Trace iodine quantitation in biological samples by mass spectrometric methods

Talanta 79 (2009) 235–242

Contents lists available at ScienceDirect

Talanta journal homepage: www.elsevier.com/locate/talanta

Trace iodine quantitation in biological samples by mass spectrometric methods The optimum internal standard Jason V. Dyke a , Purnendu K. Dasgupta b,∗ , Andrea B. Kirk b a b

Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, TX 79409-1061, United States Department of Chemistry and Biochemistry, The University of Texas at Arlington, Arlington, TX 76019-0065, United States

a r t i c l e

i n f o

Article history: Received 26 January 2009 Received in revised form 13 March 2009 Accepted 16 March 2009 Available online 26 March 2009 Keywords: Iodine ICP-MS ESI-MS Iodine-129 Isotope dilution mass spectrometry

a b s t r a c t Accurate quantitation of iodine in biological samples is essential for studies of nutrition and medicine, as well as for epidemiological studies for monitoring intake of this essential nutrient. Despite the importance of accurate measurement, a standardized method for iodine analysis of biological samples is yet to be established. We have evaluated the effectiveness of 72 Ge, 115 In, and 129 I as internal standards for measurement of iodine in milk and urine samples by induction coupled plasma mass spectrometry (ICP-MS) and of 35 Cl18 O4 − , 129 I− , and 2-chlorobenzenesulfonate (2-CBS) as internal standards for ion chromatographytandem mass spectrometry (IC-MS/MS). We found recovery of iodine to be markedly low when IC-MS/MS was used without an internal standard. Percent recovery was similarly low using 35 Cl18 O4 as an internal standard for milk and unpredictable when used for urine. 2-Chlorobenzebenzenesulfonate provided accurate recovery of iodine from milk, but overestimated iodine in urine samples by as much as a factor of 2. Percent recovery of iodine from milk and urine using ICP-MS without an internal standard was ∼120%. Use of 115 In predicted approximately 60% of known values for both milk and urine samples. 72 Ge provided reasonable and consistent percent recovery for iodine in milk samples (∼108%) but resulted in ∼80% recovery of iodine from urine. Use of 129 I as an internal standard resulted in excellent recovery of iodine from both milk and urine samples using either IC-MS/MS and ICP-MS. © 2009 Elsevier B.V. All rights reserved.

1. Introduction During the present decade there has been great interest on perchlorate in the environment [1,2]. Although unreactive at physiological pH, perchlorate is a powerful competitive inhibitor for the transport of the essential iodide ion via the sodium-iodide symporter [3,4] and excessive intake of perchlorate may result in iodine deficiency. Obviously, this is of greater concern in a population that is already iodine deficient or in a borderline status [5,6]. Iodine nutrition status is generally judged by urinary output [5,7,8]. In addition, iodine content of food for infants, notably milk, is of great importance as iodine nutrition affects the neurodevelopment of the young [9–11]. There is thus great interest in trace determination of iodine, particularly in biological fluids such as milk and urine [12,13]. In milk and urine, iodine dominantly exists as iodide. The classical approach to iodine estimation is the kinetically based Sandell–Kolthoff reaction [14] which exploits the catalysis of the oxidation of As(III) with Ce(IV) by iodide. In idealized standards, the reaction can be followed either colorimetrically (Ce(IV) is yel-

∗ Corresponding author. Tel.: +1 817 272 3171. E-mail address: [email protected] (P.K. Dasgupta). 0039-9140/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2009.03.038

low) or fluorometrically (Ce(III) is fluorescent). However, many other compounds commonly present in biological samples can interfere. Such interferences are typically removed by digesting the sample with chloric acid at 105–115 ◦ C for 30–60 min, however, it is difficult to insure that all interfering species are removed [15]. More recently, digestion with less hazardous ammonium persulfate has replaced chloric acid; it is also easier to use, however, the samples must still be heated to 91–95 ◦ C for 30 min during digestion [16]. Substantial care must be exercised during digestion to prevent the loss of volatile iodine as iodide could potentially be oxidized into iodine by the persulfate. On the positive side, the Sandell–Kolthoff approach is one of the few inexpensively accessible spectrometric approaches that can provide the requisite sensitivity for measuring iodine at low levels in these samples. Mass spectrometry has long been a formidable tool for qualitative identification; it is also extraordinarily sensitive. Iodine determination by induction coupled mass spectrometry and iodide determination by chromatography–electrospray ionization mass spectrometry have both been widely used. When it comes to quantitative analysis, mass spectrometry is plagued by a number of problems. For accurate quantitation, one must be wary of isobaric interferences and matrix effects [17]. As a general issue in ICP-MS, polyatomic interferences and oxide formation must also

