SEC-ICP-DRCMS and SEC-ICP-SFMS for determination of metal–sulfur ratios in metalloproteins

Abstract

Simultaneous determination of Fe/S and Mn/S ratios on transient signals was performed by size exclusion chromatography, hyphenated to inductively coupled plasma mass spectrometry with dynamic reaction cell technology (SEC-ICP-DRCMS), in order to characterize metalloprotein samples by their metal/sulfur ratio. Oxygen was used as the cell gas. The method eliminates the effect of polyatomic isobaric interferences at m/z = 32 by detecting sulfur as the product oxide ion ³²S¹⁶O which is less interfered. Using the same reaction gas conditions, Fe and Mn were measured at m/z 54, 56 and 55, respectively. SEC-ICP-DRCMS measurements (injection volume 20 µL, SEC flow 300 µL min⁻¹) resulted in excellent limits of detection (LOD). 4.3, 0.4, 2 and 0.6 ng g⁻¹ were assessed for ³²S¹⁶O, ⁵⁵Mn, ⁵⁴Fe and ⁵⁶Fe, respectively. Reference measurements were carried out by size exclusion chromatography inductively coupled plasma sector field mass spectrometry (SEC-ICP-SFMS), setting the mass resolution at 4500. LODs of 14, 0.5 and 0.4 ng g⁻¹ were obtained for ³²S, ⁵⁵Mn and ⁵⁶Fe, respectively. The metal/sulfur ratios of 5 commercially available metalloproteins were determined (myoglobin, haemoglobin, cytochrom c, arginase and Mn superoxide dismutase from E. coli). Two proteins were characterized after in-house heterologous expression in a host organism (Mn superoxide dismutase from Anabaena PCC 7120, catalase-peroxidase from Synechocystis PCC 6803). Different calibrants (i.e., Fe³⁺, Mn²⁺, SO₄²⁻, methionine, myoglobin) for the assessment of inter-elemental ratios have been employed. It was found that calibration using metalloprotein myoglobin is preferable to inorganic standards in terms of uncertainty of measurement. However, all metal/sulfur ratios of the investigated proteins obtained by the different methods agreed within their total combined uncertainty.

---

Introduction

Nowadays, inductively coupled plasma mass spectrometry (ICP-MS) is the most widely used on-line multi-element detector for hyphenated techniques.^1,2 Extremely low limits of detection, a wide linear dynamic range, multi-element capabilities, surveyable spectra and a high sample throughput are the benefits of the method. A trend in the field of speciation by ICP-MS is the detection and identification of ligand–metal complexes in biological samples.³ Metalloproteins, trace metal complexation in blood and blood plasma, selenoproteins in human and animal body fluids and tissues, metallodrugs and their interaction with proteins are subjects of investigation in these studies.⁴ Since their discovery, metal-binding proteins have been the focus of research in biology and medicine because of their various functions in connection with transport, storage and detoxification of both essential and toxic trace elements in different organisms. They play a crucial role in a number of key metabolic processes. One third of all enzymes are metalloproteins. The metals either fulfil structural functions and/or are essential for the catalytic process.

As such, ICP-MS does not provide any structural species information, as all molecules introduced into the high temperature ion source are broken down into atoms, which are subsequently ionized. Species identification is accomplished on the basis of retention time matching with standards solely. On the other hand, as a key advance ICP-MS is a generic detection method allowing species unspecific element selective quantification. The identification and quantification of trace elements in proteins remains an important task. Several HPLC-ICP-MS applications exploit the multi-element capability of ICP-MS to study metal-protein binding.⁵ Association of various elements with proteins has been investigated by several authors focusing on speciation in biological materials.^5–8 Bidimensional HPLC, i.e., separation by size-exclusion HPLC and subsequent SEC-size fraction analysis by ion exchange HPLC-ICP-MS, was implemented.^4,9

