoles of the aleurone layer and the embryo.^9–11 The co-localization of Fe, Zn and IP₆ in the embryo and aleurone tissues has been used to postulate that these cations are speciated with IP₆, but a very limited amount of evidence indicates that such species actually exist in planta. Therefore, the objective of the present study was to analyse the speciation of Fe and Zn in the cereal grain, using SEC-ICP-MS and barley as the experimental case. We focus especially on the interactions between Fe and Zn and their speciation with P and S, the latter two elements being major constituents of IP₆ and proteins, respectively, potentially involved in Fe and Zn binding.

Simultaneous analysis of Fe, Zn, P and S with a quadrupole-based ICP-MS system is an analytical challenge when analysing specific tissue fractions of which a limited quantity is available. In addition, the ⁴⁰Ar¹⁶O⁺ and ¹⁶O₂⁺ polyatomic interferences on the most abundant Fe and S isotopes, ⁵⁶Fe and ³²S respectively, imply that standard mode ICP-MS analysis is dependent on the detection of the much less abundant ⁵⁷Fe and ³⁴S isotopes, which severely reduces the sensitivity. The negative impacts of Ar-based interferences are typically reduced using an ICP-MS equipped with a reaction cell system. Pressurizing the ICP-MS reaction cell system with H₂ leads to a specific and efficient removal of ⁴⁰Ar¹⁶O interferences on ⁵⁶Fe, enabling analysis of ⁵⁶Fe in the low ng L⁻¹ concentration range. However, H₂ addition is detrimental to the ion transmissions of S, P and Zn. In the current study, addition of O₂ to the octopole reaction system enabled satisfactory signal intensities of S when measured as ⁴⁸SO⁺, Fe as ⁷²FeO⁺, P as ⁴⁷PO⁺ and Zn as ⁶⁶Zn⁺. The signal-to-noise ratios and LODs were markedly improved for Fe and S.

The SEC-ICP-MS analysis of the barley embryo fraction clearly showed that the chemical speciation of Zn and Fe differed. Fe mainly existed as a 12.3 kDa Fe:IP₆ oligomer, whereas Zn was mainly speciated with proteins or peptides in a species with the apparent mass of 3 kDa. Thus, current assumptions as to how Zn and Fe are stored in the mature barley grain must be re-evaluated. Analytically, the methods developed pave the way for structural identification of Fe and Zn species in the cereal grain controlling the bioavailability of these two essential elements to humans.

Experimental

Instrumental

Polishing of the grains was performed in zirconium (Zr) tubes mounted on a MM301 Retsch mixer mill (Retsch, Germany).

Before analysis, grain material was freeze dried at 0.5 mbar for 24 h (Christ Alpha 2–4, Martin Christ GmbH, Osterode, Germany) and samples were subsequently digested in a microwave oven (Multiwave 3000, mode: Synthos, Software version 2.01, Anton Paar GmbH, Graz, Austria) mounted with a 64 position carousel (64MG5, Anton Paar GmbH, Graz, Austria).

Total elemental concentrations in whole grains and tissue fractions were analyzed by ICP-OES (PerkinElmer Optima 5300DV, PerkinElmer, USA) with the following instrument settings: RF power, 1400 W; nebulizer gas flow, 0.65 L min⁻¹; auxiliary gas flow 0.2 L min⁻¹; plasma flow, 15 L min⁻¹; sample flow, 1.5 mL min⁻¹. The ICP-OES was equipped with a concentric Meinhard nebulizer and a cyclonic spray chamber. Zn, Fe, P and S were analysed in either axial (A) or radial (R) mode, using the wavelengths 206.209A, 238.216A, 214.924R and 181.983A, respectively. ICP-OES data was processed using Winlab 32 (Ver. 3.1.0.0107, PerkinElmer, USA).

SEC-ICP-MS analyses were performed on a HPLC (Agilent 1100 Series, Agilent Technologies, UK) coupled to a Diode Array Detector (DAD) and an ICP-MS (Agilent 7500ce, Agilent Technologies, UK) equipped with a PFA micro-flow nebulizer (Ezylok Micromist Nebulizer, Glass Expansion, Australia), an octopole reaction cell (ORC), and an optional mass flow controller enabling He/O₂ flow rates from 0.1 to 1 mL min⁻¹. All connections were made from PEEK, including tubing (0.17 mm id). The chromatographic data were processed using Plasma Chromatographic Software v. B-02-04 (Agilent Technologies, UK). The column thermostat used for cooling the SEC column was a Mistral (Spark, The Netherlands). Fractionated peaks were automatically collected and pooled using the Agilent 1100 Series Fraction Collector (Agilent Technologies, UK).

Chemicals and reagents

The CPI standards for ICP-MS calibration were Peak Performance: P/N 4400-ICP-MSCS, P/N4400-132565A and P/N4400-132565B from CPI International, USA. Certified reference material, NIST 8436 (durum wheat) was from the National Institute of Standards and Technology (Gaithersburg, MD, USA). 30% H₂O₂ (Riedel-de Haën, Sigma-Aldrich, Austria) and 70% HNO₃ (J. T. Baker, The Netherlands) was used as digestion media.