236

J.V. Dyke et al. / Talanta 79 (2009) 235–242

be avoided. Some of these negative effects can be overcome by specific knowledge of the behavior of the analyte of interest. In ICPMS, oxide formation can be overcome, for example, by properly choosing operating parameters, such as plasma gas composition, nebulizer flow rate, and radio frequency power [17,18]. Isobaric and polyatomic interferences can be overcome by either monitoring a different isotope of the same element or correcting for the interferant by quantitating it as well [17,19]. In ESI-MS isobaric interferences are typically overcome by preceding the MS with a liquid/ion chromatographic separation technique, which temporally separates the analytes that lead to mass fragments of identical mass to charge ratios (m/z, hereinafter the unit Thomson (Th) is used, see Ref. [20]). Non-isomeric isobaric interferences that persist in co-eluting after chromatography can be resolved in ESI-MS by tandem mass spectrometry (MS/MS) techniques. However, the one problem that cannot be solved is effects that arise from variations in the matrix components or as a result of the matrix itself. Matrix effects in MS constitute a more serious problem in two different ways. First, changing levels of electrolytes in the samples cause dramatic fluctuations in the ionization efficiency [21]. This change in ionization efficiency leads to changing signal intensities recorded by the MS. Second, matrix components can deposit in the MS, such as the entrance cones, and cause residue build up. These deposits can block the entrance to the mass analyzer and cause a reduction in signal over time [22,23]. Together or alone, ionization suppression and residue build up can lead to substantial errors. Unless compensated, such errors can amount to an order of magnitude. Effects from matrix variations and system stability can be corrected by using internal standards (IS) [24]. ISs are purposefully added to all samples and standards at equal concentrations. The IS is a component that is not indigenously present in any sample or is present at negligible concentrations compared to the amount that is added [25]. The IS should also not significantly affect the concentration of the AI present in the sample. The IS is typically added in equal amounts to all standards and samples. For an IS to function correctly, as system or matrix conditions change, any change in the signal attributed to the constant IS concentration must be reflected in a proportionate change in the analyte of interest (AI). The amount of the IS added should not markedly alter the overall sample composition/ionic strength and the IS should be of similar, but not identical, molecular/atomic weight as the analyte of interest [25,26]. The sensitivity of many mass spectrometers varies with the m/z ratio. Often low m/z ions are not as easily transferred from the ionization source to the mass analyzer compared to higher m/z ions [27,28]. In addition, the sensitivity of the instrument for the AI and IS should be similar for the IS to provide correct matrix compensation; this is especially true in ICP-MS. Finally, the IS should behave in a similar way to the AI. In ICP-MS the IS should have a similar first ionization potential to the AI [25,26]. In chromatography–ESI-MS/MS the IS should have an elution time close to that of the AI (ideally co-elute), so it elutes in a similar matrix as the AI. For this same reason, gradient elution should not be used in nonsuppressed ion chromatography (IC) or liquid chromatography (LC) when using an IS that does not fully co-elute with the AI. Isotope dilution mass spectrometry (IDMS) [29,30] is a special type of internal standardization in MS quantitation. IDMS uses an isotopically modified version of the AI as the IS. Using an isotopically modified version of the AI is the best way to meet all of the requirements of an IS. In perchlorate analysis, using quadruply 18 O-labeled perchlorate (35 Cl18 O4 , 107 Th) as an IS has proven, for example, to be the method of choice for quantitation [31,32]. An isotopically labeled IS must, nevertheless, meet the other requirements outlined earlier; in addition, within the experimental duration, the isotopic labeling must be stable [25].

We compare here the performance of several different species used as ISs for iodide/iodine analysis. For IC-MS/MS we compare the performance of 35 Cl18 O4 , 129 I, and 2-chlorobenzene sulfonate (2-CBS) as internal standards for iodide quantitation. For ICP-MS we compare the performance of 72 Ge, 115 In, and 129 I as ISs for iodine quantitation. In all cases the iodide quantitation data compensated by the appropriate IS are compared to uncorrected values and to each other. 2. Experimental 2.1. Standards and reagents All solutions were prepared using 18.2 M cm deionized water (Milli-Q Element A10 ultrapure water system, Millipore). Stock standard solutions (1000 mg/L each) of iodine, germanium, and indium were prepared by dissolving 0.1308 g of KI (www.vwrsp.com), 0.1927 g of InCl3 (www.sial.com) and 0.1441 g of GeO2 (www.Strem.com) in 100 mL of water, 2% HNO3 , and 40 mM NaOH, respectively. A 1.0 mg/L quadruply 18 O-labeled perchlorate solution (www.iconisotopes.com) and a 100 nCi carrier-free 129 I solution as KI (www.ipl.isotopeproducts.com) were purchased. Based on the certified value of the activity, and the known halflife of 129 I (1.57 × 107 years), the 100 nCi K129 I solution contained ∼56.6 mg/L 129 I. Working solutions of 1.0 mg/L of germanium, indium, and 129 I− were prepared by diluting 0.1 mL, 0.1 mL, and 1.765 mL, respectively, to a final volume of 100 mL with water. The 2-CBS solution was prepared in the laboratory from alkaline hydrolysis of the corresponding commercially available sulfonyl chloride (www.sial.com). 2.2. Ion chromatography and ESI-MS/MS All chromatography was preformed on a Dionex DX-600 ion chromatograph with an IS25 isocratic pump, EG40 eluent generator, an ASRS-Ultra 2-mm suppressor and a CD25 conductivity detector. Chromeleon 6.0 chromatography software was used for system control. Separation was performed on a Dionex IonPac AG-16 (2 mm × 50 mm) guard column and IonPac AS-16 (2 mm × 250 mm) anion separation column. The injection loop volume was 25 ␮L. The eluent used was 60 mM KOH at a flow rate of 0.25 mL/min. The IC effluent was sent to the ESI-MS/MS. A ThermoElectron TSQ Quantum Discovery Max MS was used in the negative electrospray ionization mode with a heated electrospray ionization probe (HESI) to increase sensitivity. The MS data was acquired using Xcalibur version 2.0 software package. The HESI probe temperature of the ESI-MS was set at 325 ◦ C, the ion transfer capillary was set at 275 ◦ C and the ionization potential was set at 4.5 kV. The MS operated in the selected reaction monitoring mode (SRM). The monitored ions had SRM m/z transitions of: 107.0 → 89.0 (35 Cl18 O4 − ), 127.0 → 127.0 (127 I− ), 129.0 → 129.0 (129 I− ), and 191.0 → 80.0 (2-CBS, 12 C6 H4 32 S16 O3 35 Cl− ). Iodide was quantified using the area ratio of iodide to the internal isotopic standard chosen for the correction and compared to a calibration curve composed of AI/IS ratios measured in the prepared standards. 2.3. ICP-MS An X Series II ICP-MS (www.thermo.com) was used in the direct infusion mode. The peristaltic pump built into the ICP-MS was used to prime the sample into the Peltier-cooled (3 ◦ C) nebulizer at 1.6 mL/min for 45 s and then continuously aspirate the sample into the nebulizer at 0.8 mL/min. Each measurement cycle consisted of a 20-s qualitative mass survey scan followed by three 32-s long quantitative mass

