A comparison of urinary arsenic speciation via direct nebulization and on-line photo-oxidation–hydride generation with IC separation and ICP-MS detection

Abstract

Urinary arsenic speciation provides information on recent arsenic exposure. The literature reported analysis of NIST SRM 2670 Freeze-dried Urine indicates considerable discrepancy in species specific concentration. In this study, two complementary sample introduction pathways, direct nebulization (DN) and hydride generation (HG), were utilized and compared for urinary arsenic speciation via ion chromatography (IC)-ICP-MS. The retention characteristics of arsenobetaine (AsB), arsenite [As(III)], dimethylarsinic acid (DMA), monomethylarsonic acid (MMA), arsenate [As(V)] and Cl⁻ were systematically evaluated with respect to column temperature and the (NH₄)₂CO₃ eluent molarity using the DN method. This characterization indicated that three early eluters [AsB, As(III) and DMA] were best separated at a higher column temperature and lower eluent molarity, whereas MMA, As(V) and Cl⁻ were best separated at a lower column temperature and higher eluent molarity. From these observations, a gradient elution program was developed using 40 and 70 mM (NH₄)₂CO₃ (pH 10.5) at 60 °C. This gradient condition produced satisfactory resolution for all five arsenic species with a Cl⁻ tolerance up to 0.3% w/w. In the membrane hydride generation (HG) configuration, a photo-reactor interface was installed between the column and the HG device to facilitate the detection of non-hydride active arsenic species. Isocratic elution using 40 mM (NH₄)₂CO₃ was adequate in resolving all five arsenic species while the chloride interference was removed by a gas–liquid separator. NIST SRM 2670 Freeze-dried Urine was analyzed using the DN and HG methods and the sum of the arsenical concentrations was 77.7 ± 3.5 and 71.1 ± 2.8 ng mL⁻¹, respectively.

---

Introduction

Humans are exposed to arsenic primarily through drinking water and dietary ingestion. The six most commonly found arsenic species are: arsenobetaine (AsB) or arsenocholine (AsC), dimethylarsinic acid (DMA), monomethylarsonic acid (MMA), arsenite [As(III)] and arsenate [As(V)]. These six arsenicals differ dramatically in their toxicity to humans. Arsenite is fairly toxic owing to its high affinity to the thiol group of some enzymes.^1,2 This association inactivates the enzymatic function and blocks biologically important processes. The toxic mechanism of As(V) lies in its similarity to phosphate. Arsenate uncouples oxidative phosphorylation and leads to the breakdown of energy metabolism.¹ In addition, inorganic arsenic [As(III) and As(V)] is considered to be carcinogenic.^3–7 This toxicity can be reduced via methylation. For instance, MMA and DMA exhibit intermediate toxicity, whereas AsB and AsC, the dominant arsenic species in seafood, are considered non-toxic.^1,8 Once ingested by humans, these six arsenicals are metabolized in very different ways. Inorganic arsenic is converted to MMA and DMA via reductive methylation,⁹ whereas AsB and AsC are excreted unchanged.^2,10,11 Most arsenicals are eventually eliminated in the urine.

Traditional acid digestion methods determine a "total" arsenic concentration which does not differentiate between toxic inorganic arsenic and non-toxic AsB or AsC. Therefore, from a risk assessment standpoint, a "total" arsenic concentration will produce an overestimation of the risk from exposure sources containing AsB/AsC. Arsenic speciation provides species specific concentrations and, for this reason, more accurately reflects the risk. Because most of the arsenic ingested is excreted in the urine, urinary arsenic speciation can provide information on recent exposure and shed some light on recent exposure sources. The analytical challenge in urinary arsenic speciation is to develop methods that will provide excellent sensitivity and selectivity, good matrix tolerance, and chromatographic resolution in combination with a reasonable analysis time. A variety of analytical methods have been developed for speciating arsenic concentrations in human urine.^10–20 Ion chromatography (IC) coupled with inductively coupled plasma mass spectrometric (ICP-MS) detection is one of the premier analytical techniques in arsenic speciation because it combines both the separation strength of IC and the parts per trillion (ppt) detection capability of ICP-MS. The interface between IC and ICP-MS for arsenic speciation is normally accomplished via direct nebulization (DN)^{10,12,16,17,20–27} or on-line hydride generation (HG).^28–30

NIST SRM 2670 Freeze-dried Urine (normal level) is a widely used urine reference material for speciation and "total" arsenic analysis. However, the analytical literature indicates considerable discrepancies in arsenic concentration from both speciation^{12,14,16,17,19,20} and "total" arsenic perspectives. The total recommended (not certified) arsenic concentration reported by NIST is 60 ng mL⁻¹, whereas the literature speciation results range from 60 ± 7 to 109 ± 6 ng mL⁻¹ with very poor species specific agreement. These discrepancies could be a result of interference from urinary matrix, ⁴⁰Ar³⁵Cl (m/z = 75 = As) polyatomic species (with ICP-MS detection), and incomplete chromatographic separation of early eluting species. These factors may lead to the misidentification of individual arsenicals in NIST SRM 2670 Freeze-dried Urine. In addition, it is possible that arsenicals such as As(III) and As(V) undergo a redox transformation during the sample handling and/or on-column oxidation/reduction. Finally, some of the discrepancy may be explained via the traceability of standards such as AsB to a NIST reference standard.