In this study SEC-ICP-MS will be presented as a method of characterizing already isolated metalloproteins. Conventionally, the molecular metal content of unknown or biotechnologically produced proteins is assessed by determination of the total metal concentration in solutions of isolated protein. The protein concentration of this investigated solution is obtained by photometry. The accuracy of this approach is compromised by the poor precision of the protein quantification and by the fact that the measurement solution might contain metal impurities as a result of the protein isolation procedure (e.g., affinity chromatography utilizing metal containing eluents is often employed). Since protein isolation is a multiple-step procedure, determination of the methodological metal blank is tedious. Quantification of sulfur in metalloproteins offers an alternative for protein quantification. As the amino acid sequence is known, and hence the amount of sulfur containing amino acids, the sulfur content gives an accurate value of the molar protein concentration. Consequently, the metal/sulfur ratio characterizes the stoichiometric metal content of the protein.⁷ Furthermore, SEC accomplishes separation of the metalloproteins from potential low molecular weight impurities prior to ICP-MS detection. Thus the method provides a valuable tool for quality control of metal integration in biotechnologically produced (heterologously expressed) metalloproteins.

Detection capabilities and limitations of an ICP-DRC-MS device by reporting metal–sulfur chromatograms will be illustrated. The performance of different ICP-MS instrumentation (ICP-SFMS, Element1, Thermo Finnigan, Bremen, Germany; ICP-DRC-MS, ELAN DRC II, PerkinElmer SCIEX, Concord, ON, Canada) will be compared for the simultaneous detection of problematic elements such as sulfur and iron. For the first time results obtained by DRC technology using oxygen as reaction gas for monitoring of transition metals and sulfur in transient signals will be presented. In order to obtain molar metal–sulfur ratios a correction factor accounting for the different response of the determined elements (S, Mn, Fe) was introduced. Determination of this correction factor by using inorganic standards, aminoacids or well-known metalloproteins is critically discussed.

Experimental

SEC-ICP-MS

Separation of metalloproteins from inorganic impurities was performed via SEC using a 350 × 2.0 mm PEEK column, which was packed with Sephadex G25 superfine (Sigma–Aldrich Chemie GmbH, Vienna, Austria). Ultrapure sub-boiled water and ultrapure CH₃COONH₄ (Suprapure, VWR, Darmstadt, Germany) were used for preparation of the SEC buffer. The buffer concentration was 20 mM CH₃COONH₄ (pH = 6.0). A DX 500 chromatography system (Dionex Corp., Sunnyvale, CA, USA) was operated at a flow rate of 300 µL min⁻¹. The injection volume was 20 µL. The separation column was kept at 20.0 °C throughout the measurements. The column effluent was directly coupled to the two different ICP-MS introduction systems using 0.01″ id PEEK tubing. In the case of the sector field instrument (Element 1, ThermoFinnigan, Bremen, Germany), the introduction system comprised a PFA-nebulizer (Elemental Scientific Inc., Omaha, NE, USA) and a “Scott”-type spray chamber. The introduction system of the quadrupole based ICP-MS (Elan DRC-II, PE SCIEX, Ontario, Canada) consisted of a PFA-nebulizer and a cyclonic spray chamber. Oxygen (purity 4.5, Linde Gas GmbH, Vienna, Austria) was used as reaction gas. The ICP-MS operation parameters are given in Table 1.

Table 1 ICP-MS operation parameters

	ELAN DRC II	Element 1
Nebulizer	PFA	PFA
Spray chamber	Cyclon	Scott
Nebulizer gas flow	1.0 L min^−1	1.2 L min^−1
Auxiliary gas flow	1.275 L min^−1	0.6 L min^−1
Plasma gas flow	15 L min^−1	13.0 L min^−1
ICP RF power	1075 W	1250 W
Ion lens voltage	6.35 V (fixed)	—
O2 flow rate	0.6 mL min^−1	—
RPQ	0.3	—
Axial field voltage	250 V	—
m/z measured (S, Mn)	47.97, 54.94	31.9715, 54.9375
m/z measured (S, Fe)	47.97, 53.94, 55.93	31.9721, 55.9344
Mass resolution	400	4500
Scan Mode	Peak hopping	Electric scanning
Magnet settling time	—	0.3 s (S), 0.08 s (Mn/Fe)
Mass window	—	75%
Chromatographic data points s^−1	2 s^−1	2 s^−1

