Low-level determination of six arsenic species in urine by High Performance Liquid Chromatography-Inductively Coupled Plasma-Mass Spectrometry (HPLC-ICP-MS)

Abstract

Many methods that have been used to speciate arsenic metabolites in urine are unable to adequately resolve the chromatographic peaks for arsenite (As[iii]) and arsenobetaine (AsB). We present a High Performance Liquid Chromatography-Inductively Coupled Plasma-Mass Spectrometry (HPLC-ICP-MS) method that has been optimized to reliably measure the following six arsenic species in human urine: As[iii], arsenate (As[v]), monomethylarsonous acid (MMA[iii]), monomethylarsonic acid (MMA[v]), dimethylarsinic acid (DMA[v]) and AsB. The method was evaluated with regards to changes in mobile phase, accuracy and precision. The ability to quantify the six species in a given sample depended on the low detection limits of the method—0.06 μg L⁻¹ for AsB, 0.11 μg L⁻¹ for As[iii], 0.08 μg L⁻¹ for DMA[v], 0.12 μg L⁻¹ for MMA[v] and 0.15 μg L⁻¹ for As[v]. The procedure was used to measure the six arsenic species in urine samples from 387 individuals in southeast Michigan who are chronically exposed to slightly elevated levels of arsenic in their drinking water. The DMA[v] was detected in 99.2% of samples, AsB in 98.2%, MMA[v] in 73.4%, As[iii] in 45.0%, and As[v] in 27.1%. No MMA[iii] was detected even in samples analyzed within 6 hours after collection. The results raise some doubt as to whether MMA[iii] is a significant metabolite in urine of people exposed to arsenic concentrations below 20 μg L⁻¹ in their drinking water.

Introduction

Arsenic occurs in human urine in several chemical forms which have differing toxicities. The principal species that have been reported include the oxy-anions of arsenite (As[iii]) (mainly arsenite) and arsenate (As[v]) (primarily arsenate) which are known carcinogens.¹ Other important components are the various methylated arsenic species including monomethylarsonous acid (MMA[iii]), dimethylarsinous acid (DMA[iii]), monomethylarsonic acid (MMA[v]) and dimethylarsinic acid (DMA[v]) whose toxicities and carcinogenicities remain highly contentious.^2–4 Urine may also contain other organic arsenic species such as arsenobetaine (AsB), arsenocholine, arsenosugars and arsenolipids which are derived primarily from seafood and are much less toxic than inorganic arsenic.^5,6 A single meal of seafood may increase total urinary arsenic concentrations several orders of magnitude.⁷

In epidemiological studies, urinary arsenic has been used as either a biomarker of exposure^8–10 or a biomarker of effect.^11–13 The importance of DNA methylation on the etiology and early detection (screening) of arsenic-induced cancers is currently an area of active research.^14–16 A number of studies have also reported ethnic and hereditary differences in arsenic metabolism and methylation patterns of urinary arsenic.^9,17,18 The validity of urinary arsenic as a biomarker is very much dependent on our ability to speciate the arsenic accurately. The determination of individual forms of arsenic at very low concentrations in urine still poses a challenge in terms of accuracy and precision for most analytical techniques. In the past, the most common method for determining arsenic species in urine involved high performance liquid chromatography (HPLC), with or without hydride generation, coupled to an atomic absorption spectrometer. Hydride generation was used to discriminate between the so-called toxicologically relevant As forms (i.e., As[iii], As[v], MMA[v] and DMA[v]) from arsenobetaine and other organoarsenicals that do not form volatile arsines.¹⁷ Today, a combination of an HPLC with inductively coupled plasma-mass spectrometry (ICP-MS) has become the instrument of choice for ultratrace analysis of urinary arsenic, due mainly to the superior sensitivity, high samples throughput and wide linearity range of concentration of the system.¹⁹ The High Performance Liquid Chromatography-Inductively Coupled Plasma-Mass Spectrometry (HPLC-ICP-MS) technique, however, is limited by unwanted spectral and nonspectral interferences which can change with the matrix of the sample being analyzed.^20,21 Spectral interference is particularly serious with quadrupole ICP-MS system because of its low resolution.²²