In this work, we investigated urinary arsenic speciation via two complementary methods, direct nebulization IC-ICP-MS (DN) and photo-oxidation–hydride generation–ICP-MS (HG), in an attempt to minimize possible interferences in urinary speciation analysis and produce a more complete understanding of the factors influencing arsenic speciation in urine matrices. DN offers experimental simplicity but suffers from an isobaric interference, ⁴⁰Ar³⁵Cl, in a chloride matrix such as urine, and matrix/mobile phase deposition on the sampling cones. Therefore, certain mobile phases are not instrumentally compatible and this in turn places constraints on the chromatographic development. Another concern is the presence of co-eluting hydrocarbons in a urine matrix which might enhance signal responses for early eluting species and result in higher speciation values.³¹ Hydride generation provides a complementary approach in that it removes the analyte from the hydrocarbon and chloride matrix by converting arsenicals to gaseous hydride species before they are swept into the plasma. The net effect is that the eluent, urine matrix and chloride stay in the liquid phase and are carried to waste. In addition, unwanted matrix induced chromatographic (peak broadening/splitting) and detection (sampling cone deposition) interferences are minimized via HG because of the increased analyte transport which allows for much larger sample dilution factors.

However, some multi-substituted organic arsenic species, such as AsB, AsC, TMAO (trimethylarsenic oxide) and arseno sugars, are either non-hydride active or produce involatile arsine and, therefore, cannot be directly detected by HG-ICP-MS. These species have to be converted into hydride active species such as As(V) prior to HG. This conversion can be achieved via on-line photo-oxidation.^15,32–36 The photo-oxidation interface not only allows the simultaneous detection of non-hydride active species by HG-ICP-MS, but can also serve as an on–off switch to distinguish hydride active species (normally more toxic) from non-hydride active species (normally much less toxic or non-toxic).

The objective of this work was to (a) optimize a chromatographic separation for DN which not only resolves Cl⁻ (detected as ⁴⁰Ar³⁵Cl) from all the analytes, but also offers baseline separation among the early eluting arsenicals [AsB, As(III) and DMA], which should increase the analytical confidence of the DN approach in the analysis of NIST SRM 2670 Freeze-dried Urine; (b) install a photo-oxidation interface prior to HG so that non-hydride active arsenicals can be determined by HG-ICP-MS, and develop a corresponding chromatographic scheme based on HG detection, which should provide a secondary means of analyzing NIST SRM 2670 Freeze-dried Urine which minimizes matrix induced interferences; and (c) evaluate these two modes of sample introduction for the arsenic speciation in NIST SRM 2670 Freeze-dried Urine sample. This research should assist in the development of a much needed "literature" accepted speciation concentration for the NIST SRM 2670 Freeze-dried Urine.

Experimental

Instrumentation

Important experimental parameters for IC, photo-oxidation, HG and ICP-MS are summarized in Table 1. A schematic diagram is shown in Fig. 1.

Fig. 1 Instrumentation and fluid flow diagram for IC-photo-oxidation-HG-ICP-MS.

Table 1 Paramenters for DN IC-ICP-MS and IC-photo-oxidation-HG-ICP-MS

Ion Chromatography—
Anion exchange column	ION-120, 120 mm × 4.6 mm id
Guard column	ION-120, 20 mm × 4.6 mm id
Column temperature	60 °C
Injection volume	50 µL
Eluent	DN: 40 mM (E1)/70 mM (E2) (NH4)2CO3, pH 10.5
Elution mode	gradient program (DN)—
	E1 (100%, 1.0 mL min^−1, 0.3 min), E1 → E2 (0.1 min), E2 (100%, 1.0 mL min^−1, 5.5 min),E2 → E1 (0.1 min), E1 (100%, 1.0 mL min^−1, 3.8 min), E1 (100%, 1.5 mL min^−1, 6 min),E1 (100%, 1.0 mL min^−1, 1 min)
	Isocratic program (HG)—
	40 mL (NH4)2CO3, pH 10.5, 1.0 mL min^−1
Photo-oxidation—
UV lamp wavelength	254 nm, 8 W
Reaction coil (braided)	5 m × 0.56 mm id
6% K2S2O8–4% NaON flow rate	0.5 mL min^−1
Hydride generation—
30% HCl flow rate	0.6 mL min^−1
1.5% NaBH4–0.1 M NaOH flow rate	0.5 mL min^−1
Sweep gas flow rate	0.2 L min^−1
Carrier gas flow rae	1.03 L min^−1
ICP-MS—
Acquisiton mode	Time resolved analysis
Single ion monitoring	m/z 75
Integration time	2.0 s

---

The IC and ICP-MS systems were common to both the DN and the HG interface. The IC was a Dionex GPM-2 gradient pump and the analytical column was an ION-120 (120 × 4.6 mm id) (InterAction Chromatography, San Jose, CA, USA) anion exchange column with a 20 mm guard column. Both DN and HG separation were carried out at 60 °C. The column and a 48 in stainless steel pre-heating coil were heated in a recirculated water-bath (NESLAB, RTE-100). The (NH₄)₂CO₃ eluent was pH adjusted with concentrated ammonium hydroxide solution. A gradient elution mode consisting of two eluents [40 mM (E1) and 70 mM (E2) (NH₄)₂CO₃] was used with the DN method. The gradient scheme is shown in Table 1. The total analysis time including column equilibration was about 17 min. Since Cl⁻ interference is eliminated by hydride generation, isocratic elution using 40 mM (NH₄)₂CO₃ at 1 mL min⁻¹ with a 60 °C column temperature was used for the HG method. This offers baseline separation among the three early eluters and also a reasonable analysis time (12 min). Because the LC was not electrically interfaced with the ICP-MS, a simultaneous post-column injection of AsB standard was used as a drift standard (labeled as t = 0 STD in the figures) and served as a time zero marker in the chromatogram. In the case of HG, the use of AsB as the drift standard also served as a constant check on the photo-reactor because AsB is not hydride active without the pre-oxidation step.