Myo-inositol-2-monophosphate (IP₁) and monocalcium-myo-inositol-1,2,3,4,5,6-hexakis-phosphate (Ca:IP₆) were from Sigma-Aldrich, Austria. Zn₇MT2a (from rabbit liver) was obtained from Bestenbalt LLC, Estonia. Cu–Zn superoxide dismutase (from bovine erythrocytes), myoglobin (from horse heart), carbonic anhydrase (from bovine erythrocytes) and vitamin B₁₂ (cyanocobalamin) were all purchased from Sigma-Aldrich, Austria. The enzymes protease XIV (bacterial; from Streptomyces griseus, pepsin (from porcine stomach mucosa) and phytase (from wheat), were also from Sigma-Aldrich, Austria. The reaction gas (10% O₂ in He) was purchased from AGA, Sweden.

All methanol used was of CHROMASOLVE quality (Riedel-de Haën, Sigma-Aldrich, Austria). The tetrabutylammonium hydroxide solution (∼40%) and citric acid-monohydrate were from Fluka (Sigma-Aldrich, Austria). Potassium hydroxide was from Merck, Germany. TRIS-HCl, TRIS-base and EDTA were from Sigma-Aldrich, Austria. Quartz sand (SiO₂) from Fluka (Sigma-Aldrich, Austria) was used in all extractions.

Milli-Q element water (18.2 Ω) was used throughout the experiment (Milli-Q Plus, Millipore Corporation, MA, USA).

Analytical procedures

Plant material and tissue fractionation. Grains of field-grown barley (Hordeum vulgare cv. Golden Promise) were used in all experiments. The grains were rinsed rapidly 3 times in Milli-Q water and subsequently freeze dried for 24 h. Only grains weighing 40–50 mg were used to get a homogeneous bulk grain sample. The awns and embryo of each grain were removed and separated by hand using a Teflon-coated scalpel. The rest of the grain was transferred to a Zr tube, where the inner walls were coated with sand paper (P-120). The Zr tube was agitated at 30 shakes s⁻¹ for 2 × 20 min in order to polish off the bran (the fused testa and aleurone tissue layers) from the endosperm (the main constituent of white flour). The abraded material was carefully collected from the Zr tube and the sandpaper.

The required polishing time was optimized using colour indicators for Fe (blue) and Zn (red). The staining procedure is described below. When the grain could no longer be stained it was concluded that only endosperm remained. All tissue fractions were weighed and summed up for comparison with the whole grain weight.

Tissue staining methods. 500 mg L⁻¹ of DTZ (diphenyl thiocarbazone) dissolved in 100% MeOH was used as Zn-indicator, and 2% potassium-ferrocyanide (Perl’s Preussian blue) as Fe-indicator. For Zn staining, the barley grains were put in the DTZ-solution for 15 min, followed by washing 3 times in Milli-Q water and air drying.¹² For Fe staining, the grains were soaked in 2% HCl solution for 10 min, then an equal amount of indicator solution was added. After a further 15 min, the grains were treated in the same way as in the Zn staining procedure.¹³

Elemental mass balance analysis of tissue fractions. Before analysis, samples were digested using a high-throughput micro-digestion technique.¹⁴ The micro-scaled digestion was carried out in disposable standard glass vials (Wheaton® 15 × 46 mm, Cap 13-425) that were acid washed prior to use. The vials were closed with special PEEK screw caps (MG5, Anton Paar GmbH, Graz, Austria) and disposable lip-type PFTE seals (Mat. No. 41186, Anton Paar GmbH, Graz, Austria) which ensured tightness and stability at the elevated temperatures.

For each tissue fraction, 7 replicates of 10–15 mg were transferred to the digestion vials. A mixture of 250 μL 30% H₂O₂ and 500 μL concentrated HNO₃ (70%) was added and vials were closed with seals and screw caps and eventually microwaved for 90 min using the following temperature program: 10 min ramping to 140 °C; temperature maintained for 80 min; cooling for 10 minutes. Samples were cooled in a freezer before uncapping to reduce the venting pressure. Finally, samples were diluted to a final concentration of 7% HNO₃ and analysed directly in the vial using ICP-OES.

For validation of the multi-elemental analysis, certified reference material was used (NIST 8436, from durum wheat).

Extraction procedure. 20 mg of embryo tissue were extracted with 2 mL of degassed 50 mM TRIS-HCl buffer (pH 7.5) and 300 mg of acid-washed quartz sand. The extractions were performed over a period of 60 min using an ice-cold mortar and a pestle under a flow of N₂ gas in order to prevent oxidation. The pestle was used every 15 min during the incubation period to homogenize the sample. After 60 min, the samples were centrifuged (16 000 × g, 2 °C, 15 min). The supernatant was ultra-filtered with a 50 kDa cut-off spin filter (Microcon centrifugal filter device; YM-50, Millipore, USA). Several extraction buffers (ammonium acetate, 50–100 mM, ±NaCl or ±DTT, pH 7.0–7.8) were tested, but with no significant improvement compared to 50 mM TRIS-HCl (pH 7.5) buffer, which was subsequently used in all extractions.

The enzymatic extractions were performed in a mortar with a pestle at pH 7.5 in 2 mg mL⁻¹ phytase/50 mM TRIS-HCl, and in 0.2 mg mL⁻¹protease XIV/TRIS-HCl. The extracts were incubated at 37 °C for 18 h, then pelleted and the supernatant ultra-centrifuged using a Microcon filter with 50 kDa cut-off. Protease XIV has broad substrate specificity and is stable at pH 5–9. It hydrolyzes peptide bonds on the carboxyl side of aspartic and glutamic acid.