---

Metalloprotein samples, chemicals and standards

Table 2 lists the investigated metalloproteins. Myoglobin (purity 95–100%, source: horse heart), hemoglobin (purity 95–100%, from bovine blood) and cytochrom c (purity 95%, source: horse heart) were purchased at Sigma Aldrich GmbH, Vienna, Austria. The preparation of catalase-peroxidase (purity > 99%, source: Synechocystis PCC 6803), a multi-functional heme enzyme with catalase, peroxidase and haloperoxidase activity, has been described elsewhere.¹⁰ Mn superoxide dismutase (purity > 99%: source, Anabaena PCC 7120) was prepared according to a procedure described by Atzhofer et al.¹¹ Catalase-peroxidase and Mn superoxide dismutase are available at Planta Naturstoffe GmbH, Vienna, Austria. L-Arginase (source, bovine liver) was purchased at Fluka, Buchs, Switzerland. Mn superoxide dismutase (purity > 91%, source: E. coli) was purchased at Fluka. All protein preparations were stored at −80 °C and diluted in 20 mM CH₃COONH₄ (pH = 6.0) prior to measurement.

Table 2 Stoichiometric metal and sulfur content of analyzed metalloproteins

Metalloprotein	Molecular weight /Dalton	Metal atoms per molecule	Sulfur containing amino acids per molecule	Stoichiometric sulfur ∶ metal ratio	Protein concentration in measurement solution^a/nmol g^−1
a Assuming 100% purity.
Myoglobin from horse heart	16 954	1 Fe	2	2	5.90
Hemoglobin from bovine blood	61 968	4 Fe	10	2.5	2.26
Cytochrom c from horse heart	11 702	1 Fe	4	4	12.8
Catalase-peroxidase from Synechocystis PCC 6803	168 910	2 Fe	54	27	0.592
Mn superoxide dismutase from E. coli	91 736	4 Mn	12	3	5.32
Mn superoxide dismutase from Anabaena PCC 7120	56 344	2 Mn	10	5	7.09
Arginase from bovine liver	69 478	2 Mn	7	3.5	1.64

---

Methionine (p.a.), S, Fe and Mn ICP-MS standard solutions for preparation of calibration standards were purchased at VWR (Darmstadt, Germany). The accuracy of the Mn and Fe calibrations was assessed by measurement of a standard reference material (TM-27.2, National Water Research Institute, Canada).

In this work, reference is made to the terminology: accuracy of measurement, standard uncertainty, expanded uncertainty and coverage factor according to the Guide to the Expression of Uncertainty in Measurement.¹² All combined uncertainties were calculated according to the ISO/GUM guide using the uncertainty propagation procedure.¹³ Dedicated software (GUM-workbench software, Metrodata GmbH, Grenzach-Wyhlen, Germany) was employed for the calculations based on the numerical method of differentiation.¹⁴ Processing of chromatographic data was carried out using Chromeleon (Version 6.40., Dionex Corp., Sunnyvale, CA, USA).