A Hewlett-Packard (Avondale, PA, USA) 4500 Series bench top ICP-MS system was used. The rf power was set at 1200 W. The plasma gas and auxiliary gas flow rates were 15 and 1.0 L min⁻¹, respectively. Arsenic species were identified by chromatographic retention times. Chromatographic integration was performed using the software provided with the HP 4500 instrument. For the DN method, the instrument was tuned with 10 ppb As(III) solution; for the HG method, it was tuned with 0.1 ppb As(III)–10% HCl and 1% NaBH₄–0.1 M NaOH.

The IC effluent was mixed with K₂S₂O₈ solution at a three-way PTFE manifold mixer and passed through the braided reaction coil under UV irradiation in the PHRED photo-reactor (see Fig. 1). The percentage of K₂S₂O₈ and NaOH used in the reaction was optimized based on the responses of 6 ppb AsB at 25% HCl and 1% NaBH₄ (results not shown). The post-column photochemical reactor PHRED³⁷ (Fig. 1) was purchased from Aura Industries (Staten Island, NY, USA). The reactor compartment was equipped with an 8 W UV lamp (254 nm), and a polished stainless steel support plate with holes at the bottom which allows PTFE tubing of the photo-reaction coil to extend in/out of the reactor. For better irradiation efficiency, the original flat-shaped PHRED braided reaction coil (designed to fit on the bottom support plate under the UV lamp) was replaced with 5 m of thin-walled microbore PTFE tubing (i.d. 0.56 mm, wall thickness 0.25 mm) (Cole-Parmer, Vernon Hills, IL, USA), which was braided around the UV lamp as described in the literature.^38,39 This braided configuration was used throughout this work in an effort to minimize peak broadening which resulted from the additional interfaces. The on-line photo-oxidation efficiency depends on UV light transmission through the reactor coil (which is related to coil material, wall thickness and distance to the UV light source) and photolysis time. This braided configuration increased the AsB signal intensity by 90% compared with the PHRED flat braided coil (data not shown). Excessively long reactor coils, especially knitted ones, aggravated the system back-pressure (causing the peristaltic pump to produce inconsistent flow), which caused severe plasma pulsing and for this reason were avoided. The 5 m reactor coil was found to allow enough time for most organoarsenicals to break down without sacrificing the system stability. Because a cooling device for the UV lamp was not provided with the PHRED, a warm-up period of 15–20 min was applied (UV light, IC eluent and all the reagent flows were switched on) so that thermal equilibrium was reached at the photo-reaction coil before data were collected.

The HG interface is comprised of two mixing tees (shown in Fig. 1), a 750 µL knitted hydride reaction coil (Dionex, Sunnyvale, CA, USA) and a gas permeable membrane which serves as a gas–liquid separator. The membrane HG unit was constructed similarly to that described by Magnuson et al.^29,40 The membrane utilized was expanded polytetrafluoroethylene (ePTFE) microporous tubing purchased from International Polymer Engineering (Tempe, AZ, USA). The effluent from the photo-reactor was acidified with 30% HCl at a three-way PTFE manifold, then mixed with 1.5% NaBH₄–0.1 M NaOH solution at the second three-way PTFE manifold. This set of parameters were determined to be the optimum for the photo-oxidation membrane HG system. The HG reaction was completed as the mixed flow passed through the knitted hydride reaction coil, with a residence time of ca. 20 s. Arsine and other gases/vapors were separated from the liquid waste at the membrane gas–liquid separator and swept to the plasma by using a sweep gas (0.2 L min⁻¹) and a carrier gas (1.03 L min⁻¹, see Fig. 1). To maintain the stability of the plasma, measures should be taken to reduce the gas pressure inside the membrane gas–liquid separator and minimize the pulsation of peristaltic pumps which were used for reagent delivery (smaller id peristaltic pump tubings operated at higher rpm produce less pulsation).

Reagents

All reagents and arsenic solutions were prepared in de-ionized (Millipore, Bedford, MA, USA), distilled (DDI) water in fresh Nalgene polyethylene or PTFE bottles. All dilutions were made on a mass basis. All sample preparations were handled in a Class 100 clean air hood to avoid contamination. Arsenic standards and post-column reaction solutions were prepared daily.

The arsenite [As(III)] and arsenate [As(V)] standards were obtained from Spex Certiprep (Metuchen, NJ, USA). Dimethylarsinic acid (DMA) and disodium methylarsenate (MMA) were obtained from Chem Services (West Chester, PA, USA), and arsenobetaine (AsB) was synthesized by the Department of Chemistry at University of British Columbia (Vancouver, Canada). All standard materials were verified for total arsenic concentration against NIST SRM 1643c using ICP-AES and ICP-MS.

(NH₄)₂CO₃ and K₂S₂O₈ (ACS reagent grade) were purchased from Aldrich (Milwaukee, WI, USA). Trace metal grade ammonium hydroxide solution (Fisher Scientific, Pittsburgh, PA, USA) was used for eluent pH adjustment. Optima ammonium hydroxide solution (As < 100 ppt) (Fisher Scientific) trace metal grade HCl (Fisher Scientific), and NaBH₄ (97 + %) (Johnson Matthey, Ward Hill, MA, USA) were used for HG. NIST SRM 2670 Freeze-dried Urine (normal level) (NIST, Gaithersburg, MD, USA) was used as a reference standard and analyzed by both DN and HG. The sample was first reconstituted with 20 mL of DDI water, then further diluted with DDI water to the desired concentration. The recommended total As concentration in reconstituted NIST SRM 2670 Freeze-dried Urine was 60 ng mL⁻¹. The reconstituted urine sample was stored at 4 °C and used within 1 week. The urine sample was filtered through a 0.45 µm disk prior to injection.