A goal of this study is to develop a method for measuring the six arsenic species (As[iii], As[v], MMA[iii], MMA[v], DMA[v], and AsB) in urine samples at ultratrace levels found in an unexposed population of southeast Michigan using the HPLC-ICP-MS technique. The study was prompted by the finding that the techniques used in many previous studies involved considerable overlapping of AsB and As[iii] peaks and could have resulted in over-reporting of As[iii] concentrations. Another goal of this report is to develop a simple method for overcoming the ArCl⁺ (caused by the chloride in the sample) spectral interference without resorting to the standard interference models. The method was applied to hundreds of samples with a view to identifying other arsenic species that may be present in human urine.

Experimental

Standards

Stock arsenic[iii] and arsenic[v] standards were prepared by dissolving the correct amount of arsenic[iii] oxide (Aldrich, 99.99%) and sodium arsenate dibasic heptahydrate (Aldrich, American Chemical Society [ACS] Reagent), respectively, in water to obtain 1000 mg L⁻¹ arsenic (As). Stock MMA[v] and DMA[v] solutions were prepared by dissolving the correct amount of monosodium acid methane arsonate sesquihydrate (Chem Service, 96.5%) and cacodylic acid (Sigma, 98%), respectively, in water to obtain 1000 mg L⁻¹ As. Stock arsenobetaine standard was prepared by diluting arsenobetaine calibration solution (European Commission, Community Bureau of Reference, 434 mg L⁻¹ As) to 10 mg L⁻¹ in water. The source of MMA[iii] was the diiodomethylarsine (CH₃AsI), which was obtained from Dr William Cullen at the University of British Columbia (Vancouver, CA).²³ MMA[iii] was prepared by dissolving the appropriate amount of CH₃AsI₂ in water to obtain 1000 mg L⁻¹ As. MMA[iii] is formed after hydrolysis of CH₃AsI₂ in water. We prepared 10 mg L⁻¹ stock standards from the 1000 mg L⁻¹ standards and diluted daily to obtain the working standards. All standards were stored at 4 °C.

Urine certified reference material (National Institute for Environmental Studies, NIES NO. 18, Tsukuba, Ibaraki, Japan) was prepared by dissolving the appropriate amount of urine powder in water to obtain 69 ± 12 μg L⁻¹ AsB, 36 ± 9 μg L⁻¹ DMA and 137 ± 11 μg L⁻¹ total As. A 2 mg L⁻¹ germanium (Ge) internal standard solution was prepared from the stock Ge solution (Inorganic Ventures, 1000 mg L⁻¹ Ge) by dilution with the same mobile phase being used in the analytical column that day.

Reagents

Methanol (Fisher, certified electronic grade, 99.8%), tetrabutylammonium hydroxide (TBAH) solution (Fluka, puriss p.a., ~40% in water), ammonium phosphate monobasic (Fisher, 98.9%) and malonic acid (Fisher, reagent grade) were used to make the mobile phase. Mobile phase pH was adjusted using dilute solutions of sodium hydroxide (Fisher, certified ACS, 97.5%) or phosphoric acid (Mallinckrodt, argon [Ar], 85%). The mobile phase was then filtered to protect the column through a 0.2 μm pore diameter nylon membrane filter (Millipore, GNWP 047 00) and sonicated for 30 minutes to remove dissolved gases. Water was obtained from a Milli-Q Water Purification System using deionized water as the input.

Instrumentation

Details of the instrumentation are given in Table 1. The HPLC system included an Alltech model 426 isocratic HPLC pump, Alltech model 530 column heater set at 50 °C, column with precolumn cartridge, and six-port manual injection valve. The columns used included a Phenomenex Gemini C18 analytical column (4.6 × 150 mm) with guard cartridge and a Phenomenex Gemini C18 analytical column (4.6 × 250 mm) with guard cartridge. The effluent from the column was injected directly into the ICP-MS nebulizer through PFA (perfluoroalkoxy) tubing.

Table 1.