Results and discussion

Optimization of DRC conditions

In most applications so far the dynamic reaction cell has been operated with specific parameters for a selected isotope. However, when Mn/S and Fe/S ratios of metalloproteins are determined by SEC-ICP-DRCMS, reaction gas conditions have to be optimized suitably for simultaneous S, Fe and Mn detection. Generally, when a transient signal needs to be analyzed within a short period of time, the variation of the gas density in the DRC is not possible. Analyte signals are acquired on a timescale of milliseconds, which currently exceeds the necessary stabilization time of the conditions inside the reaction cell. Hence, the same gas conditions have to be applied. Concerning S, the thermochemistry is unfavorable for charge transfer reaction for elimination of the O₂⁺ interference since the interfering ion has a similar ionization potential to the analyte.¹⁵ A recent study employed Xe for partial removal of the O₂⁺ interference.⁷ As an alternative approach oxygen was used successfully to convert S ions via an exothermal reaction into the molecular ion of ³²S¹⁶O (m/z 48).¹⁶ In this case the adjustable DRC bandpass must include both the reacting analyte ions and the product ions. At m/z 48 several potential interferences (e.g., ⁴⁸Ti, ⁴⁸Ca, ³¹P¹⁶O¹H) can still hamper the accurate quantification of sulfur as ³²S¹⁶O. However, in our study inorganic impurities are separated from the proteins under investigation. Moreover, reference measurements by SEC-ICP-SFMS at an m/Δm of 4500 excluded the occurrence of such interferences in the presented samples. In our study oxygen was evaluated as reaction gas for additional Mn and Fe detection. The dependence of Mn and Fe signal intensity on the oxygen flow as well as the oxide formation rate of the selected isotopes was critically assessed. Fig. 1 depicts signal intensity versus reaction gas flow for ³²S¹⁶O, ⁵⁵Mn, ⁵⁴Fe, ⁵⁶Fe, ⁵⁷Fe and ⁵⁸Fe blank and standard solutions, respectively. At low oxygen flows, however, the cell gas is primarily cell gas entrained from the ion optics chamber.¹⁵ Consequently, much more oxygen might be present in the cell (as much as 17% oxygen) and the monitored curves might not represent the real figures. In accordance with theory, product ions such as ³²S¹⁶O, ⁵⁵Mn¹⁶O and ⁵⁶Fe¹⁶O show a maximum at moderate oxygen gas flow of 0.6 mL min⁻¹. Generally, under these conditions, MnO and FeO formation was in the range of 2% and regarded as negligible. Mn ions that do not react with the reaction gas experience multiple collisions, which can cause collisional focusing towards the cell axis and thus better transmission. The net effect on transmission depends on the ion–neutral mass ratio, the energy of the ions, rf-field and gas pressure.¹⁵Fig. 1 shows that Mn transmission experiences a maximum at 0.7 ml min⁻¹. Compared with the standard operation mode, the sensitivity of Mn determination is decreased (350 000 cps versus maximum 150 000 cps in the DRC mode). Moreover, the low background signal at m/z 56 indicated efficient removal of ArO⁺ ions. The other minor Fe isotopes (m/z 54, 57) showed improved sensitivity by taking advantage of collisional focusing with the pressurized cell at 0.7 mL min⁻¹ oxygen flow. Due to the low signal/noise ratio ⁵⁸Fe was not considered in the present study. As a result of these findings, the gas density parameter were adjusted to 0.6 mL min⁻¹ oxygen flow, in order to minimize the signal for ⁴⁰Ar¹⁶O⁺, but retaining the ³²S¹⁶O, ⁵⁵Mn ⁵⁶Fe, ⁵⁴Fe, ⁵⁷Fe ion transmission as high as possible.

Fig. 1 Optimization of O₂ reaction gas flow rate for simultaneous determination of Mn, Fe and S on transient signals using a solution containing 100 ng g⁻¹ S and 20 ng g⁻¹ Fe and Mn.

Linearity of concentration versus signal curves is a prerequisite for the assessment of accurate metal–sulfur ratios in metalloproteins. Fig. 2 shows calibration graphs monitored for the selected cell parameter using inorganic standard solutions diluted in the SEC eluent buffer. Concentrations were selected on the basis of expected concentrations for metalloproteins analysis. Operational parameters are summarized in Table 2. Linearity was found for the isotopes ³²S¹⁶O, ⁵⁵Mn, ⁵⁴Fe and ⁵⁶Fe. It was not possible, however, to calibrate ⁵⁷Fe. On the basis of this experiment, the isotopes ⁵⁶Fe and ⁵⁴Fe were selected for measurement by SEC-ICP-DRCMS. Table 3 gives the detection limits obtained by SEC-ICP-DRCMS. The detection limits were based on three times the standard deviation of the baseline signal quantified by peak height calibration. The respective peak heights were obtained from the average of 4 consecutive SEC-ICP-MS determinations of myoglobin standards (Fe concentration of 329 ng g⁻¹, S concentration of 378 ng g⁻¹, 20 µL injection volume, HPLC flow 300 µL min⁻¹ CH₃COONH₄ buffer). For Mn the detection limit was obtained from the average of 4 consecutive FI determinations of a 200 ng g⁻¹ standard (20 µL injection volume, HPLC flow 300 µL min⁻¹ acetate buffer). As the reference method in this study SEC-ICP-SFMS was used at the mass resolution setting of R = 4500, allowing interference free determination of S, Fe and Mn. As can be seen in Table 3, the corresponding detection limits are excellent and comparable to those obtained by SEC-ICP-DRCMS.