Results and discussion

Optimization of chromatography for sample introduction via direct nebulization

Table 2 provides a comparison of some literature results for arsenic speciation of NIST SRM 2670 Freeze-dried Urine (normal level). As stated earlier, some of these discrepancies may be partially explained by matrix-induced incomplete chromatographic resolution of early eluting species and/or misidentifying and quantifying Cl⁻ as MMA or As(V). Tsalev et al.¹⁵ and Le and Ma¹⁹ utilized HG detection coupled with AAS or AFS, which eliminates the Cl⁻ interference. Ritsema et al.¹⁶ separated Cl⁻ from the arsenic species chromatographically using IC-ICP-MS. Although ArCl interference was also addressed by Heitkemper et al.,¹² Goessler et al.¹⁴ and Zheng et al., ²⁰ interference correction or subtraction was not performed. ArCl is probably the main reason why an unusually high As(III) concentration was reported by Heitkemper et al.¹²

Table 2 Comparison of concentrations of arsenic species in SRM 2670 Freeze-dried Urine (normal level)^a

	AsB	DMA	As (III)	MMA	As(V)	Sum
a Two-sided significance level for paired sample t-test using α = 0.05: AsB, 0.5583; DMA, 0.0260; MMA, 0.1657.b n = 0.05, μg L^−1 ± 2σ.c N.f., not found, or below detection limit.d N.d., not determined.
DN^b	17.8 ± 1.1	48.2 ± 2.3	N.f.^c	9.8 ± 0.3	1.3 ± 0.2	77.1 ± 3.5
HG^b	18.5 ± 4.2	43.6 ± 3.8	N.f.	9.0 ± 2.1	N.f.	71.1 ± 2.8
Ref. 19	N.f.	49 ± 5	N.f.	11 ± 3	N.f.	60 ± 7
Ref. 16	16.0 ± 1.1	47.5 ± 1.6	<0.4	9.4 ± 0.9	6.9 ± 0.8	79.8 ± 4.4
Ref. 20	15.5 ± 1.0	52.1 ± 2.2	<2.5	16.0 ± 0.6	N.d.^d	84.4 ± 3.3
Ref. 14	21.2 ± 3.7	48.2 ± 2.4	15.0 ± 4.5	9.5 ± 3.0	2.9 ± 0.7	96.8 ± 14.3
Ref. 17	35.6 ± 7.8	48.3 ± 2.8	<4	18.4 ± 4.4	<4	102.3
Ref. 12	N.d.	45.5 ± 3.5	52.6 ± 4.1	9.9 ± 1.4	0.8 ± 0.5	109 ± 6

---

In order to minimize these potential errors, chromatographic optimization was conducted via the DN approach prior to the analysis of an NIST SRM 2670 Freeze-dried Urine sample. The retention characteristics of arsenic species on an anion exchange column are pH dependent because of the pK_a differences between the arsenic species. Arsenobetaine (pK_a = 4.7) is either a cation or a zwitterion and elutes in or near the void volume. Dimethylarsinic acid (pK_a = 6.2) and As(III) (pK_a = 9.2) carry either a partial or unit apparent charge, hence both are weakly retained by the column and normally elute immediately after AsB. To achieve satisfactory separation of these three species, a relatively weak eluent with the appropriate pH adjustment is required. In addition, if DN is utilized, eluent compatibility with ICP-MS detection and the chromatographic resolution of Cl⁻ is required. Chloride (detected as ⁴⁰Ar³⁵Cl) normally appears as a broad peak and produces analyte peak broadening because of its relatively high concentration. Therefore, the ultimate chromatographic goal is to design a separation scheme with good matrix tolerance that preserves the chromatographic resolution of the early eluters [AsB, As(III) and DMA], and provides adequate separation of the Cl⁻ interference from all the target arsenicals. This often leads to a compromise between resolution and analysis time. Larsen et al.²² reported that the ION-120 anion exchange column produces near baseline resolution of the three early eluters [As(III), AsB and DMA] using 100 mM (NH₄)₂CO₃ as eluent at pH 10.5 at 50 °C. Utilizing these conditions, Cl⁻ elutes 100 s after As(V) and its intensity is negligible with injections of up to 1% NaCl. These separation conditions are problematic when elevated concentrations of AsB are experienced because the resolution between AsB and DMA is inadequate. One approach to improve AsB–DMA separation is by controlling the chromatographic temperature. Fig. 2(a) and (b) illustrate the temperature effects on early eluting peaks with a 30 and 70 mM (NH₄)₂CO₃ mobile phase, respectively. The shift in retention times is most pronounced for the weaker eluent [30 mM (NH₄)₂CO₃, Fig. 2(a)]. In both Fig. 2(a) and (b), DMA and As(III) shift to longer retention times whereas AsB is unaffected by column temperature. This shift in retention time is caused by the charge difference (neutral vs. −1). In general, the higher the charge, the more the anion will be retained when the column temperature is increased. On the other hand, with a fixed column temperature [comparing the two solid or the two dotted lines in Fig. 2(a) and (b)], DMA and As(III) elute much faster with a stronger eluent [70 mM (NH₄)₂CO₃] and the resolution among the three species decreases. Based on these observations, a higher column temperature and a lower eluent molarity should be applied to achieve the best separation among AsB, DMA and As(III). The complete separation of these early eluting peaks should minimize their potential misidentification.