Final HPLC and ICP-MS operating conditions

HPLC
Column	Phenomenex Gemini C18, 5 mm, 250 × 4.6 mm (00G-4435-E0)
Precolumn	Phenomenex Security, Guard Cartridges (AJ0-7596)
Column heater	50 °C with mobile phase preheating loop
Mobile phase	4% (v/v) MeOH, 5 mM TBAH, 10 mM ammonium phosphate, pH 9.5
Flow	700 μL min−1
Injection volume ICP-MS	20 μL
Mode	Normal mode, no reaction gas
Forward power	1550 W
Plasma gas	15 L min−1 Ar
Intermediate gas	1 L min−1 Ar
Carrier gas	0.85 L min−1 Ar
Auxiliary gas	0.2 L min−1 Ar
Sample depth	8.0 mm
Extraction lens	0.0 V
Einzel 1,3	−80 V
Einzel 2	12 V
Cell entrance	−15 V
Cell exit	−10 V
Quadrupole bias	−3 V
Octopole bias	−6 V
Plate bias	−45 V
Nebulizer	MicroMist (glass expansion)
Cones	nickel-plated, sampler (1.0 mm orifice) and skimmer (0.7 mm orifice)
Oxides	(140Ce16O+/140Ce+) > 1.0%
Doubly charged	(140Ce2+/140Ce+) < 2.0%
Acquisition	Time-resolved analysis mode with 0.1,0.1 and 0.5 seconds/point for Cl, Ge, and As, respectively

The ICP-MS system used was an Agilent 7500c. The instrument was set to monitor signals at 75 amu (As), 72 amu (Ge) and 35 amu (Cl). Column eluent was introduced into the ICP-MS through a PFA sample line. The line was connected to a tee connecter that also accepted the 2 mg L⁻¹ germanium ISTD (internal standard) diluted in mobile phase. A peristaltic pump supplied the ISTD to the tee. The ISTD solution was reduced to approximately 50 μg L⁻¹ through dilution by the eluent stream. The tee connector was grounded to the ICP-MS chassis through a Pt wire in contact with the sample solution to reduce signal fluctuations from any electrostatic charging caused by the ISTD peristaltic pump. After the tee, the sample was directed into a MicroMist micro-uptake glass concentric nebulizer (Glass Expansion, Australia, AR35-1-FSS04EX). The Peltier-cooled Scott-style spray chamber was jacketed and maintained at 2 °C to increase the signal stability and reduce water vapor in the sample mist. A quartz torch with a 2.5 mm id sample tube was used to prevent clogging caused by salt formation. Nickel-plated sample and skimmer cones (1.0 mm and 0.7 orifices, respectively) were used.

ICP-MS operating conditions and typical lens voltages are listed in Table 1. Conditions were optimized daily for high sensitivity of ⁵⁹Co and ⁸⁹Y while maintaining low oxide formation (¹⁴⁰Ce¹⁶O⁺/¹⁴⁰Ce⁺ > 1.0%) and low doubly charged species (¹⁴⁰Ce²⁺/¹⁴⁰Ce⁺ < 2.0%) for a solution with 10 μg L⁻¹ of Ce. The extraction lens voltage was set near 0.0 volts throughout the study to maintain stability in the instrument response despite the high dissolved solids in the eluent. The electron multiplier and detector were optimized daily using the Agilent Chemstation software and the pulse and analog detection modes of the detector were tuned to accommodate a large response range. Chromatograms for As, Ge, and Cl were collected by the Agilent Chemstation software using time-resolved analysis. Arsenic chromatograms were adjusted by the software to account for signal drift relative to the standard set using the Ge ISTD counts. Chlorine was monitored to determine the position of chloride ion in order to ensure resolution with peaks of interest.