Fig. 2 Signal to concentration linearity of different Fe isotopes measured by ICP-DRCMS using the cell conditions listed in Table 1.

Table 3 Comparison of LODs obtained by SEC-ICP-SFMS and ICP-DRCMS using O₂ as reaction gas^a

Measured isotope	SEC-ICP-DRCMS/ng g^−1	SEC-ICP-SFMS/ng g^−1
a The values were calculated from 3 standard deviations of the blank signal. 20 µL of sample were injected into a SEC flow of 300 µL min^−1. b Measured as ^32S ^16O.
^32S	4.3^b	14
^55Mn	0.4	0.5
^54Fe	2	—
^56Fe	0.6	0.4

---

Determination of molar S/Fe ratios in metalloproteins

Four Fe-containing metalloproteins, namely myoglobin, hemoglobin, cytochrom c and catalase-peroxidase were investigated by SEC-ICP-MS. The protein concentration of the injected solutions is listed in Table 2. As can be seen in Fig. 3 separation of the commercially available hemoglobin is an absolute prerequisite since inorganic impurities obfuscate the otherwise adequate accuracy of S/Fe determination. The protein elutes at 2 min with the implemented experimental set-up. Table 4 lists the results obtained by 3 different calibration strategies using ICP-SFMS detection. Correction factors considering the different sensitivities of Fe and S detection in order to assess the correct molar ratio were introduced. These factors were determined by the following methods. (1) Flow injection of inorganic Fe and S standards in the form of Fe³⁺ and SO₄²⁻ was used for calibration and compared with (2) flow injection of inorganic Fe in the form of Fe³⁺ and S in the form of methionine. (3) Moreover, SEC-ICP-SFMS measurements using myoglobin as standard were employed. Periodic injection of 0.2 M HCl has found to improve the robustness of the hyphenation.

Fig. 3 Rapid separation of metalloprotein hemoglobin from inorganic Fe³⁺ and SO₄²⁻ using size exclusion chromatography in combination with (a) ICP-SFMS and (b) ICP-DRCMS. ⁵⁶Fe and ³²S measurements by ICP-SFMS were performed at high mass resolution (m/Δm = 4500). In the case of ICP-DRCMS, sulfur was measured as ³²S¹⁶O using oxygen as the reaction gas. The separation by SEC is a prerequisite for successful elimination of inorganic impurities and determination of accurate sulfur/metal ratios.

Table 4 Stoichiometric S/Fe ratio in Fe containing metalloproteins determined by SEC-ICP-SFMS and SEC-ICP-DRCMS^a

Protein	Purity (%)	S/Fe ratio theoretical	S/Fe ratio (SEC-ICP-SFMS) calibrated by SO4–Fe^3+via flow-injection	S/Fe ratio (SEC-ICP-SFMS) calibrated by methionin/Fe^3+via flow-injection	S/Fe ratio (SEC-ICP-SFMS) calibrated by S/Fe in myoglobin	S/Fe ratio (SEC-ICP-DRCMS) calibrated by S/Fe in myoglobin
a The stoichiometric ratio obtained by SEC-ICP-SFMS was obtained utilizing three different calibrants. The uncertainty given with the values represents the total combined uncertainty of the method (coverage factor 2; 4 replicate measurements).
Myoglobin	>95	2	1.89 ± 0.22	2.17 ± 0.21	—	—
Hemoglobin	>95	2.5	2.41 ± 0.40	2.77 ± 0.41	2.54 ± 0.38	2.81 ± 0.08
Cytochrome c	>95	4	3.46 ± 0.37	3.97 ± 0.32	3.65 ± 0.28	3.92 ± 0.08
Catalase-peroxidase	>99	27	20.5 ± 2.1	23.6 ± 1.8	21.7 ± 1.6	24.2 ± 0.6