Fig. 2 Effect of column temperature and eluent molarity on the separation resolution of AsB, DMA and As(III). (a) 30 mM (NH₄)₂CO₃ (baseline is elevated for better viewing); (b) 70 mM (NH₄)₂CO₃. Solid lines, 50 °C; dashed lines, 80 °C. AsB, 1 ppb; DMA, 0.4 ppb; As(III) and MMA 0.6 ppb.

The second chromatographic concern is to resolve the Cl⁻ interference from all the arsenicals. On an ION-120 column, with (NH₄)₂CO₃ as the eluent, Cl⁻ tends to co-elute with MMA and As(V) and this could be a source of misidentification. To sort out the optimum chromatographic conditions, the retention times of all five arsenicals and Cl⁻ were compared at different eluent molarities and column temperatures. Five eluent concentrations [30, 40, 50, 60 and 70 mM (NH₄)₂CO₃] and four column temperatures (50, 60, 70 and 80 °C) were selected for study. For the purpose of identifying each arsenical without interference of Cl⁻, Cl⁻ was injected alone as 1% NaCl, whereas all other arsenicals were injected in one sample. Fig. 3 is a 3-D plot of the retention time profiles of As(V) (black) and Cl⁻ (light gray) as a function of (NH₄)₂CO₃ concentration and column temperature. The dashed line surface, which corresponds to a shorter retention time, is graphically below the solid line surface. The 3-D graph shows that the retention of As(V) is more affected by eluent molarity and column temperature than is that of Cl⁻, presumably owing to the charge difference. The chromatographic resolution between As(V) and Cl⁻ is maximized at points in Fig. 3 where the two surfaces show a maximum separation [i.e., 30 mM (NH₄)₂CO₃, 80 °C; 70 mM (NH₄)₂CO₃, 50 °C]. The chromatograms associated with these two points are shown as in Fig. 4 in order to ensure data clarity. Chromatographic conditions of 30 mM (NH₄)₂CO₃ at 80 °C produce good chromatographic resolution between MMA, Cl⁻ and As(V) [see Fig. 4(a)]. However, the extremely broad Cl⁻ peak begins to cause integration problems at higher chloride concentrations. In addition, the relatively weak eluent will lead to a separation with relatively poor (ionic) matrix tolerance. Chromatographic conditions of 70 mM (NH₄)₂CO₃ at 50 °C produce excellent resolution between MMA, As(V) and Cl⁻, as shown in Fig. 4(b). Under these conditions, the analysis time is much shorter and Cl⁻ is downstream of all the five arsenic species. Therefore, the peak tailing of the chloride does not affect the separation and/or data processing. The observations inFig. 3 and 4 also suggest that the retention behaviors of As(V) and Cl⁻ are different on an ION-120 anion exchange column, with As(V) being more sensitive to the changes in eluent concentration and column temperature. This difference makes 70 mM (NH₄)₂CO₃ at 50 °C the best separation conditions for these three late eluting species.

Fig. 3 Retention times of As(V) and Cl⁻ as a function of column temperature and eluent molarity.

Fig. 4 Retention times of MMA, As(V) and Cl⁻ at (a) 30 mM (NH₄)₂CO₃, 80 °C and (b) 70 mM (NH₄)₂CO₃, 50 °C. MMA, 0.6 ppb; As(V), 1 ppb; Cl⁻, 0.1%.

The chromatographic studies reported in Fig. 2 indicate that the best chromatographic resolution between AsB, DMA and As(III) is obtained at a low eluent concentration and high column temperature [e.g., 30 mM (NH₄)₂CO₃, 80 °C). These conditions are contrary to the high eluent concentration and low column temperature [e.g., 70 mM (NH₄)₂CO₃, 50 °C] required for the separation of Cl⁻ and As(V). Isocratic conditions of 70 mM (NH₄)₂CO₃ and 50 °C can be used, but early eluting baseline resolution can only be maintained with an injection of 1 ppb AsB, 0.4 ppb DMA and 0.6 ppb As(III) or less. This limited resolution could be further reduced when a heavy matrix such as urine is present in the sample. Hence a gradient separation was developed. Fig. 5 is a chromatogram which indicates the separation achieved after gradient optimization. The column temperature selected was 60 °C and the eluent program is made up of a combination of 40 mM (NH₄)₂CO₃ and 70 mM (NH₄)₂CO₃ and is outlined in Table 1. The 40 mM (NH₄)₂CO₃ eluent at 60 °C gives a good separation among the three early eluters, whereas 70 mM (NH₄)₂CO₃ facilitates the elution of MMA and As(V) and also ensures that Cl⁻ is separated from the rest of the arsenicals. The analysis can be completed within 17 min including column re-equilibration. Up to 0.3% Cl⁻ can be tolerated without loss of the resolution among the early eluting species while maintaining the resolution of As(V) and Cl⁻.

Fig. 5 Chromatogram of five arsenic species in a 0.1% Cl⁻ matrix resulting from gradient elution with ICP-MS detection. AsB, 2 ppb; DMA, 1 ppb; As(III), DMA and As(V), 5 ppb.