Elemental total concentrations of the extractions were analysed by ICP-MS, using NIST 8436 (durum wheat) as CRM and CPI standards for calibration. 0.5 mL of each extraction was diluted 10 times with 3.5% HNO₃ prior to analysis. All calibration standards and blanks were prepared in the same matrix.

Column and column materials. Several column types and chromatographic methods were tested in order to identify the most robust system, maintaining the integrity of the metal binding species. Also, different reverse phase and ion exchange columns were tested, but with poor results due to de-stabilization of the metal binding species. Size exclusion chromatography markedly improved the stability of the species, and different stationary phases were tested, including cross-linked dextran and agarose gel beads, which were superior in terms of species stability. The Superdex 75 10/300 GL column (Glass, 10 × 300 mm, 13 μm cross-linked agarose/dextran, Amersham Biosciences, USA) with an optimum separation range between 0.7 and 70 kDa was selected for all further experiments.

SEC-ICP-MS analysis. The SEC column was mass calibrated by DAD at 214 nm and with elemental detection, using Cu–Zn superoxide dismutase (32 kDa, containing Zn, Cu and S), myoglobin (16.95 kDa, containing Fe and S), Zn₇MT2a (6.59 kDa, containing Zn and S), vitamin B₁₂ (1.36 kDa, containing Co and P) and myo-inositol-2-monophosphate (0.46 kDa, containing P). A linear regression curve was constructed from the retention times and the log transformed molecular masses of the calibrants.

The ICP-MS was tuned in standard mode and the tune settings were extrapolated to He mode using the ChemStation software (Agilent Technologies, UK). The flow rate of O₂ (10% in He) to the octopole reaction system (ORS) was initially set to 0.5 mL min⁻¹ and subsequently optimized by monitoring the ion intensity at m/z 48 (³²S¹⁶O) using a tune solution of S, P (25 μg L⁻¹), Fe and Zn (5 μg L⁻¹) in 50 mM TRIS-HCl (pH 7.5) with fixed octopole and quadrupole bias voltages set at −16 V. The bias voltage settings were then optimized in order to obtain the ideal compromise setting, providing the highest oxide formation of P, S and Fe at m/z 47, 48 and 72, while ensuring the highest ion transmission for Zn at m/z 66.

During method development, the following oxide and non-oxide ions were monitored: ²⁴Mg⁺, ³¹P⁺, ⁴⁴Ca⁺, ⁴⁷PO⁺, ⁴⁸SO⁺, ⁵⁷Fe⁺, ⁷²FeO⁺, ⁶⁶Zn⁺, ⁶⁷Zn⁺, ⁶⁸Zn⁺, ⁵⁵Mn⁺, ⁵⁹Co⁺, ⁶⁰Ni⁺, ⁶³Cu⁺, ⁶⁵Cu⁺ and ¹¹⁴Cd⁺. In order to correct for plasma instability, vitamin B₁₂ (cobalamine ) was added as internal standard to all extracts and monitored as ⁵⁹Co⁺.

External calibration curves were measured by direct aspiration into the ICP-MS run in oxygen mode, where the signals of the calibration standards were recorded directly from the tune window. From the linear regression of these calibration curves, the LOD values were calculated.

Several different mobile phases (potassium phosphate, ammonium acetate, TRIS-HCl), concentrations (1–100 mM) and pH values (4–8) were tested to obtain the most robust SEC and ICP-MS conditions. Based on this, TRIS-HCl (50 mM, pH 7.5) was selected as mobile phase with a flow rate of 1.0 mL min⁻¹. Five to 100 μL of sample was injected on the SEC column, using a 25 min (1500 s) run time. The column temperature was kept at 5 °C during the whole analysis using a Mistral column thermostatic unit (Spark, The Netherlands).

After each run the column was rinsed by an automatic injection program set at 5 repetitive 20 μL injections of 10 mM EDTA/50 mM TRIS-HCl (pH 7.5), with a 3 min delay between each injection, as described in detail by Persson et al. (2006).¹⁵ The column was cleaned more rigorously every 10^th injection by injection of 500 μL solution consisting of pepsin (1 mg mL⁻¹), phytase (2 mg mL⁻¹), NaCl (0.5 M) and acetic acid (10%), followed by incubation for 1 h at 37 °C and re-equilibration with mobile phase.

The external calibration for quantification of chromatographic peaks was performed using flow injection analysis where the calibrants were injected into the ICP-MSvia the HPLC, without column. Elemental standards were used for Fe and P external calibration, but for Zn : S ratios, Cu–Zn superoxide dismutase (SOD) was used as calibrant. Measured relative to a series of reference metallopeptides, this improved the accuracy of the Zn : S ratios, probably by ensuring uniform peak shapes of the ⁴⁸SO⁺ and the ⁶⁶Zn⁺ signals, hence enabling a more accurate integration of the peaks.