J.V. Dyke et al. / Talanta 79 (2009) 235–242

scans. Once the sampling was complete, the auto sampler probe was washed in DI water for 1 min before storage and reuse. 2.4. Sample preparation and addition of internal standard Most biological samples must be digested, extracted or otherwise processed before they can be analyzed. Ideally, spiking with an IS should be conducted as the very first step; this is especially true of IDMS. Addition of such an IS as the first step can compensate for any analyte loss during processing. However, for ISs that are not simple isotopically substituted analogs, if analyte loss does occur during sample processing, differential loss behavior of the IS and the AI can render accurate quantitation problematic. Fortunately, initial experiments in the present case for either the milk or urine samples showed that statistically the results did not differ whether the IS was spiked as a first step or just before analysis, suggesting that there was no (differential) loss of the analyte (or the IS) during processing. For the purposes of the present paper, the following reasons led us to chose to generate detailed data where the IS was added just before analysis: (1) the primary motivation was to identify the best (as well as acceptable) IS markers for IC–ESIMS/MS and ICP-MS determination of iodide/iodine and not to be confounded by recovery issues of specific sample processing methods, (2) much more sample is processed than is analyzed. Often, the processed/extracted samples are diluted further prior to analysis. Adding IS at the first stage and having a reasonable concentration of the IS in the final analyzed sample requires IS amounts that can increase the analytical cost/sample, especially when the IS is expensive, as is often true for an isotopically labeled IS. However, as a general analytical practice we do recommend that the IS be added to the sample prior to any manipulations. Human milk samples were processed by a streamlined procedure that has been described in detail elsewhere [33]. Briefly, 20 mL aliquots of the milk sample were put in 50-mL capacity centrifuge tubes and centrifuged at 16,000 rpm at 15 ◦ C for 25 min (Sorvall RC-6+, www.thermo.com; Fiberlite F-13-14X50CY rotor, www.piramoon.com). The liquid portion of the resulting sample was decanted to the top of a prewashed PL-10 Centricon Plus-20 centrifugal filter device (UFC2 LGC 24, www.millipore.com). The Centricon devices were centrifuged at 5000 rpm at 15 ◦ C for 90 min. A 0.5 g aliquot of prewashed Amberlyst 15 macroreticular cationexchange resin (www.sial.com) was added to the dialyzate; the sample was vortexed for 5 min and allowed to stand for 10 min. The sample was then taken into a prewashed 10-mL disposable syringe and passed through a prewashed 25 mm, 0.45 ␮m pore size nylon membrane syringe filter prior to analysis. Extracted milk samples were spiked with varying levels of iodide for standard recovery experiments. Milk samples were diluted 4× and 10× with water for IC-MS/MS and ICP-MS analysis, respectively. Samples for IC–ESI-MS/MS analysis were spiked to a concentration of 10 ␮g/L of 129 I− , 5 ␮g/L of 35 Cl18 O4 − , and 25 ␮g/L of 2-CBS prior to being loaded into the auto sampler (Finnigan Surveyor Auto sampler Plus, www.thermo.com). Samples for ICP-MS analysis were spiked to a concentration of 10 ␮g/L of 129 I− , 72 Ge, and 115 In prior to analysis. The milk samples used in this study had been analyzed before in a previous study [6]. The samples were selected such that one sample in each experiment had moderate to low levels of iodide present and the other had high levels of iodide present. Urine samples were filtered with a 0.45 ␮m nylon syringe filter and spiked with varying levels of iodide for standard recovery experiments. The samples were then diluted 5× and 10× with water for IC-MS/MS and ICP-MS analysis, respectively. Samples for ICMS/MS analysis were spiked to a concentration of 10 ␮g/L of 129 I− , 5 ␮g/L of 35 Cl18 O4 − , and 25 ␮g/L of 2-CBS prior to analysis. Samples for ICP-MS analysis were spiked to a concentration of 10 ␮g/L of 129 I− , 72 Ge, and 115 In prior to analysis. The urine samples used in

237

this study had been collected for a previous study [6] and analyzed before. The samples were selected such that one sample in each experiment had moderate to low levels of iodide present and the other had high levels of iodide present. 2.5. Standard addition and recovery Stock urine and milk extract samples were split into several equal aliquots and spiked with a high concentration 127 I standard to have final spiked iodide concentrations ranging between 0 ␮g/L and 150 ␮g/L I. The samples were then spiked with an individual IS whose efficacy was being tested. 2.6. Temporal reproducibility We conducted a study where various processed milk and urine samples (a total of 25), spiked with the appropriate IS, were each analyzed twice by IC-MS/MS and one particular additional milk sample was additionally inserted randomly in this sequence 10 times, each time being analyzed twice like all the other samples. The results for these 10 repeated probe samples (n = 2 ea) are discussed. 3. Results and discussion 3.1. Internal standard selection The choice of an appropriate IS for mass spectrometric iodine quantitation has not been obvious. Leading researchers in the field had used isotopically labeled perchlorate as an IS for both perchlorate and iodide quantitation [13]. Although this may provide some correction, it may not fully correct for matrix effects due to the different retention times and ionization behavior of the analytes. A more appropriate selection for an iodide IS might be a compound that co-elutes with iodide. It is known that 4-chlorobenzene sulfonate (4-CBS) co-elutes with perchlorate on certain IC columns [34]. Although iodide and perchlorate do not co-elute, they are both highly polarizable spherical monoanions (the reason why perchlorate physiologically competes with iodide) and elute close to each other. It is reasonable that a different chlorobenzenesulfonate isomer may co-elute with iodide on certain IC columns. We found that of the three isomers (2-, 3-, 4-), 2-CBS elutes the closest (retention time, tR , 7.3 min) to iodide (tR , 6.7 min) on an AS-16 (Dionex) IC column under our standard isocratic elution conditions (Fig. 1) and could even be made to co-elute at higher eluent strengths. The other alternative for an iodine IS is an iodine isotope. Iodine127 is the only naturally occurring stable isotope of iodine. There are several radioactive iodine isotopes that originate from the nuclear fission of uranium and plutonium [35] or from the cosmic ray spallation of atmospheric xenon [36]. Many of these isotopes, especially 131 I, are used in medicine as tracers for nuclear imaging, and for treating various thyroid-related conditions. The major problem with using a radiotracer in mass spectrometry is contamination of the equipment as well as vacuum and vent systems with radioactive material. This will be a particular problem with radioiodine because it is so easily taken up by the mammalian thyroid. Of the 14 major radioactive isotopes of iodine, only 129 I has a half-life (t½ ) > 60 d. 129 I has a t½ = 15.7 million years [37,38]. 129 I is the only isotope of iodine that would not lead to an unacceptable extent of radioactive contamination when used at measurable concentrations. This is because the half-life of 129 I is so long that an appreciable concentration represents very little radioactivity; for example, a 5 mL sample aliquot that has been spiked with 10 ␮g/L level of 129 I contains 387 pmol 129 I, amounting to <16 pCi in radioactivity. Relative to this, the Nuclear Regulatory commission considers the handling and purchase of up to 100 nCi of 129 I exempt from