Sample collection and sample preparation

Urine samples were collected from adults enrolled in a case-control study conducted in 11 counties of Southeastern Michigan. Drinking water in Southeastern Michigan contains elevated levels of arsenic.^24,25 Details on the Michigan Bladder Cancer Case-Control Study have been published elsewhere.²⁶ Briefly, cases were bladder cancer cases aged 21–80 (at the time of diagnosis) enrolled through the Michigan State Cancer Registry. Controls were selected from an age-weighted list using a random-digit dialing procedure and were matched to control in terms of age, gender, and county of residence. Demographic, occupational, and residential information were obtained through phone and personal interviews. The home of each participant was visited to collect water, toenails, urine, and saliva samples.

A subsample of 387 participants provided spot urine samples. The research team provided participants with a sample collection kit which included an acid washed 120 mL polyethylene cup in a plastic biohazard bag, disinfection wipes, gloves and instructions. Participants were asked to collect their urine samples, which were then picked up by the research staff. Samples were immediately frozen in dry ice and transported to the Trace Metal Laboratory at the University of Michigan, School of Public Health where they were stored at −20 °C until analysis. Before analysis samples were quickly thawed to room temperature and filtered through a 0.2 μm pore diameter PTFE syringe filter (Whatman Puradisc, 6785-2502). No preservatives were added to samples, and were injected into the HPLC undiluted. Concentrations were adjusted to the mean specific gravity of the samples (1.018 g mL⁻¹). Details on specific gravity adjustments have been described previously.²⁷

Quality control/quality assurance

Calibration standards were prepared in water. Deionized water was sonicated and later used to prepare reference materials and mobile phase. Deionized water was sonicated as a precautionary step to limit the possible oxidation of arsenic species. The ICP-MS was equilibrated daily with the mobile phase to be used in each experiment. Standard and reference materials were injected 3 times into the 1% (v/v) carrier stream. Duplicate injections and duplicate preparations were analyzed daily along with the calibration check standards. Urine samples were frozen in dry ice immediately after collection and stored at −20 °C until analysis. Samples were analyzed in batches ranging from 15–30 samples per day. At the end of the day calibration standards were checked to verify calibration and instrument performance.

Results and discussion

This study grew out of our initial attempt to use the method for arsenic speciation developed by Le et al.²⁸ with the exception that the column was switched to the Phenomenex Gemini, 5 μm, 150 × 4.6 mm column at a flow rate of 700 μL min⁻¹. The mobile phase was 5 mM TBAH, 3 mM malonic acid and 5% (v/v) methanol at pH 6.2. The column heater was set at 50 °C and the output was connected to the ICP-MS nebulizer. The separation between most of the species, As[iii], DMA[v], MMA[v], and As[v] was acceptable, with retention times of 2.5, 3.8, 5.1 and 7.6 minutes, respectively. However, we were unable to separate AsB from As[iii] by this method. This was not a problem for Le et al.²⁸ as the HG-AFS (hydride generation atomic fluorescence spectrometer) system they used was not capable of detecting AsB.

Effect of pH

Changing the pH of the mobile phase was not effective in resolving the close similarity in the retention times of As[iii] and AsB (Fig. 1). Furthermore, DMA[v] and MMA[v] were coeluted at higher pH values. The retention time for As[v] was found to increase with the increase in pH (Fig. 1).

Fig. 1.

Change in retention time of five arsenic species with increasing mobile phase pH using the Gemini, 5 μm, 150 × 4.6 mm column with 5 mM tetrabutylammonium hydroxide, 3 mM malonic acid and 5% (v/v) methanol.

Effect of the mobile phase

Due to the necessity to resolve AsB from As[iii], we decided to switch to another mobile phase system. For these experiments, we initially used the adapted ion-pairing chromatography method of Le et al.²⁸ in conjunction with the phenomenex Gemini silica-organic column. This column can be used over a broad pH range, convenient for the deprotonation of arsenous acid (H₃AsO₂, pK_a = 9.3) to H₂AsO₂⁻ and for enhanced separation of the various arsenic species. The mobile phase was 10 mM ammonium phosphate at pH 8.7 with 5 mM tetrabutylammonium hydroxide (TBAH) added as the ion-pairing reagent and 4% (v/v) methanol added in order to enhance the sensitivity of the ICP-MS.