Ion pairing chromatography . The IP_1–6 ligands were identified using the ion pairing chromatography method by Helfich and Bettmer (2004).¹⁶ The column was a reverse phase C₁₈ Phenomenex Luna (5 μm particle size, 100 Å pore size, 150 × 2.00 mm ID, Phenomenex, Germany). The mobile phases were Milli-Q elemental water (A) and 10% MeOH/20 mM citrate/0.04% TBA (tetrabutylammoniumhydroxide) (B). The pH of mobile phase B was adjusted to 6, using KOH. The following gradient was used: 0 min: 25% B → 5 min (300 s): 100% B → 25 min (1500 s): 100% B → 35 min (2100 s): 25% B. The outlet from the HPLC was hyphenated to the ICP-MS where the inositol phosphates (IP₁–IP₆) were detected as ⁴⁷PO⁺ in O₂ mode. The IP₆ standard was incubated 1 : 1 with a 2 mg mL⁻¹ phytase/50 mM TRIS-HCl solution for 8 h at room temperature. After injection of this solution on the column, the IP₁–IP₅ standards could be identified.

Results and discussion

Elemental concentrations in tissue fractions

Barley grains were fractionated into the following tissue types: awns (outermost layer), embryo, bran layers (including the testa and the nutrient-dense aleurone layer), and endosperm. Less than 3% of the total grain material was lost in the fractionation procedure. The largest tissue fraction by weight was the endosperm, followed by the bran layers, the awns and finally the embryo (Table 1). The endosperm accounted for approximately 68% of the total grain dry weight and the bran layers around 22%, whereas the embryo only accounted for approximately 4%.

Table 1 Whole barley grain and its main tissue fractions (values are means ± std, n = 7). Each tissue type was weighed and analysed separately on ICP-OES. The recovery shows how much of the total (weight or amount of element) that were accounted for

	Total	Awn	Bran layers	Embryo	Endosperm	Recovery (%)
			mg
Dry weight	43 ± 3	2.4 ± 0.1	9.5 ± 0.3	1.8 ± 0.1	29 ± 0.5	98
Elements			μg g^−1
Zn	30 ± 1	15 ± 0.6	47 ± 1	164 ± 4	14 ± 1	91
Fe	50 ± 3	67 ± 3	128 ± 6	80 ± 3	25 ± 1	105
P	3740 ± 11	1330 ± 70	6710 ± 90	13 600 ± 170	1640 ± 80	87
S	1470 ± 20	490 ± 30	1740 ± 11	3990 ± 120	1160 ± 60	93

---

The recoveries of Zn, Fe, P and S, measured as the ratio between the accumulated content of the four grain fractions and that of the whole grain, ranged from 87% to 105% (Table 1). Thus, the procedure used was robust and sample loss and contamination were reduced to an acceptable level. The coefficient of variance (CV%; n = 7) ranged between 1 and 6%, depending on the element and the tissue in which it was analysed (Table 1).

The concentrations of Zn and Fe in the whole grain (Table 1) were within the normal range of 20–40 μg Zn g⁻¹ DW and 40–80 μg Fe g⁻¹ DW observed for field grown barley.^17,18 The embryo was the most elementally dense tissue fraction and, especially for Zn and P, the concentrations were 4–5 fold higher than those in the whole grain. The embryo was the only compartment where the Zn concentration exceeded that of Fe.

The non-uniform distribution of Fe and Zn among the different grain tissue types is stunning. For example, the dry weight of the embryo was only 4% of the total grain weight, but contributed to approximately 23% of the total Zn content. On the contrary, the endosperm only contributed approximately 33% of the total Zn content even though it accounted for 68% of the total dry weight. In practice, this implies that removing the outer layers and the embryo may reduce the Zn content by more than 60% compared to that of the whole grain. Polishing of grains to remove the embryo and grain layers is common practice throughout the world, mainly to avoid problems with rancidity during grain storage. The negative consequence of polishing is that a major part of the essential trace elements is lost, contributing to the Fe and Zn deficiency problems described above.

Initial SEC-ICP-MS speciation analyses

Size exclusion chromatography (SEC) using dextran covalently bound to porous agarose beads has previously been shown to facilitate the speciation analysis of labile coordination complexes, causing only marginal disintegration of the species of interest.^15,19–21 The superiority of agarose–dextran gels in terms of maintaining the integrity of the Zn and Fe species extracted from the embryo tissue was similarly observed in the present work.

In the initial SEC-ICP-MS analyses, Fe mainly overlapped with P at R_t = 720 s, equivalent to a molecular size of 12.3 ± 3.1 kDa (data not shown). Additionally, a minor Fe/P peak with the apparent mass of 0.5 ± 0.2 kDa eluted at R_t = 920 s.

Zinc, on the other hand, eluted at R_t = 864 s, clearly separated from both the first and second P/Fe peaks. The apparent mass of the Zn peak was 3.0 ± 0.7 kDa. These initial results clearly indicated a de-coupling of the Fe and Zn speciation, but the speciation could only partly be analysed due to poor sensitivity in standard mode, especially for ³⁴S⁺ and ⁵⁷Fe⁺. The sensitivity for Fe could be significantly improved by running the multi-elemental speciation analysis in H₂ mode, whereas sufficient S sensitivity was lacking both in standard and H₂ mode. As sensitive measurement of S was crucial in order to reveal the presence of peptide species, development of a method allowing simultaneous detection of S, Fe, Zn and P was required.