Urinary analysis by direct nebulization

Fig. 6 is a chromatogram of NIST SRM 2670 Freeze-dried Urine (normal level). The Freeze-dried Urine was reconstituted and diluted by a factor of 10 prior to analysis. This chromatogram indicates the presence of AsB, DMA, MMA and As(V). Arsenite was not detected and only a trace of As(V) was found. The presence of Cl⁻ was confirmed by the simultaneous detection of a peak at the same retention time at m/z 51(³⁵Cl¹⁶O⁺). The urine matrix did not produce significant peak retention time shifts. In addition, no significant reduction in instrument sensitivity was observed during the 2 d when the urine samples were analyzed. In Fig. 6, the peak labeled with an asterisk is an unidentified arsenic species which constitutes 2% of the total peak area for all species. This sample was fortified with MMA and the unidentified peak did not change in intensity. This peak was investigated further via the HG method (Fig. 7). The speciation results for NIST SRM 2670 utilizing the DN are compared in Table 2 with those obtained by the HG method and those cited in the literature.

Fig. 6 Chromatogram of NIST SRM 2670 Freeze-dried Urine (normal level, diluted 10-fold) by direct nebulization IC-ICP-MS. *, Un-identified species.

Fig. 7 Chromatogram of NIST SRM 2670 Freeze-dried Urine (normal level, diluted 50-fold) by IC-photo-oxidation-HG-ICP-MS. Inset: undiluted urine. *, Unidentified species.

Urinary arsenic determination by IC–photo-oxidation–hydride generation–ICP-MS

Hydride generation converts arsenicals into gaseous arsine which is removed from the liquid phase at the membrane gas–liquid separator before analysis by ICP-MS. It eliminates Cl⁻ as an isobaric interference and dramatically simplifies chromatography. An isocratic elution, using only 40 mM (NH₄)₂CO₃ at 60 °C, was found adequate to separate all five arsenic species, which in turn shortened the analysis time. Because of the enhanced sensitivity offered by HG, samples can be prepared with a much higher dilution factor, which significantly minimizes the urine matrix effect on the chromatography. In addition, HG offers an added capability in distinguishing complex arsenicals (non-hydride active, such as AsB and arsenosugars) from simple arsenicals (hydride active, such as MMA and DMA), which could provide some clues in identifying arsenic species. The design of the photo-reactor reaction coil should maximize the oxidation efficiency while minimizing the chromatographic peak broadening and the system operating back-pressure. If gas producing eluents such as (NH₄)₂CO₃ cannot be avoided, their concentration should be kept to the minimum, because CO₂ competes with AsH₃ to cross the membrane and increases the pressure inside the membrane, which contributes to the plasma pulsing.

Fig. 7 is a chromatogram of a 50-fold dilution of NIST Freeze-dried Urine (normal level). Broader peaks (relative to DN) are produced by the two additional mixing manifolds used in the HG interface. The increased baseline noise is presumably due to the formation of CO₂ upon acidification of the eluent just prior to the membrane HG. The CO₂ causes added gas pressure in the membrane and produces a pressure pulse within the sample introduction channel of the plasma. This back-pressure problem was not observed in previously reported research using a phosphate buffer.^29,40

A comparison ofFig. 6 and 7 shows varied relative intensity of AsB vs. DMA which is a result of the difference in their oxidation efficiency in the photo-reactor. The unknown species labeled with an asterisk in Fig. 6 was not found in a diluted (50-fold) urine sample using HG. An undiluted sample was analyzed by HG and the peak of interest is shown as the inset in Fig. 7. As(V) was also identified in this undiluted sample. A complete chromatogram was not collected for the undiluted sample because of the extremely high count rates associated with DMA, etc. This unknown peak is hydride active without the use of photo-oxidation, which indicates that it is not a highly substituted arsenic species. The presence of this unidentified species in both DN and HG methods confirms the presence of an As atom within its structure. The fact that it eluted on the anion exchange column close to MMA indicates that it may have a −2 charge similar to MMA at pH 10.5. Zheng etal.²⁰ also reported an unknown species in NIST SRM 2670 Freeze-dried Urine on a Hamilton PRP X-100 column using a pH 2.91 tartaric acid buffer, and this species was suspected to be trimethylarsenic oxide (TMAO), which should elute in the void volume given our chromatographic separation.

The total arsenic value for NIST SRM 2670 Freeze-dried Urine (normal level) is not certified but its recommended value is 60 ng mL⁻¹. Table 2 shows the speciation results for NIST SRM 2670 Freeze-dried Urine (normal level) obtained by both DN and HG. Literature results are also listed for comparison. The sums of the species obtained by DN and HG are 77.1 ± 1.1 and 71.1 ± 2.8 µgL⁻¹, respectively. The unknown species found by both DN and HG was not included in the calculation. t-Tests were performed on both species specific and the summation values in the DN and HG data sets using α = 0.05. A species specific comparison indicates that AsB concentration determined via HG and DN are within 4% (0.7 ppb). The t-test results show that the difference between two AsB values are insignificant. This indicates that the photo-oxidation HG is an effective means of determining AsB. The MMA concentrations determined are within 8% (0.8 ppb) and this difference is also insignificant according to the t-test. The largest and most significant difference (t = 0.026) between the two techniques was observed for DMA (a 4.6 ppb difference or 11%). The reason for this difference is currently unclear. In the DN case, it could be associated with the co-eluting urine matrix component which might affect the signal response. For HG, the urine matrix effect was minimized via the high dilution factor and hydride conversion. Although the DMA concentrations are statistically different between the two techniques (because of their relatively small variances), the comparison with the literature reported concentrations indicates an overall mean of 47.8 ppb. Therefore, DMA is the only arsenical which demonstrates good across-literature agreement. The sums of arsenic concentrations from both methods are also significantly different, mainly owing to the difference in DMA concentrations.