238

J.V. Dyke et al. / Talanta 79 (2009) 235–242

Fig. 1. IC-MS/MS chromatogram of (A) the total ion current, (B) 5 ␮g/L 35 Cl18 O4 − , (C) 5 ␮g/L 127 I− , (D) 10 ␮g/L 129 I− , and (E) 25 ␮g/L 2-chlorobenzene sulfonate.

regulations [39]. 129 I decays by weak ␤- and ␥-emissions (64 keV and 25 keV, respectively, [38]), compared to this, on an average the human body contains 0.12 ␮Ci worth of 40 K, which also primarily emits ␤- and ␥-emissions (520 keV and 160 keV, respectively) [40,41]. The naturally occurring ratio of 129 I to 127 I is estimated at 1.5 × 10−12 [42]; it is found in higher ratios only around regions that have experienced nuclear fallout [43,44]. Since iodide derived from 127 I and 129 I are chemically equivalent, 129 I should be an ideal choice as an IS for iodide/iodine quantitation in any type of mass spectrometry; there are, however, no reports of its use thus far in chromatography–MS/MS analysis. In the case of ICP-MS, germanium, as germanium dioxide, has been claimed to be an appropriate IS for iodide analysis [45,46]. Germanium is an attractive element to use as an IS because it is not abundant in the earth’s crust (1.5 mg/kg) [47] and is not likely to be abundant in biological samples. Importantly, it also has a first ionization potential (IP) of 7.9 eV that is not far from that of iodine’s (10.45 eV) [48]. Excluding astatine, radon and xenon (which cannot be incorporated into a sample as a tracer for obvious reasons) the 10 elements closest in first IP (spanning 9.4–13.6 V) to iodine are Zn, Se, As, S, Hg, P, C, Br, Cl, and H. These are all likely to be present in biological samples, in varying and possibly significant levels. One drawback for 72 Ge is that its m/z ratio (72 Th) is much different from that of iodine (127 Th); this may put iodine and germanium into different intrinsic gain regions of the ICP-MS. Indium (115 In) is a very common IS used in ICP-MS analysis, widely chosen because of its low natural abundance and an intermediate value of 115 Th that falls nearly in the middle of the m/z values for the elements across the periodic table; it is considered a

good IS for many of the mid mass range elements. However, while indium has a close m/z value to that of iodine, at 5.79 eV its first IP is much lower [48]. As with IC–ESI-MS/MS, 129 I should be the best IS for iodine analysis as well. Being the same element, they of course have the same IP. The very small difference between the m/z value of the two isotopes puts them into the same sensitivity range of the ICPMS. The analytical problem of measuring iodine in foods and other biological material has been a challenging one and it has been so acknowledged for some time [49]. The advent of the ICP-MS did not necessarily mean that this was accepted as the best and most accurate solution. The use of 129 I as an IS for iodine determination by ICP-MS was simultaneously introduced by three different groups [50–52]. Haldimann et al., one of these pioneers, have continued to use 129 I as IS [53–55] and although they initially reported significant iodide memory effects in the nebulizer and some isobaric interference from 129 Xe present in the argon gas. Others have used 129 I as an IS with solid samples with the further addition of Pd, this iodophilic element prevents iodine loss during ashing and removal of organics [56]. Recently Bruchert et al. have used 129 I enriched iodide and iodate in gel electrophoresis coupled to an ICP-MS to measure these species in aerosol samples [57]. However, overall, not many have adopted 129 I as the IS of choice in iodine determination. Beginning with NHANES 2000, the Centers for Disease Control had adopted ICP-MS as the preferred method of iodine analysis in urine but utilized 130 Te as the IS [58]. The difference plots vs. the Sandell–Kolthoff method suggested that significant differences may exist for individual samples. Similarly, recent work by others continues to use other ISs, e.g., 103 Rh [59].

J.V. Dyke et al. / Talanta 79 (2009) 235–242

239

Fig. 2. Iodide recoveries (a) for a milk sample (Milk-1, Table 1) and (b) for a urine sample (Urine-2, Table 1). Different ISs were used to correct the iodide data obtained on an IC-MS/MS. The error bars indicate ±1 standard deviation.