Speciation analysis in O₂ mode

Sulfur analysis using quadrupole-based ICP-MS is a challenging task, mainly due to the following three problems: (i) high ionization energy of S, which implies that only a limited fraction of the total S is ionized in the plasma source; (ii) heavy interference by ¹⁶O₂ on the major ³²S isotope, restricting detection to the ³⁴S-isotope with only 4.3% abundance; (iii), interference from a range of polyatomic species derived from nitrogen and oxygen (e.g.¹⁵N¹⁸O¹H⁺, ¹⁶O¹⁸O⁺, ¹⁷O₂⁺) on the ³⁴S-isotope, causing an elevated background signal that further impedes detection.

In order to achieve a sufficient S signal, O₂ was added to the octopole, converting the major S isotope (³²S) into the S oxide ion (⁴⁸SO⁺) via a reaction which is thermodynamically favourable (ΔH_f = −6.2 kJ mol⁻¹). Using this approach it was possible to avoid the ¹⁶O₂ interference and measure S as the oxide at m/z 48, where isobaric overlap with the main Ti isotope (73.7% abundancy) is found, but fortunately not with ion intensities significantly exceeding the background noise. Also Ca has an isotope on m/z 48 which might bias the ⁴⁸SO⁺ signal. However, the inherently low concentration of this phloem-immobile element in the cereal grain, in combination with the low abundance of the ⁴⁸Ca isotope (0.187%) implies that the interference from Ca should be insignificant. This assumption was supported by the fact that no co-eluting signal was detected for the ⁴⁴Ca isotope (2.086% abundance) in the chromatogram, strongly indicating that the isobaric interference from the ⁴⁸Ca signal was negligible.

Using O₂ addition to the octopole, the signal-to-noise ratio for S detection at ⁴⁸SO⁺ was improved markedly relative to measuring ³⁴S in standard mode (Fig. 1). The LOD was improved from 25 to 3.4 μg S L⁻¹. There is no data available from previous studies using this analytical approach on an ICP-MS equipped with an octopole reaction cell system. However, using oxygen addition to an ICP-MS system with dynamic reaction cell (DRC), Smith et al. (2003)²² reported an increase in S sensitivity by a factor of 100. LOD values of 0.2 μg S L⁻¹ and 4.3 μg S L⁻¹ have been achieved by other groups, also on DRC systems.^23,24

Fig. 1 Signal-to-noise ratios of Zn, Fe, P, S, where O₂ mode (dotted lines) is compared to standard mode ICP-MS (solid lines).

Matrix-based polyatomic interferences on m/z 48 could potentially occur from the nitrogen-derived ions ³⁴S¹⁴N⁺, ¹⁴N¹⁶O¹⁸O⁺ and ¹⁴N¹⁷N₂⁺. The occurrence of such interferences was checked by spiking the sample into a N-free mobile phase (0.1% TFA) and comparing it to the N-containing TRIS-HCl buffer. The background signals were similar in these two mobile phases, strongly indicating interference-free sulfur detection (data not shown). These findings fit with thermodynamic data indicating that oxygen transfer to nitrogen or to an additional oxygen atom is thermodynamically unfavourable under the conditions given.^22–24

As is evident from Fig. 2, the O₂ mode enabled sufficient sensitivity to allow speciation analysis of S thereby providing information which was not feasible in either standard, H₂ or He mode.

Fig. 2 SEC-ICP-MS chromatogram showing an embryo sample where the ion intensity signals for ³⁴S⁺ (standard mode) and ⁴⁸SO⁺ (oxygen mode) are overlaid.

Using the same approach as described above, P could be measured as ⁴⁷PO⁺ (Fig. 1). The signal-to-noise ratio for ⁴⁷PO⁺ was almost similar to that of ³¹P in standard mode and consequently the LOD was only marginally improved, decreasing from 6.1 to 4.6 μg P L⁻¹ as measured by direct aspiration into the ICP-MS.

In O₂ mode, measurement of Zn as ⁶⁶Zn¹⁶O⁺ at m/z 82 resulted in a drastic drop in sensitivity (data not shown) and might lead to isobaric overlap with ⁸²Se⁺. Consequently, Zn was analysed as ⁶⁶Zn⁺ with potential interferences from a number of S containing polyatomic ions such as ³⁴S¹⁶O₂⁺, ³²S¹⁶O¹⁸O⁺, ³²S¹⁷O₂⁺, ³²S³⁴S⁺ and ³³S₂⁺. However, spiking the TRIS buffer with methionine as S source showed that the impact of these polyatomics were insignificant under the conditions given. Measuring ⁶⁶Zn⁺ in oxygen mode resulted in a 15% decrease in ion intensity due to decreased ion transmission. However, the background signal was also decreased, resulting in a modest increase in signal-to-noise ratio (Fig. 1).