A comparison between the literature reported concentrations for AsB (15.5–35.6 ng mL⁻¹), As(III) (0–52.6 ng mL⁻¹), MMA (9.0–18.4 ng mL⁻¹) and As(V) (0–6.9 ng mL⁻¹) indicates poor agreement. The poor agreement between AsB and As(III) may be due to the tendency of As(III) and AsB to co-elute on most anion exchange columns over a large pH range which increases the chances of misidentifying AsB as As(III) or vice versa. Some of the smaller differences observed for AsB and As(III) may be induced by ionic strength based peak broadening, which in turn causes poor peak integration. The poor agreement for MMA and As(V) could be induced by these species chromatographically co-eluting with chloride. In addition, the use of extremely weak eluents in gradient separations [to facilitate AsB and As(III) separation] can allow the As(V) in this weak eluent to preconcentrate on the head of the column. The As(V) from the weak eluent is then mobilized when the gradient is initiated. The net effect is a false As(V) peak produced by contamination in the weak eluent. Finally, it may be difficult to ensure that As(III) is not being converted into As(V) by on-column oxidation. Therefore, to obtain an accurate arsenic speciation analysis, a good chromatographic scheme which offers baseline separation between AsB and As(III) and good matrix tolerance is essential. Again, the chromatographic optimization for the DN and the use of HG were investigated in an attempt to minimize some of these potential problems.

The sum of the species reported in the last column indicates that DN, HG and all but one literature value are biased high relative to the recommended value. The value reported by Le and Ma¹⁹ shows complete agreement at 60 ± 7 ng mL⁻¹. However, AsB was not found and included in the 60 ± 7 ng mL⁻¹ concentration and the reported DMA (49 ± 5 ng mL⁻¹) and MMA (11 ± 3 ng mL⁻¹) concentrations are in good agreement with other values reported in Table 2. These differences in reported arsenic speciation values in NIST SRM 2670 Freeze-dried Urine samples further indicate the imminent need for a certified reference material for arsenic species in a urine matrix.

Comparison of direct nebulization and hydride generation approaches

DN and HG differ in sample introduction, sensitivity, selectivity and experimental complexity. HG offers near 100% analyte transport efficiency compared with 1–2% for DN and for this reason increased sensitivity is a benefit of HG. In addition, the eluent and other matrix components are separated from the analytes and carried to the waste via HG. This eliminates the need to resolve Cl⁻ chromatographically from other arsenicals and allows for a wider choice of eluent. A gradient elution program had to be utilized to separate all arsenicals and chloride with DN, whereas isocratic conditions were found to be adequate for HG. As a result, a shorter analysis time was achieved for the HG method. HG also provides additional selectivity by distinguishing hydride-active species, which are normally more toxic [i.e., As(III), As(V), DMA and MMA], from non-hydride active species, which usually are less or non-toxic, (i.e., arseno sugars, AsB and AsC). Photo-oxidation can serve as an on–off switch for the non-hydride forming arsenicals. This is helpful if co-elution of a complex arsenical and a simple arsenical is suspected. The improved sensitivity for HG also allows more dilute samples to be analyzed, which further reduces the matrix induced chromatographic peak retention time shifts and peak splitting. However, compared with DN, more experimental parameters are involved in the photo-oxidation–hydride generation method, which adds more instrumental complexity. The greatest limitation for the HG method is probably the background or reagent contamination, especially from the NaBH₄. This significantly limits the linear range for quantitative applications. Methods to purify NaBH₄ further are needed in this regard.

Conclusion

Two complementary methods, direct nebulization (DN) and hydride generation (HG), were investigated for arsenic speciation in urine. These two methods differ in sample dilution factor, sample introduction and separation conditions. For DN, resolution among the early eluting species [AsB, DMA, As(III)] is improved at higher column temperatures and lower eluent molarity, whereas lower column temperatures and higher eluent molarity are needed to resolve Cl⁻ from MMA and As(V). These opposing resolution requirements made a gradient elution necessary for DN. A maximum of 0.3% Cl⁻ can be tolerated without loss of the resolution among the early eluting species while maintaining the resolution of As(V) and Cl⁻. HG detection reduces the required resolution by eliminating Cl⁻ interference and thereby allows for an isocratic separation. The HG interface is extremely sensitive, requiring a 50-fold dilution of the urine matrix prior to analysis. This dilution minimizes peak broadening induced by high ionic strength matrices, but mobile phase and reagent purity can be problematic for HG. Non-hydride active arsenicals such as AsB can be detected by installing a photo-oxidation interface between IC and HG, thereby facilitating a direct comparison with the DN approach. Increased photo-oxidation efficiency can be achieved by using thin-walled microbore reaction tubing braided around the UV lamp in the photoreactor. A braided photo-reaction coil also minimizes peak broadening. In addition, an unknown arsenic species was found in NIST SRM 2670 Freeze-dried Urine by both DN and HG.

NIST SRM 2670 Freeze-dried Urine (normal level) was analyzed by both DN and HG. The paired sample t-tests indicated that the sums of the individual arsenic concentrations obtained by these two methods are significantly different, mainly owing to the significant difference in DMA concentrations. These statistically significant differences between the two techniques are relatively small in comparison with the variations reported in the literature.