3.2. Standard addition and recovery A good way to measure the effectiveness of an IS is to perform a standard addition experiment. Data from spiked samples (see Section 2) were collected and processed using the respective IS response for correction. These corrected data were compared among themselves as well as to results obtained without any correction. The correspondence of the slope of the standard addition plot (vide infra) to unity indicates the efficacy of a particular IS for the intended correction. 3.3. Spike recovery for added iodide IC-MS/MS Fig. 2a and b shows the standard addition results of a milk and urine sample, respectively. When no ISs were used, poor recoveries (slope of the standard addition curve 0.323 ± 0.090 to 0.854 ±0. 251) were observed for both milk and urine. These data are summarized in Table 1. Obviously, the quantitation of iodide/iodine is quite poor without an IS. The use of 18 O enriched perchlorate as an IS for iodide turns out to be a poor choice as well: the recoveries ranged from 38.0 ± 7.4% to 161.6 ± 15.3%. Three of the four samples analyzed showed recoveries below 78.1%, while one gave a significantly superquantitative recovery. This is likely due to the significantly different retention times of the two analytes in the chromatographic system resulting in varying amounts of other ion precursors eluting in the different temporal windows from one sample to another. The matrix composition thus changes dramatically in the windows that iodide and perchlorate elute and the two analytes are thus detected in very different matrices in different samples, regardless of any similarity between the analytes themselves. If in fact the two analytes also differ in their ionization behavior, further inaccuracies in quantitation will occur. It is to be noted that perchlorate was analyzed using the MS/MS mode and was fragmented in the second quadrupole of the triple quadrupole system during the analysis, while iodide, a monoatomic species, was not fragmented. Rather in the latter case

we sought to fragment most everything else that may be present with iodide in the second quadrupole such that the only 127 Th species to enter the third quadrupole should be iodine. However, even if the latter strategy is successful, it cannot compensate if differing extents of 127 I− are formed in the ionization process and are selected by the first quadrupole. Using 2-CBS as an IS has the advantage of more accurately reflecting the background matrix of the iodide peak as it elutes very close to iodide. In milk samples, 2-CBS indeed appears to be a good IS for iodide. It shows recoveries that range from 102.1 ± 14.6% to 104.9 ± 2.3% for the two test milk samples. However, the recovery of iodide using 2-CBS as an IS in urine samples is very different. In urine the iodide spike recovery was found to be between 206.2 ± 6.7% and 217.6 ± 19.1%. Since the IS correction is calculated as an area ratio of the iodide signal to the IS signal a recovery of 200% indicates that the 2-CBS signal has been reduced by ∼50% relative to iodide. It appears that some other species present in the Urine matrix may be selectively suppressing the ionization and detection of 2-CBS. Even though 2-CBS elutes near iodide and works well for milk samples, it is clear that it cannot be applied to all samples. The best results were obtained with 129 I-iodide as the IS. For the two initial test samples, the recoveries were 101.4 ± 6.2% and 98.9 ± 6.2%, respectively, for milk and urine (Table 1). These were by far the best recoveries for iodide in both milk and urine samples taken together. The isotopic IS has the advantage that it not only coelutes with iodide but it behaves identically to the analyte in the ionization process and has a very similar mass. The disadvantage to using the 129 I isotope is that trace amounts of 127 I are present within the stock solution. This results in a small amount of 127 I being added to the sample. However, since the same amount is added to all samples and standards, the minute amount is corrected by running a blank sample in the calibration curve. The cost of 129 I is not insignificant; 10 mL of a 5 mg/L stock solution costs over US$500 whereas 2-CBS can be readily made from 2-chlorobenzenesulfonyl chloride at a negligible cost.

Table 1 Iodide recoveries in milk and urine samples using different internal standards for signal correction on an IC-MS/MS. Slopes and correlation coefficients resulting from duplicate analysis of matrix spiked at five levels with 127 I and quantified using four different IS approaches. % Recovery, IC-MS/MS 129

No IS Milk #1 Milk #2 Urine #1 Urine #2

44.3 32.3 85.4 62.7

± ± ± ±

2

2.4 (r = 0.9910) 9.0 (r2 = 0.8101) 25.1 (r2 = 0.7937) 24.6 (r2 = 0.6838)

I

103.8 98.9 96.2 101.6

18

2-CBS ± ± ± ±

2

5.7 (r = 0.9912) 2.7 (r2 = 0.9977) 3.2 (r2 = 0.9966) 5.2 (r2 = 0.9922)

104.9 102.1 206.2 217.6

± ± ± ±

2

2.3 (r = 0.9986) 14.6 (r2 = 0.9422) 6.7 (r2 = 0.9969) 19.1 (r2 = 0.9773)

O perchlorate

56.5 38.0 78.1 161.6

± ± ± ±

3.5 (r2 = 0.9885) 7.4 (r2 = 0.8984) 2.4 (r2 = 0.9971) 15.3 (r2 = 0.9739)

240

J.V. Dyke et al. / Talanta 79 (2009) 235–242

Fig. 3. Iodide recoveries from (a) milk and (b) urine sample. Different ISs were used to correct the iodide data obtained on an ICP-MS. The error bars indicate ±1 standard deviation.

3.4. Spike recovery for added iodide ICP-MS

3.5. Temporal reproducibility

Iodide spike recoveries were greater than 100% when no IS was used to correct the raw ICP-MS data, as seen in Fig. 3a and b and Table 2. Iodine recovery was acceptable in only one of the urine samples (104.1 ± 0.8%). The other three samples exhibited a recovery averaging 120%. While 115 In may be a widely used IS for ICP-MS analysis, it performed poorly for iodine determination for both milk and urine, showing the lowest recoveries (61.9 ± 0.9% and 53.0 ± 7.1%, respectively) of the three evaluated ISs. The lower IP of indium (5.8 eV vs. 10.5 eV for I) allows it to be more easily ionized by the argon plasma than iodine which results in underestimation of the latter. Germanium has been reported as an appropriate IS for iodide analysis [45] and we have ourselves shown its suitability as an IS for iodine in iodized table salt samples [46]. However, diluted table salt solutions at a more or less constant concentration represent a more or less invariant matrix compared to biological samples such as milk and urine. With 72 Ge is used as an IS for iodine analysis in milk it shows recoveries are only slightly high (108 ± 1%). However, the recoveries were much lower (80.5–88.7%) for urine samples. The presence of Ge in urine is not the likely reason as Ge urinary concentrations in populations not subjected to occupational exposure to Ge are usually below the typical ICP-MS LOD for Ge (0.25 ␮g/L) [60]. Compared to In, the first ionization potential of Ge (7.9 eV) is closer to that of iodine but the m/z value is much less which may affect the correction. In any case, 72 Ge obviously would not appear to be a generally applicable choice as an IS for iodine analysis in all biological matrices. As with IC-MS/MS, 129 I was the best choice as an IS for iodide analysis by ICP-MS as well. Iodide spike recoveries were 99.1 ± 2.2% for milk and 102.6 ± 5.7% for urine. No other IS provided recoveries in both matrices that were as good.