Iron measured as ⁵⁷Fe⁺ gave low signal intensity already in standard mode and with a further decrease in oxygen mode, the signal was hardly detectable. The system was therefore optimized to analyse Fe as an oxide similar to the principles outlined for S and P. Measuring the product ion ⁷²FeO⁺ enabled use of the most abundant ⁵⁶Fe isotope (91.75%) instead of ⁵⁷Fe with only 2.12% abundance. The signal-to-noise ratio was increased markedly by this approach and the LOD was lowered almost 20-fold to 0.5 μg Fe L⁻¹ (Fig. 1). Simultaneous analysis of S and Fe was achieved by Hann et al. (2004)²³ with a DRC system, but with optimization of the ⁵⁶Fe and ⁵⁴Fe signals, while the signal of the ⁷²FeO⁺ product ion was too low to allow Fe detection at m/z 72. The reaction between Fe and O₂ producing ⁷²FeO⁺ has a ΔH_f of +79 kJ mol⁻¹ and is not a spontaneous reaction under the conditions occurring in the octopole.²⁵ The ⁷²FeO⁺ formed in the octopole in the present work probably originated from a two-step process in which Fe was first converted to FeO₂⁺ followed by reaction with oxygen. The formation of FeO₂⁺ is highly spontaneous with a ΔH_f of −356 kJ and may occur in the flame of the plasma.²⁶ FeO₂⁺ can then react with one additional oxygen in the octopole to form the FeO⁺ ion. This reaction is also thermodynamically favourable with a ΔH_f = −58 kJ.

At m/z 72 there is a potential isobaric overlap with the main isotope of Ge (⁷²Ge). However, Ge is only found in ultra-trace quantities in biological tissues, including the grain tissues used in the present work (data not shown). It will therefore not significantly influence the accuracy of the method.

The low background signal at 72 m/z indicated that no significant interferences were present, although theoretically both plasma- and matrix-based interferences from ³⁶Ar₂⁺, ³⁶Ar³⁶S⁺, ³⁶S¹⁸O₂⁺, ³⁶S₂⁺, ³⁷Cl¹⁷O¹⁸O⁺ and ³⁵Cl³⁷Cl⁺ could occur at this m/z value. Fortunately, these isotopes have low abundances and high ionization potentials, and did not, therefore, cause significant analytical bias.

The accuracy of analysing P, S, Fe and Zn in O₂ mode was tested on a number of reference proteins with known Zn/S and Fe/S ratios. Very good agreement with the theoretical sulfur : metal ratios was obtained (Table 2).

Table 2 The theoretical and the analysed sulfur : metal ratios of the reference proteins used. The proteins were measured in the concentration range 0.8–3 μM (n = 3)

Enzyme/protein	Carbonic anhydrase (bovine erythrocytes)	Metallothionein 2A (rabbit liver)	Myoglobin (horse heart)
Molecular weight/Da	28 983	6581	16 954
Histidines	11	0	11
Methionines	4	1	2
Cysteines	0	20	0
Theoretical S : metal ratio	4	3	2
Analysed S : metal ratio	4.08 ± 0.02	3.06 ± 0.06	1.91 ± 0.05

---

Using the oxygen mode it was now possible to analyse P, S, Fe and Zn in one single chromatographic run (Fig. 3). Fe overlapped with P as a major (12.3 kDa) and minor peak (0.5 kDa), whereas Zn eluted in one single peak (3.0 kDa) between both the Fe and P signals. Sulfur appeared as two distinct peaks of which the first marginally co-eluted with the first eluting Fe/P signal. The second S peak perfectly co-eluted with the Zn signal at 864 s, indicating the presence of a peptide ligand.

Fig. 3 SEC-ICP-MS speciation of Fe, P, Zn and S in an embryo sample, run in O₂ mode. The column was mass-calibrated with reference compounds as indicated at the top of the figure.

Ligand identification

The ratio of the S : Zn peak was 43 ± 6.2 S atoms per Zn, which indicates the presence of several peptides having only a limited fraction of the thiol groups coordinated with Zn. The latter was supported by the fact that reverse-phase chromatography of the collected peak, with UV and ESI-MS detection, generated a range of peaks. Given the apparent complexity of the S ligand identification, we decided to focus on identification of the Fe/P peaks in the rest of this paper.

In order to identify the P-containing ligands from the SEC chromatogram, an extracted sample was incubated with phytase, the enzyme responsible for successive dephosphorylation of IP₆ in a stepwise and ordered manner.²⁷

An equal volume of the extractions incubated with phytase was injected on the SEC column at times 0, 2, 4 and 8 h. The 12.3 kDa P-peak gradually decreased in size during incubation with the phytase enzyme (Fig. 4). At the same time, the 0.5 kDa P peak increased equivalently. Since the phytase enzyme is very specific to IP₆,²⁸ the shift from the 12.3 kDa to the 0.5 kDa peak clearly showed that the apparent IP₆ oligomer was successively degraded into lower IPs (IP_1–5). Also the co-eluting Fe peak of the 12.3 kDa oligomer eluted later in the chromatogram (data not shown), confirming that Fe was indeed speciated with IP₆ or with a mixture of different IPs.

Fig. 4 Incubation of the extracted embryo sample using the IP₆-degrading enzyme phytase. The ⁴⁷PO⁺ signal at time zero is identical to the ⁴⁷PO⁺ signal in Fig. 3.

In order to identify the particular IPs involved in forming an oligomer with Fe, ion pairing chromatography was implemented in hyphenation to O₂ mode ICP-MS.¹⁶ The Fe/P peaks from the SECchromatography (Fig. 3) were collected and re-injected on this system, where a standard containing IP₁–IP₆ had been identified. The largest Fe/P peak from the SEC chromatogram (peak 1) consisted mainly of IP₆ molecules (Fig. 5), and the smallest Fe/P peak (peak 2) was dominated by IP₁ and orthophosphate PO₄⁻³. In the largest Fe/P peak there were also small amounts of IP₅, IP₄ and IP₁ present.