---

In Table 4 the ratios obtained with the different calibration strategies are given with total combined uncertainties calculated according to Eurachem guidelines. Standard uncertainties associated with each of the individual variables of a simple measurement equation were combined using the method of propagation of uncertainty. For the comparison of the two different ICP-MS detection methods, calibration by myoglobin was selected, since in this case the standard preparation (mixture of Fe and S solutions) and different transport efficiencies for biomolecules and inorganic standards do not contribute to the final total combined uncertainty of S/Fe ratio determination. The S/Fe intensity ratios could be determined with excellent standard uncertainties (precision of N = 4 determinations) ranging from 1.5–3% and 0.7–1.2% for SEC-ICP-SFMS and SEC-ICP-DRCMS, respectively. For the FI experiments using ICP-SFMS detection precisions of 1.3–3% were assessed. As can readily be seen in Table 4 all molar ratios agree well within their total combined uncertainties showing the accuracy of species unspecific calibration of the relative sensitivity factor for S and Fe by FI and SEC-ICP-MS. Moreover, the suitability of both ICP-MS detection methods was found. In the case of catalase-peroxidase an S/Fe ratio of 22 was experimentally found instead of the theoretically predicted value of 27, revealing a yield > 100% for the metal integration in the recombinant metalloprotein. This could indicate unspecific binding of hemin to the protein surface, since during expression of catalase-peroxidase in E. coli hemin has been added to increase the yield of recombinant heme protein.¹⁰

Determination molar S/Mn ratios

Table 5 lists the investigated Mn containing metalloproteins and corresponding S/Mn ratios. The S/Mn molar ratios were obtained from FI calibration of the relative sensitivity factor using inorganic standard solutions (Mn concentration of 200 ng g⁻¹, S concentration of 1000 ng g⁻¹, 20 µL injection volume, HPLC flow 300 µL min⁻¹ acetate buffer). The precision of S/Mn intensity ratio determination ranged around 3% and 4% for ICP-DRCMS and ICP-SFMS detection, respectively. All S/Mn ratio values given in Table 5 agree within their total combined uncertainty. Both SEC-ICP-SFMS and SEC-ICP-DRC-MS investigation of this metalloprotein set revealed S/Mn ratios higher than the theoretical target values, indicating the presence of protein impurities (e.g., ≈9% in E. coli MnSOD) and/or incomplete metal integration. The low manganese content in recombinant MnSOD from Anabaena could correspond with its recently reported low specific activity in comparison with homologous MnSOD from E. coli.¹¹

Table 5 Stoichiometric S/Mn ratio in Mn containing metalloproteins determined by ICP-DRCMS and ICP-SFMS^a

Protein	Purity (%)	S/Mn ratio theoretical	S/Mn ratio (ICP-SFMS)	S/Mn ratio (ICP-DRCMS)
a The stoichiometric ratio was determined after calibration with a solution containing Mn and methionine. The uncertainty given with the values represents the total combined uncertainty of the method (coverage factor 2; 4 replicate measurements). b Estimated form amino acid sequence of arginase from human liver.
Mn superoxide dismutase from E. coli	>91	3	3.93 ± 0.40	4.03 ± 0.35
Mn superoxide dismutase from Anabaena PCC 7120	>99	5	17.6 ± 2.4	16.8 ± 1.5
Arginase	Unknown	3.5^b	84.8 ± 13	88.1 ± 6.8

---

In the case of commercially available arginase a ratio of 87 was found instead of an estimated ratio of 3.5. Fig. 4 shows the SEC-ICP-DRC-MS analysis of arginase. Quantification of the protein via the sulfur signal revealed a concentration of 1.60 ± 0.13 nmol g⁻¹ corresponding to the target value of 1.64 nmol g⁻¹ proving incomplete metal integration. The immense Mn signal at 4 min indicates significant manganese release due to protein degradation and unfolding processes during storage.