Four percent MeOH was chosen from past experience with the ICP-MS indicating that addition of alcohols to samples can cause a dramatic increase in ionization of some elements. Arsenic is very sensitive to this effect. The methanol concentration combined with higher-than-usual ICP-MS sample flow rate (700 μL min⁻¹) and mobile phase matrix deposition can lead to faster-than-usual deterioration of sensitivity. However, because sensitivity was excellent, the increased maintenance time was weighed against the need for achieving very low detection limits. For samples with high levels of arsenic, either the sample flow rate or MeOH concentration may be reduced to decrease routine maintenance of the ICP-MS sample introduction system.

Effect of retention times

It was crucial to ensure that chloride in the urine samples would not interfere with the detection and quantification of arsenic species, since ArCl⁺ (75 amu) is not resolved with As (75 amu) with a quadrupole mass analyzer. Adequate resolution between the arsenic species is necessary for accurate quantification of those species. Initially, we used pH 8.7, which resulted in the elution of all species within 11 minutes with reasonable resolution. However, the chloride peak overlapped considerably with that of MMA[v]. By increasing the pH to 9.2, the retention time became longer resulting in adequate separation of MMA[v] and chloride peaks.

Furthermore, these conditions with the new pH provided for near-baseline separation of the AsB, As[iii] and DMA[v] peaks. Using the mobile phase and pH, we were able to successfully resolve AsB, As[iii], DMA[v], MMA[v] and As[v] peaks with eluting times of 2.4, 3.1, 3.8, 6.9 and 10.3 minutes, respectively, as shown in Fig. 2. The total runtime was approximately 13 minutes.

Fig. 2.

Chromatogram of a urine sample using the Gemini, 5 μm, 150 × 4.6 mm column with 10 mM ammonium phosphate, 5 mM tetrabutylammonium hydroxide and 4% (v/v) methanol at pH 9.2.

Fig. 3 shows a chromatogram of a urine sample spiked with five arsenic species. The chloride peak, monitored at 35 amu, is shown as a dotted line. As can be seen, the chloride peak was perfectly resolved from those of arsenic species of interest at a retention time of 9.8 minutes, permitting the use of normal mode operation (i.e., no reaction gases, such as helium) and the omission of interference equations. The separation between DMA[v], Cl⁻, and MMA[v] was good enough that even large changes in Cl⁻ concentration in different urine samples (resulting in a broader Cl⁻ elution and potential for interference) were noninterfering. This method, which is relatively rapid, can be used in studies where the presence of MMA[iii] and DMA[iii] is not expected in the sample.

Fig. 3.

Chromatogram of a urine sample, spiked with 10 μg L⁻¹ of AsB, As[iii], DMA[v], MMA[v], As[v] using the Gemini, 5 μm, 250 × 4.6 mm column with 10 mM ammonium phosphate, 5 mM tetrabutylammonium hydroxide and 4% (v/v) methanol at pH 9.5.

Separation of MMA[iii]

When a standard solution containing As[iii] and MMA[iii] was injected into the 150 mm Gemini column which had provided adequate resolution of the peaks of the five arsenic species and allowed for complete separation of the ArCl⁺ interference, one large, broad peak at approximately 3.2 minutes was obtained. Since it was not possible to change the pH of the mobile phase using the 150 mm Gemini without affecting the separation of the other peaks, it became necessary to switch to a longer column while maintaining the same conditions (10 mM ammonium phosphate, 5 mM tetrabutylammonium hydroxide, and 4% (v/v) methanol at pH 9.2). The switch to a Phenomenex Gemini C18, 5 μm, 250 × 4.6 mm column increased the runtime significantly. At pH 9.2, however, chloride interfered with MMA[v] peak and the MMA[iii], and As(iii) peaks were not adequately resolved. When the pH adjusted up to 9.5, all peaks for the six species were resolved and there was no overlap with the chloride peak (Fig. 4). The eluting times for AsB, MMA[iii], As[iii], DMA[v], MMA[v] and As[v] were 3.87, 5.02, 5.48, 6.66, 12.35 and 19.38 minutes, respectively. Total runtime was approximately 23 minutes.

Fig. 4.