Fig. 5 Ion pairing chromatography of collected peaks from the SEC-ICP-MS analysis. The IPs were detected as ⁴⁷PO⁺ in O₂ mode.

Since the molecular weight of IP₆ is only 660 Da, the 12.3 kDa peak may represent an oligomer of many IP₆ molecules, stabilized by hydrogen bonding and ionic bonding with cations, including Fe. This hypothesis was tested by analysing the sample under the same conditions but with the pH lowered to 2, which destabilizes the strong tendency of IP₆ monomers to form hydrogen and ionic bonds in an IP₆ oligomer. Indeed, the low pH delayed the elution of the P peak by approximately 120 s, now eluting as a 0.7 ± 0.3 Da compound (data not shown), which fits well with the molecular size of one single IP₆ molecule.

The ratio between P and Fe in the 12.3 kDa Fe:IP₆ oligomer was quantified using external calibration of ⁴⁷PO⁺ and ⁷²FeO⁺. The ratio was found to be 27.7 ± 0.4 (n = 3) P atoms per Fe atom. Since each IP₆ molecule contains 6 P atoms, the IP₆ : Fe ratio would be 4.6. With the apparent mass of 12.3 kDa, these data indicate the following stoichiometry: Fe₄(IP₆)₁₈ (molecular mass: 12.1 kDa). However, it should be noted that traces of several cations were found in the P-containing fractions and that the existence of pure Fe–P species in planta is rather unlikely.

Extraction and speciation efficiencies

Only a limited proportion of the total content of Zn, Fe, P and S was extractable in 50 mM TRIS-HCl buffer (Table 3). The extractability was 15 and 18% for Zn and Fe, respectively, and slightly higher for P and S; 38 and 56%, respectively. The low extractability of Fe and Zn was probably due to the presence of water-insoluble precipitates, including IP₆ complexes. In addition, adsorption to negatively charged cell walls may have limited the extractability. The speciation efficiency, i.e. the ratio between eluted and total amount added to the column, was 80% of the TRIS-extracted Fe and Zn (Table 3). The remaining 20% represented free ions and labile species degrading on the column. Under the conditions given, free ions were quantitatively trapped on the column material, and could only be eluted by a repetitive injection of EDTA.¹⁵

Table 3 The extraction efficiencies for TRIS-HCl, phytase and protease XIV. The speciation efficiency is the percentage of each element passing through the SEC column

Element	Total conc./μg g^−1	TRIS extraction/μg g^−1	Phytase extraction/μg g^−1	Protease XIV extraction/μg g^−1	Speciated (% of total)
Zn	164 ± 3.8	23 ± 3.8	17 ± 2.1	128 ± 15	11 ± 0.7
Fe	80 ± 3.3	14 ± 2.0	27 ± 1.9	9 ± 3.6	15 ± 1.7
P	13 600 ± 170	5160 ± 260	11 600 ± 1700	3870 ± 460	32 ± 1.3
S	3990 ± 120	2250 ± 65	2320 ± 79	3830 ± 180	37 ± 2.7

---

In order to investigate how representative the speciation analysis was for the embryo samples, two additional extraction procedures were performed, using either protease or phytase enzymes.

Incubation with protease XIV caused the extraction efficiency of Zn to increase 5-fold compared to the TRIS extraction (Table 3). 78% of the total Zn could now be extracted whereas the extraction efficiency of Fe was largely unaffected. At the same time, the amount of extracted S almost doubled compared to that in TRIS and increased to 96% of the total S, whereas the extraction of P did not increase at all. Based on the increase in extraction efficiency with protease XIV relative to the TRIS extraction, the following elemental series could be deduced, going from highest increase to lowest, i.e. decreasing importance of peptide speciation:

Ca > Zn > Mn > Mg > S > Cu > K > P > Fe

Interestingly, incubation with phytase produced the opposite results to those observed for protease: the extraction efficiency increased for Fe, but not for Zn (Table 3). Compared to the TRIS extraction, the extraction efficiency in the presence of phytase was doubled for Fe but was unchanged for Zn. Also P was more than doubled whereas S was not affected. The P extracted with phytase now accounted for 85% of the total P. Listed according to decreasing extraction efficiency with phytase relative to TRIS, the elemental sequence was:

Mn > Mg > P > Fe > Cu > K > Ca > S > Zn

The fact that phytase increased the extraction efficiency of Fe and P, but not that of Zn and S, and that protease XIV increased the extraction efficiency of Zn and S but not that of Fe and P, strongly supports the results from the speciation analysis, showing that Fe was speciated with IP₆ in contrast to Zn, which was speciated with peptides.

Conclusions

Simultaneous SEC-ICP-MS speciation analysis of S, P, Fe and Zn was possible with a method based on the analysis of ⁶⁶Zn⁺ and the oxide product ions ⁴⁸SO⁺, ⁴⁷PO⁺ and ⁷²FeO⁺. By addition of O₂ to the octopole reaction system the signal-to-noise ratios and LODs were markedly improved for Fe and S.