Measurement reproducibility is an important consideration – in theory, even if an IS indicates low recovery, it can be successfully used if the recovery is consistently reproducible across samples. With many real samples, especially biological samples, traces of salt and organic matter can be deposited on various system components, especially the sample/skimmer cones between the atmospheric pressure region and the high vacuum region. In ICP-MS, material is also deposited in the torch. In IC-MS/MS, the ion transfer tube and electrospray needles are other susceptible regions. All can affect the eventual signal. To an extent, a good IS may be able to correct for such changing conditions within the MS. If it is unable to do so, a consistent change in the quantitative results will be seen as the same sample is run repeatedly. As described in the experimental section, the IC-MS/MS results for a total of 25 samples, each spiked with the appropriate IS, were each analyzed twice with one particular additional milk sample additionally inserted randomly in this sequence 10 times, each time being analyzed twice like all the other samples. The results for these 10 repeated probe samples (n = 2 ea) are shown in Fig. 4, as other samples are run. All results are reported with the first measurement of the test probe (Sample #3) normalized as 100; (based on the average of the measurements based on the 129 I IS, this sample contained 78.4 ± 4.9 ␮g/L I). Note that #1–15 are milk samples while #16–34 (with the exception of the repeated probe milk samples # at 23, 26, 29, 32, (and 35)) are urine samples. It will be observed that without any IS, even before the urine samples are run, the variability within individual measurements are very high, reaching to >50% in relative standard deviation. After beginning to run the urine samples, the situation only gets worse: The results range from ∼45% to ∼127% of the original value. Using the 35 Cl18 O4 as IS provided somewhat better reproducibility and

Table 2 Iodide recoveries in milk and urine samples using different internal standards for signal correction on an ICP-MS. Slopes and correlation coefficients resulting from duplicate analysis of matrix spiked at five levels with 127 I and quantified using four different IS approaches. % Recovery, ICP-MS 129

No IS Milk #1 Milk #2 Urine #1 Urine #2

118.0 122.0 119.5 104.1

± ± ± ±

2

1.9 (r = 0.9993) 1.8 (r2 = 0.9994) 3.8 (r2 = 0.997) 0.8 (r2 = 0.9998)

72

I

98.3 99.9 104.6 100.6

± ± ± ±

2

1.7 (r = 0.9991) 1.4 (r2 = 0.9994) 5.8 (r2 = 0.9910) 0.9 (r2 = 0.9998)

115

Ge

108.0 108.1 80.5 88.6

± ± ± ±

2

1.4 (r = 0.9995) 2.0 (r2 = 0.9990) 5.0 (r2 = 0.9886) 0.7 (r2 = 0.9998)

In

64.7 61.9 53.0 78.6

± ± ± ±

1.5 (r2 = 0.9985) 0.9 (r2 = 0.9993) 7.1 (r2 = 0.9486) 0.9 (r2 = 0.9996)

J.V. Dyke et al. / Talanta 79 (2009) 235–242

241

Fig. 4. Reproducibility of iodide measurement in a milk sample using different internal standards for signal correction. Data were collected using an IC-MS/MS. The error bars indicate ±1 standard deviation.

Fig. 5. Reproducibility of iodide measurement in a milk sample using different internal standards for signal correction. Data were collected using an ICP-MS. The error bars indicate ±1 standard deviation.

there was no consistent trend with time. However, the results varied from 86% to 126% relative to the initial measurement. Using 2-CBS on the other hand, one sees a continuous and consistent increase in the IS-corrected signal with time, indicating that the detection of the 2-CBS derived ion was being increasingly affected by the changing conditions of the MS; by the end of the experiment, the same sample was being measured at twice its original value. Using 129 I as the IS was much better in comparison to the above – the average of all 10 runs was 103.8 ± 4.3% of the first run. To be fair, even 129 I was not a panacea; while significant underestimation was rare, up to 10% overestimation was observed on two occasions. This should be regarded as the ultimate performance limits in absolute accuracy with these types of samples. A similar experiment was conducted with 21 milk and urine samples on an ICP-MS, in between which the same probe milk sample as above was interspersed 10 times. In this case, each sample was measured three times. The results are shown in Fig. 5. Using no IS here actually provides good reproducibility (99.4 ± 0.7% of first run) until the urine samples are run (beginning with run 17). From that point, the signal consistently increases with time. Exactly the opposite is observed when the results are corrected with 72 Ge as an IS. With only milk samples run, the results showed excellent reproducibility (100.6 ± 0.4%) but soon as urine samples were analyzed, the recovery started decreasing consistently. This is mirror image behavior in comparison with the raw signal (no IS) and suggests that the 72 Ge IS signal probably remains invariant – the observed results are largely, if not wholly, due to the 127 I signal actually increasing with time. Overall this behavior was understandable. Considerable care was taken to remove fats, proteins, and all material with MW > 10,000 from the from the milk matrix. The resulting sample is very clean as compared to the raw milk. On the other hand, the urine samples were merely diluted 10-fold. The urine matrix consists primarily of water, salts, urea/ammonia, and several small organic molecules. Urine samples contained significantly more dissolved solids than the processed milk samples. In the case of IC-MS/MS, the chromatography itself provides for significant sample cleanup