Chromatogram of aqueous solution of AsB, As[iii], DMA[v], MMA[v], As[v], 10 μg L⁻¹, and MMA[iii], unknown concentration using the Gemini, 5 μm, 250 × 4.6 mm column with 10 mM ammonium phosphate, 5 mM tetrabutylammonium hydroxide and 4% (v/v) methanol at pH 9.5.

Recoveries and detection limits

The five arsenic species exhibit a small difference in sensitivity, 2.5, 1.8, 2.2, 2.0, and 2.3 × 10⁵ counts μg L⁻¹ for AsB, As[iii], DMA[v], MMA[v] and As[v], respectively. This might be due to a difference in the ionization potentials of the different species using the current mobile phase and plasma conditions. As can be seen, the ordinary linear regression of the data for the species gave correlation coefficients of at least 0.99 using equally spaced standards. The peak area relative standard deviation (RSD) for three injections of a 20 μg L⁻¹ aqueous standard was less than 2% for all but As[iii], which had an acceptable relative standard deviation of 5.7%.

Table 2 shows the results of a urine sample spiked with 10 μg L⁻¹ of each of the five main arsenic species and recoveries for certified reference material injections. Spike sample peak area precision was very good, less than 2.5% for all five of the main arsenic species. Spike sample peak recoveries were all within 10%. The certified reference material (NIES no. 18) showed good recovery for AsB and DMA[v] at 89 and 94%, respectively, the only two species certified in that standard. While the CRM contained only two arsenic species, the spiked urine samples showed excellent recovery for all five of the main arsenic species found in urine. Thus, we are confident that the matrix has a minimal effect on sample recovery relative to the aqueous standards.

Table 2.

Quantitative results for arsenic speciation method in spiked sample and CRM

	AsB	As[iii]	DMA[v]	MMA[v]	As[v]
Spiked urine sample (10 μg L−1):
Original result (μg L−1)	3.7	0.9	11.4	6.2	2.1
Average spike result	13.2	11	22.4	16.6	12.6
Peak area RSD	1.6	1.9	0.9	2.5	2.4
Average recovery (%)	96	101	110	104	105
Reference material: (NIES No. 18)
Certified (μg L−1)	69 ± 12		36 ± 9
Measured (μg L−1)a	61 ± 0.6		34 ± 0.4
Recovery (%)	89		94

^a99% confidence interval.

The detection limits of speciation were taken to be three times the standard deviation of the background fluctuation of a 10 μg L⁻¹ standard solution, divided by the sensitivity of the instrument for each species on any given day based upon the peak height of each species in the standard solution. The detection limit for all species was less than 0.15 μg L⁻¹ using a 20 μL injection volume. The high sensitivity of this method is due to three main factors: (i) the complete separation between all of the peaks and chloride permits the omission of any interference equations or reaction gas (helium) both of which increase detection limits relative to normal mode operation if it can be done interference-free; (ii) the combination of a highly sensitive ICP-MS instrument with the addition of methanol to the mobile phase allows for very low detection limits for arsenic compounds; and (iii) the low background fluctuation of the chromatographic trace reduces instrumental noise, primarily due to the steady input flow from the HPLC system and the very good precision afforded by the MicroMist nebulizer in conjunction with the Peltier-cooled spray chamber of the ICP-MS.

Urine sample results

Most of the urine samples analyzed showed excellent peak separation. During the course of analysis, we were able to detect two hitherto unknown arsenic peaks. One of the peaks occasionally coeluted with DMA[v], between 6 and 8 minutes. Few of the previous studies had the resolution capacity of our method, hence this particular compound was probably misclassified and reported as DMA[v]. The other unknown occurred after MMA[v], near 12–14 minutes and was completely separated from the analyte peaks. A chromatogram of a urine sample showing the coeluting peak near DMA[v] as well as the resolved unknown near 12.7 minutes is shown in Fig. 5.

Fig. 5.

Urine sample showing two unknown arsenic species, one coeluting with DMA(V) and the other near 12.7 minutes.