When the method was applied to extracts of the embryos of barley seeds, Fe mainly co-eluted with P as a 12.3 kDa Fe:IP₆ oligomer, while Zn co-eluted only with S, indicating a peptide ligand. Incubation with phytase strongly increased the extraction efficiency of Fe and P from the embryo, but not that of Zn. In contrast, Zn and S extractability were strongly increased following incubation with protease. These data suggest that the well-established dogma stating that Zn and Fe share the same speciation and that their binding and bio-availability in the cereal grain are controlled by IP₆ is questionable. Future investigations of Fe and Zn speciation in the endosperm, which is the most important grain fraction with respect to human nutrition may benefit from the analytical progress presented here.

Acknowledgements

The authors are grateful to Bente Broeng and Bente Postvang for their skilful technical assistance. This work was supported financially partly by The Danish Research Council for Technology and Production Sciences (project 23-04-0082), HarvestPlus, the EU-FP6 projects META-PHOR (FOOD-CT-2006-03622), PHIME (FOOD-CT-2006-016253) and the ICROFS project ORGTRACE.

References

1. R. M. Welch and R. D. Graham, Field Crops Res., 1999, 60, 1–10.
2. T. Ma and Y. L. Kou, Journal of Guangdong Micronutrient Science, 2003, 10, 46–47.
3. X.-E. Yang, W.-R. Chen and Y. Feng, Environ. Geochem. Health, 2007, 29, 413–428.
4. V. Raboy, Phytochemistry, 2003, 64, 1033–1043.
5. E. Vasca, S. Materazzi, T. Caruso, O. Milano, C. Fontanella and C. Manfredi, Anal. Bioanal. Chem., 2002, 374, 173–178.
6. H. Persson, M. Türk, M. Nyman and A.-S. Sandberg, J. Agric. Food Chem., 1998, 46, 3194–3200.
7. J. Torres, S. Dominguez, M. F. Cerda, G. Obal, A. Mederos, R. F. Irvine, A. Diaz and C. Kremer, J. Inorg. Biochem., 2005, 99, 828–840.
8. C. De Stefano, D. Milea and S. Sammartano, Thermochim. Acta, 2004, 423, 63–69.
9. J. N. Lott and E. Spitzer, Plant Physiol., 1980, 66, 494–499.
10. E. N. C. Mills, M. L. Parker, N. Wellner, G. Toole, K. Feeney and P. R. Shewry, J. Cereal Sci., 2005, 41, 193–201.
11. F. Hatzack, K. S. Johansen and S. K. Rasmussen, J. Agric. Food Chem., 2000, 48, 6074–6080.
12. L. Ozturk, M. A. Yazici, C. Yucel, A. Torun, C. Cekic, A. Bagci, H. Ozkan, H.-J. Braun, Z. Sayers and I. Cakmak, Physiol. Plant., 2006, 128, 144–152.
13. S. Krishnan, K. Datta, N. Baisakh, M. De Vasconcelos and S. K. Datta, Curr. Sci., 2003, 84, 1232–1234.
14. T. H. Hansen, K. H. Laursen, D. P. Persson, S. Husted and J. K. Schjoerring, unpublished work.
15. D. P. Persson, T. H. Hansen, P. E. Holm, J. K. Schjoerring, H. C. B. Hansen, J. Nielsen, I. Cakmak and S. Husted, J. Anal. At. Spectrom., 2006, 21, 996–1005.
16. A. Helfrich and J. Bettmer, J. Anal. At. Spectrom., 2004, 19, 1330–1334.
17. D. J. Reuter, D. G. Edwards and N. S. Wilhelm, in Plant Analysis, ed. D. J. Reuter and J. B. Robinson, CSIRO Publishing, Collingwood, 2nd edn, 1997, vol. 2, ch. 5, pp. 84–95.
18. S. Husted, B. F. Mikkelsen, J. Jensen and N.-E. Nielsen, Anal. Bioanal. Chem., 2004, 378, 171–182.
19. V. Vacchina, K. Połeć and J. Szpunar, J. Anal. At. Spectrom., 1999, 14, 1557–1566.
20. J.-C. Shen, J. L. Liu, Z.-X. Zhuang, X.-R. Wang and F. S. C. Lee, Talanta, 2006, 69, 988–995.
21. J. Szpunar and R. Lobinski, Anal. Bioanal. Chem., 2002, 373, 404–411.
22. J. Smith, I. D. Wilson, L. Weidolf, F. Abou-Shakra and M. Thomsen, Chromatographia, 2004, 59, 165–170.
23. S. Hann, G. Koellensperger, C. Obinger, P. G. Furtmüller and G. Stingeder, J. Anal. At. Spectrom., 2004, 19, 74–79.
24. D. R. Bandura, V. I. Baranov and S. D. Tanner, Anal. Chem., 2002, 74, 1497–1502.
25. NIST_homepage: http://kinetics.nist.gov/kinetics/index.jsp.
26. B. Kellogg and K. K. Irikura, J. Phys. Chem. A, 1999, 103, 1150–1159.
27. L. Bohn, L. Josefsen, A. S. Meyer and S. K. Rasmussen, J. Agric. Food Chem., 2007, 55, 7547–7552.
28. X. G. Lei and J. M. Porres, Biotechnol. Lett., 2003, 25, 1787–1794.

---