(very little of the original accompanying material elutes in the window where the column effluent is directed to the MS to monitor the peak(s) of interest), whereas in the absence of a chromatography step preceding it, direct determination by ICP-MS is affected much more significantly. The use of 115 In as IS results in unpredictable behavior. Prior to running urine samples, the signal of the probe milk sample was more or less reproducible (102.8 ± 3.4%), it increased abruptly to 123.6% of the first run after the first urine sample and decreased slowly thereafter as more urine samples were run. Again, 129 I showed the greatest stability of all ISs used; the performance in this case was nearly flawless. All data ranged between 99.5% and 100% of the first run and the worst relative standard deviation was 0.4%. The average of all 10 runs was 99.7 ± 0.1% of the first sample. This suggests that under normal circumstances iodine measurements will be largely unaffected by the cleanliness of the sample and the mass spectrometer with the use of this IS.

4. Conclusions For almost any real sample analysis by mass spectrometry, reliable quantitative results cannot be obtained without the use of an appropriately chosen IS. Some ISs, such as 2-CBS or 72 Ge, may be useful for iodide/iodine analysis for certain types of samples, but it is clear that neither are appropriate for all types of biological samples. Other ISs, such as 18 O enriched perchlorate and 115 In do not appear to be very reliable for iodide/iodine measurements. By far the best IS to use for iodide analysis is the isotope 129 I. Although it is radioactive, the 16 million year half-life of 129 I leads to very little cause for concern from its radioactivity. A long half-life also means that the concentration of a standard solution does not change appreciably. It is an excellent choice as an IS for iodide/iodine analysis by MS; its low natural abundance helps the cause. It is versatile in its ability to provide matrix corrections for both milk and urine and also in compensating for changing conditions in the mass spectrometer itself.

242

J.V. Dyke et al. / Talanta 79 (2009) 235–242

Acknowledgments This research was supported by the State of Texas Advanced Research Program through 3644-0007-2006. Auxiliary support from the Gerber Foundation and the Centers for Disease Control is also acknowledged. References [1] P.K. Dasgupta, K. Martinelango, W.A. Jackson, T.A. Anderson, K. Tian, R.W. Tock, S. Rajagopalan, Environ. Sci. Technol. 39 (2005) 1569–1575. [2] A.B. Kirk, P.K. Martinelango, K. Tian, A. Dutta, E.E. Smith, P.K. Dasgupta, Environ. Sci. Technol. 39 (2005) 2011–2017. [3] O. Dohan, C. Portulano, C. Basquin, A. Reyna-Neyra, L.M. Amzel, N. Carrasco, Proc. Natl. Acad. Sci. U.S.A. 104 (2007) 20250–20255. [4] N. Tran, L. Valentin-Blasini, B.C. Blount, C.G. McCuistion, M.S. Fenton, E. Gin, A. Salem, J.M. Hershman, Am. J. Physiol. Endocrinol. Metab. 294 (2008) E802–E806. [5] E.N. Pearce, H.R. Bazrafshan, X. He, S. Pino, L.E. Braverman, Thyroid 14 (2004) 327–328. [6] P.K. Dasgupta, A.B. Kirk, J.V. Dyke, S.-I. Ohira, Environ. Sci. Technol. 42 (2008) 8115–8121. [7] J.G. Hollowell, N.W. Staehling, W.H. Hannon, D.W. Flanders, E.W. Gunter, G.F. Maberly, L.E. Braverman, S. Pino, D.T. Miller, P.L. Garbe, D.M. DeLozier, R.J. Jackson, J. Clin. Endocrinol. Metab. 10 (1998) 3401–3408. [8] S.-I. Ohira, A.B. Kirk, J.V. Dyke, P.K. Dasgupta, Environ. Sci. Technol. 42 (2008) 9419–9423. [9] B.S. Hetzel, Lancet 2 (1983) 1126–1127. [10] P. Laurberg, S.B. Nohr, K.M. Pedersen, A.B. Hreidarsson, S. Andersen, I.B. Pedersen, N. Kudsen, H. Perrild, T. Jorgensen, L. Ovesen, Thyroid 10 (2000) 951–963. [11] J.B. Stanbury, A.E. Ermans, P. Bourdoux, C. Todd, E. Oken, R. Tonglet, G. Vidor, L.E. Braverman, G. Medeiros-Neto, Thyroid 8 (1998) 83–100. [12] A.B. Kirk, J.V. Dyke, C.F. Martin, P.K. Dasgupta, Environ. Health Perspect. 115 (2007) 182–186. [13] B.C. Blount, L. Valentin-Blasini, Anal. Chim. Acta 567 (2006) 87–93. [14] E.B. Sandell, I.M. Kolthoff, Mikrochim. Acta 1 (1937) 9–25. [15] J. Benotti, N. Benotti, S. Pino, H. Gardyna, Clin. Chem. 11 (1965) 932–936. [16] S. Pino, S.L. Fang, L.E. Braverman, Clin. Chem. 42 (1996) 239–243. [17] D.A. Skoog, F.J. Holler, T.A. Nieman, Principles of Instrumental Analysis, 5th ed., Harcourt Brace, New York, 1998, pp. 262–269. [18] P. Dulski, Fresenius J. Anal. Chem. 350 (1994) 194–203. [19] K.J.R. Rosman, P.D.P. Taylor, Pure Appl. Chem. 70 (1998) 217–235. [20] R.G. Cooks, A.L. Rockwood, Rapid Commun. Mass Spectrom. 5 (1991) 93. [21] S. Tan, G. Horlick, J. Anal. At. Spectrom. 2 (1987) 745–763. [22] T. Akagi, T. Hirata, A. Masuda, Anal. Sci. 6 (1990) 397–400. [23] D.T. Heitkemper, J.A. Caruso, Appl. Spectrosc. 44 (1990) 228. [24] E.H. Evans, J.J. Giglio, J. Anal. At. Spectrom. 8 (1993) 1–18. [25] S.M. Nelms, ICP Mass Spectrometry Handbook, Blackwell, Boston, 2005, pp. 150–175. [26] J.J. Thompson, R.S. Houk, Appl. Spectrosc. 41 (1987) 801–806. [27] E.C. Kempen, J.S. Brodbelt, Anal. Chem. 72 (2000) 5411–5416. [28] J.C. Schwartz, M.W. Senko, J.E.P. Syka, J. Am. Soc. Mass Spectrom. 13 (2002) 659–669. [29] J.D. Fassett, P.J. Paulsen, Anal. Chem. 61 (1989) 643A–649A.