Results of arsenic species in spot urine samples from 387 adults who participated in the study are shown in Table 3. The concentrations presented were adjusted by specific gravity to account for the dilution factor.²⁷ Arsenic values were transformed to a log scale in order to perform statistical analysis. DMA[v] was detected in 99.2% of samples, AsB in 98.2%, MMA[v] in 73.4%, As[iii] in 45.0%, and As[v] in 27.1%. This sensitive method enabled us to achieve a higher percent of samples with detectable inorganic arsenic and MMA[v] than what has previously been reported in similar epidemiological studies.²⁹ Neither MMA[iii] nor DMA[iii] was detected in any urine sample even those that were processed in less than six hours after collection (n = 123).

Table 3.

Summary of arsenic species in urine samples

Arsenic species	A mean/μg L −1	G mean/μg L −1	Range/μg L −1	% BDL	Percentiles
					5%	75%	95%
As[iii]	0.340	0.266	BDL-3.40	55	BDL	0.290	1.0
As[v]	0.307	0.244	BDL-7.0	73	BDL	0.165	0.752
DMA[v]	5.57	3.73	0.150–74.0	0.8	0.878	7.15	16.0
MMA[v]	0.995	0.590	BDL-18.0	26	BDL	1.0	3.39
AsB	13.6	4.44	0.166–257.0	1.8	0.541	9.71	72.12
SumAsa	7.18	5.17	0.749–74.6	—	—	—	—

^aSumAs = Sum of As[iii], As[v], MMA[v], DMA[v], or the biologically active arsenic species.

A mean = Arithmetic mean; G mean = Geometric mean; BDL = Below the detection limit.

Although MMA[iii] and DMA[iii] have been reported in urine of people exposed to high levels of arsenic in their drinking water,^10,30,31 many other studies have not been able to detect these urinary arsenic species especially in unexposed populations.^32–34 One reason may be the fact that these species are very unstable and can be completely oxidized to DMA within a day even when stored at −20 °C.³⁵ It has also been suggested that use of uncharacterized standard prepared by treating DMA with sulfur-containing reductants has led many previous studies to misclassify thio-DMA as DMA[iii] in human urine.³⁶ Besides the misidentification of DMA[iii] because of its chromatographic properties, a recent paper by Lindberg et al.³⁷ suggests that thio-DMA was at least contributing to the analytical signal assigned to DMA[iii] in the often-cited study of Valenzuela et al.¹⁰ which found high levels of DMA[iii] in fresh urine samples that were analyzed within 6 hours after collection. Considerable uncertainty thus still surrounds the measurements of MMA[iii] and DMA[iii] that have been reported in the literature.

Even the metabolic role of MMA[iii] and DMA[iii] in the biotransformation of arsenic and the accumulation of these species in urine are being challenged. The Challenger³⁸ pathway produces “free” DMA[iii] and MMA[iii] in a series of oxidation/reduction steps that involve As[iii], MMA[v], and DMA[v] en route to trimethylarsine oxide (TMAO). Since TMAO is not common in urine of people exposed to inorganic arsenic, the Challenger pathway basically leads to the highly reactive DMA[iii] as the end product, which would seem rather odd. The recently proposed Hayakawa pathway involves nonoxidative methylation of arsenic and places MMA[iii] as a precursor of MMA[v] and DMA[iii] as a precursor of DMA[v], hence provides a better explanation for the high levels of DMA[v] and MMA[v] and the absence of TMAO in human urine which has been reported in many studies.³⁷ Our results are consistent with the Hayakawa model in the sense that we were unable to detect any MMA[iii] or DMA[iii] even in fresh urine samples analyzed within six hours after collection.³⁹

Lindberg et al.³⁷ recently reported the presence of thio-DMA[v] in 44% of urine samples from women in Bangladesh exposed to high levels of arsenic in their drinking water. We believe that one of the unidentified peaks in our chromatograms is probably due to this compound. Because Lindberg et al.³⁷ used different column and mobile phase, the chromatographic characteristics of their peaks are not readily comparable with those of the present study and hence cannot be used to confirm the identity of our mysterious peaks.