References

1. P. J. Craig, Organometallic Compounds in the Environment, Wiley, New York, 1986..
2. H. B. F. Dixon, Adv. Inorg. Chem., 1999, 44, 191.
3. World Health Organization, Environmental Health Criteria, Vol. 18: Arsenic, WHO, Geneva, 1981..
4. A. Furst, in Arsenic: Industrial, Biomedical, Environmental Perspectives, ed. W. H. Lederer and R. J. Fensterheim, Van Nostrand Reinhold, New York, 1983, pp. 151–165..
5. V. F. Simmon, in Arsenic: Industrial, Biomedical, Environmental Perspectives, ed. W. H. Lederer and R. J. Fensterheim, Van Nostrand Reinhold, New York, 1983, pp. 166–172..
6. I. Harding-Barlow, in Arsenic: Industrial, Biomedical, Environmental Perspectives, ed. W. H. Lederer and R. J. Fensterheim, Van Nostrand Reinhold, New York, 1983, pp. 203–209..
7. Arsenic: Exposure and Health Effects, ed. C. O. Abernathy, R. L. Calderon and W. R. Cappell, Chapman and Hall, London, 1997..
8. W. R. Cullen and K. J. Reimer, Chem. Rev., 1989, 89, 713.
9. M. Stoeppler and M. Vahter, in Trace Elements in Biological Specimens, ed. R. F. M. Herber and M. Stoeppler, Elsevier, Amsterdam, 1994, pp. 291–320..
10. X. C. Le, W. R. Cullen and K. J. Reimer, Clin. Chem., 1994, 40, 617.
11. T. R. Irvin and K. D. Irgolic, Appl. Organomet. Chem., 1988, 2, 509.
12. D. Heitkemper, J. Creed and J. Caruso, J. Anal. At. Spectrom., 1989, 4, 279.
13. E. Crecelius and J. Yager, Environ. Health Perspect., 1997, 105, 650.
14. W. Goessler, D. Kuehnelt and K. J. Igrolic, in Arsenic: Exposure and Health Effects, ed. C. O. Abernathy, R. L. Calderon and W. R. Cappell, Chapman and Hall, London, 1997, pp. 33–44..
15. D. L. Tsalev, M. Perling and B. Welz, Analyst, 1998, 123, 1703.
16. R. Ritsema, L. Dukan, T. R. I. Navarro, W. Van Leeuwen, N. Oliveira, P. Wolfs and E. Lebret, Appl. Organomet. Chem., 1998, 12, 591.
17. M. Moldovan, M. M. Gómez, M. A. Palacios and C. Camara, Microchem. J., 1998, 59, 89.
18. J. C. Ng, D. Hohnson, P. Imray, B. Chiswell and M. R. Moore, Analyst, 1998, 123, 929.
19. X. C. Le and M. Ma, Anal. Chem., 1998, 70, 1926.
20. J. Zheng, W. Kosmus, F. Pichler-Semmelrock and M. Köck, J. Trace Elements Med. Biol., 1999, 13, 150.
21. S. H. Hansen, E. H. Larsen, G. Prital and C. Cornett, J. Anal. At. Spectrom., 1992, 7, 629.
22. E. H. Larsen, G. Pritzl and S. H. Hansen, J. Anal. At. Spectrom., 1993, 8, 557.
23. C. Demesmay, M. Olle and M. Porthault, Fresenius' J. Anal. Chem., 1994, 348, 205.
24. H. Ding, J. Wang, J. G. Dorsey and J. A. Caruso, J. Chromatogr., A, 1995, 694, 425.
25. P. Teräsahde, M. Pantsar-Kallio and P. K. G. Manninen, J. Chromatogr., A, 1996, 750, 83.
26. X. C. Le and M. Ma, J. Chromatogr., A, 1997, 764, 55.
27. S. Londesborough, J. Mattusch and R. Wennrich, Fresenius' J. Anal. Chem., 1999, 363, 577.
28. W. C. Story, J. A. Caruso, D. T. Heitkemper and L. Perkins, J. Chromatogr. Sci., 1992, 30, 992.
29. M. L. Magnuson, J. T. Creed and C. A. Brockhoff, J. Anal. At. Spectrom., 1996, 11, 893.
30. T. Dagnac, A. Padró, R. Rubio and G. Rauret, Talanta, 1999, 48, 763.
31. E. H. Larsen and S. Stürup, J. Anal. At. Spectrom., 1994, 9, 1099.
32. R. H. Atallah and D. A. Kalman, Talanta, 1991, 38, 167.
33. A. G. Howard and L. E. Hunt, Anal. Chem., 1993, 65, 2995.
34. R. Rubio, J. Albertí, A. Padró and G. Rauret, Trends Anal. Chem., 1995, 14, 274.
35. J. Albertí, R. Rubio and G. Rauret, Fresenius' J. Anal. Chem., 1995, 351, 415.
36. X. Zhang, R. Cornelis, J. De Kimpe and L. Mees, Anal. Chim. Acta, 1996, 319, 177.
37. H. Joshua, presented at the 36th Annual Eastern Analytical Symposium and Exposition, Somerset, NJ, November 1997..
38. R. G. Luchtefeld, J. Chromatogr. Sci., 1985, 23, 516.
39. J. R. Poulsen, K. S. Birks, M. S. Gandelman and J. W. Birks, Chromatographia, 1986, 22, 231.
40. M. L. Magnuson, J. T. Creed and C. A. Brockhoff, J. Anal. At. Spectrom., 1997, 12, 689.

---

Footnote

† National Research Council Postdoctoral Fellow.

---