Fig. 4 Analysis of metalloprotein arginase by SEC-ICP-DRCMS. It was found that only 6% of the protein is present as metalloprotein. Moreover a high concentration of Mn and sulfur containing impurities was detected.

Conclusion

Using oxygen as reaction gas simultaneous ICP-DRC-MS detection of S, Fe, Mn could be performed. Both SEC-ICP-DRC-MS and SEC-ICP-SFMS provided a valuable tool for the accurate determination of the sulfur–metal ratios. Relative sulfur/metal sensitivity factors could be determined by species unspecific calibration using inorganic standard solutions. The accuracy of both FI and SEC-ICP-MS calibration was satisfying.

Investigation of selected commercially available and in-house produced metalloproteins clearly showed the necessity of separation of the metalloproteins prior to sulfur/metal molar ratio determination. Evidently, the accuracy and precision of conventional methods relating total metal concentrations to protein concentration determined by UV photometry is compromised. In this situation SEC-ICP-MS is a promising alternative for supporting the development of metalloprotein expression as a reliable tool for quality control in biotechnological production of metalloproteins. However, it has to be stressed that the implemented SEC method separates only inorganic impurities from proteins. Due to this low separation efficiency, in our study unambiguous metal/sulfur ratios in terms of metal integration are only provided for pure, isolated protein samples.

References

1. K. Sutton, R. C. Sutton and J. A. Caruso, J. Chromatogr. A, 1997, 789, 85.
2. F. Vanhaecke and G. Köllensperger, Detection by ICPMS, Handbook of Elemental Speciation, ed. R. Cornelis, Wiley & Sons, New York, in the press.
3. J. Szpunar, C. Witte, R. Lobinski and F. Adams, Fresenius’ J. Anal. Chem., 1995, 351, 351.
4. J. Szpunar and R. Lobinski, Pure Appl. Chem., 1999, 71, 899.
5. J. Wang, R. S. Houk, D. Dreessen and D. R. Wiederin, J. Biol. Inorg. Chem., 1999, 364, 546.
6. H. Goenaga Infante, K. Van Campenhout, R. Blust and F. C. Adams, J. Anal. At. Spectrom., 2002, 17, 79.
7. D. Pröfrock, P. Leonhard and A. Prange, Anal. Bioanal. Chem., 2003, 377, 132.
8. C. N. Ferrarello, M. R. Fernandez de la Campa and A. Sanz-Medel, Anal. Bioanal. Chem., 2002, 373, 412.
9. J. Szpunar and R. Lobinski, Anal. Bioanal. Chem., 2002, 373, 404.
10. C. Jakopitsch, F. Rüker, G. Regelsberger, M. Dockal, G. A. Peschek and C. Obinger, Biol. Chem., 1999, 380, 1087.
11. W. Atzhofer, G. Regelsberger, U. Jacob, G. A. Peschek, P. G. Furtmüller, R. Huber and C. Obinger, J. Mol. Biochem., 2002, 321, 479.
12. Quantifying Uncertainty in Analytical Measurement, EURACHEM, London, 2000.
13. Guide to the Expression of Uncertainty in Measurement, International Organization for Standardization, ISO/GUM, Geneva, Switzerland, 1995, ISBN 92-67-10188-9.
14. J. Kragten, Analyst, 1994, 119, 2161.
15. S. D. Tanner, V. I. Baranov and D. R. Bandura, Spectrochim. Acta, Part B, 2002, 57, 1367.
16. D. R. Bandura, V. I. Baranov and S. D. Tanner, Anal. Chem., 2002, 74, 1497.

---