[30] A. Lasztity, M. Viczian, X. Wang, R.M. Barnes, J. Anal. At. Spectrom. 4 (1989) 761–766. [31] United States Environmental Protection Agency, Determination of Perchlorate in Drinking Water by Liquid Chromatography Electrospray Ionization Mass Spectrometry http://www.epa.gov/safewater/methods/ pdfs/methods/met331 0.pdf Accessed January 6, 2009. [32] P.K. Martinelango, J.L. Anderson, P.K. Dasgupta, D.W. Armstrong, R.S. Al-Horr, R.W. Slingsby, Anal. Chem. 77 (2005) 4829–4835. [33] J.V. Dyke, A.B. Kirk, P.K. Martinelango, P.K. Dasgupta, Anal. Chim. Acta 567 (2006) 73–78. [34] L. Yu, Q. Cheng, J. Canas, L. Valentin-Blasini, B.C. Blount, T. Anderson, Anal. Chim. Acta 567 (2006) 66–72. [35] B.C. Purkayastha, G.R. Martin, Can. J. Chem. 34 (1956) 293–300. [36] D.R. Schink, P.H. Santschi, O. Corapcioglu, P. Sharma, U. Fehn, Earth Planet. Sci. Lett. 135 (1995) 131–138. [37] J.F. Emery, S.A. Reynolds, E.I. Wyatt, G. Gleason, Nucl. Sci. Eng. 48 (1972) 319–323. [38] Argonne National Laboratory, Human Health Fact Sheet, Iodine, http://www.ead.anl.gov/pub/doc/iodine.pdf Accessed January 14, 2008. [39] United States Nuclear Regulatory Commission, October 1, 2007, 10 CFR 30.71 Schedule B, http://www.nrc.gov/reading-rm/doc-collections/ cfr/part030/part030-0071.html. [40] G. Kramer, Environmental and Background Radiation — Food and the Human Body, http://www.hps.org/publicinformation/ate/q322.html. [41] Argonne National Laboratory, Potassium-40, http://www.ead.anl.gov/ pub/doc/potassium.pdf. [42] J.E. Moran, U. Fehn, R.T.D. Teng, Chem. Geol. 152 (1998) 193–203. [43] U. Rao, U. Fehn, Y. Muramatsu, H. McNeil, P. Sharma, D. Elmore, Environ. Sci. Technol. 36 (2002) 1271–1275. [44] G. Krupp, D.C. Aumann, J. Environ. Radioact. 46 (1999) 287–299. [45] T. Eickhorst, A. Seubert, J. Chromatogr. A 1050 (2004) 103–109. [46] P.K. Dasgupta, Y. Liu, J.V. Dyke, Environ. Sci. Technol. 42 (2008) 1315. [47] D.R. Lide, CRC Handbook of Chemistry and Physics, 74th ed., CRC Press, Boca Raton, FL, 1993, p. 14-9. [48] Environmental, Chemistry & Hazardous Materials News, Careers & Resources, Periodic Table of Elements Sorted by 1st Ionization Potential (eV), http://environmentalchemistry.com/yogi/periodic/1stionization.html Accessed January 14, 2009. [49] J.A. Pennington, S.A. Schoen, G.D. Salmon, B. Young, R.D. Johnson, R.W. Marts, J. Food Compos. Anal. 8 (1995) 171–217. [50] G. Radlinger, K.G. Heumann, Anal. Chem. 70 (11) (1998) 2221–2224. [51] M. Haldimann, B. Zimmerli, C. Als, H. Gerber, Clin. Chem. 44 (1998) 817–824. [52] Y. Gelinas, G.V. Iyengar, R.M. Barnes, Fresenius J. Anal. Chem. 5 (1998) 483–488. [53] M. Haldimann, A. Eastgate, B. Zimmerli, Analyst 125 (2000) 1977–1982. [54] M. Haldimann, A. Alt, A. Blanc, K. Blondeau, J. Food Compos. Anal. 18 (2005) 461–471. [55] M. Haldimann, R. Wegmuller, M. Zimmermann, Eur. Food Res. Technol. 218 (2003) 96–98. [56] M. Resano, E. Garcia-Ruiz, L. Moens, F. Vanhaecke, J. Anal. At. Spectrom. 20 (2005) 81–87. [57] W. Bruchert, A. Helfrich, N. Zinn, T. Klimach, M. Breckheimer, H.W. Chen, S.C. Lai, T. Hoffmann, J. Bettmer, Anal. Chem. 79 (2007) 1714–1719. [58] K.L. Caldwell, C.B. Maxwell, A. Makhmudov, S. Pino, L.E. Braverman, R.L. Jones, J.G. Hollowell, Clin. Chem. 49 (2003) 1019–1021. [59] P. Macours, J.C. Aubry, B. Hauquier, J.M. Boeynaems, S. Goldmanc, R. MorenoReyes, J. Trace Elem. Med. Biol. 22 (2008) 162–165. [60] H.A. Roels, J.P. Buchet, J. Environ. Monit. 3 (2001) 67–73.