C. Derrick
Quarles
Jr
*,
Patrick
Sullivan
,
Nick
Bohlim
and
Nathan
Saetveit
Elemental Scientific, Inc., 7277 World Communications Dr., Omaha, NE, USA. E-mail: derrick.quarles@icpms.com
First published on 20th May 2022
This work focuses on providing fast and reliable separations of arsenobetaine (AsB), trimethylarsine oxide (TMAO), dimethylarsinic acid (DMA), monomethylarsonic acid (MMA), arsenocholine (AsC), arsenite (As(III)), and arsenate (As(V)). Two different methods are presented: (1) a one-column method for the determination of AsB, DMA, MMA, AsC, As(III), and As(V) with a separation time of ∼2 minutes and (2) a two-column method for the determination of AsB, TMAO, DMA, MMA, AsC, As(III), and As(V) with a separation time of ∼4.5 minutes. Recovery of the two methods falls between 94 and 107%. Methods were evaluated for accuracy by analyzing proficiency samples from Centre de Toxicologie du Québec (CTQ) and New York Department of Health (NYDOH). Correlation between the measured values and reference values was very good, with a <4.5% difference in results. Limits of detection in a urine matrix ranged from 2.8–6.0 ng L−1 As and 4.1–9.1 ng L−1 As for the one- and two-column methods, respectively.
Arsenic is perhaps one of the most studied elements for understanding its chemical species and potential toxicity. Arsenic is a naturally occurring element in the Earth's crust and is found in soil around the world at varying amounts, 0.1–40 mg kg−1.4 Arsenic is or has been used in pesticides, herbicides, food additives, drugs, poisons, and chemical weapons.4–6 Arsenic can be found in many different chemical forms. The inorganic forms, arsenite (As(III)) or arsenate (As(V)), are more toxic due to higher bioavailability. Some organic forms (dimethylarsinic acid (DMA) and monomethylarsonic acid (MMA)) are less toxic due to a lower bioavailability, while other organic forms of arsenic such as arsenobetaine (AsB), arsenocholine (AsC), and trimethylarsine oxide (TMAO) are considered non-toxic.7–11
The most common routes of arsenic exposure are from drinking water, food consumption (such as rice or seafood), or industrial exposures.2,8,12 Arsenic is excreted in the urine, therefore measuring urinary arsenic levels can help identify any exposure that had occurred within the previous 24–48 h.13 However, simply measuring the total arsenic levels will not reveal the full impact of the potential exposure. To fully assess the overall health implications for individuals with elevated levels of arsenic, the chemical form of the arsenic species must be identified. The most common methodology for measuring total arsenic is performed using an inductively coupled plasma-mass spectrometer (ICP-MS), whereas determining the arsenic species is typically done by chromatographic separation prior to introduction to the ICP-MS. A large amount of work has been dedicated to this topic in recent decades, however the resulting methods are generally lengthy, offer only a sub-set of the arsenic species, and/or have high operational costs.2,11,14–26 Wegwerth et al. presented the fastest method (2 minutes) for arsenic speciation to date but it did not include TMAO.25 Ciardullo et al. presented a method for seven arsenic species (AsB, AsC, DMA, MMA, TMAO, As(III), and As(V)), however 25 minutes were needed to complete the separation.15
In this work, the sample introduction system for total arsenic and the chromatographic separation of the arsenic species are performed within a single platform automation system (prepFAST IC) connected to a single ICP-MS. This provides automation in the sample preparation and delivery, but also reduces the potential bias of having two completely different setups for these measurements. Two different arsenic speciation methods were developed and evaluated for AsB, AsC, DMA, MMA, As(III), As(V), and TMAO. The methods were evaluated for column recovery, accuracy, precision, and limits of detection. Accuracy of the methods were evaluated by analyzing proficiency testing samples from the Centre de Toxicologie du Québec (CTQ) and New York Department of Health (NYDOH).
Fig. 2 Chromatographic separation of AsB, TMAO, As(III), DMA, AsC, MMA, and As(V) using the (a) one-column and (b) two-column methods. Each species was spiked into the urine sample at 50 μg L−1 As. |
Author | Year | Matrix | As species | Number of species | Time (min) |
---|---|---|---|---|---|
a Methods presented in this publication. | |||||
Quarlesa | 2022 | Urine | AsB, As(III), DMA, AsC, MMA, As(V) | 6 | 2 |
Quarlesa | 2022 | Urine | As(III), DMA, AsC, MMA, As(V), AsB, TMAO | 7 | 4.5 |
Langasco19 | 2022 | Rice | As(III), DMA, MMA, As(V) | 4 | 10 |
Barnet14 | 2021 | Rice | As(III), DMA, MMA, As(V) | 4 | 7 |
Hwang17 | 2021 | Fish | AsC, AsB, As(III), DMA, MMA, As(V) | 6 | 7.5 |
Kara18 | 2021 | Rice | AsC, AsB, As(III), DMA, MMA, As(V) | 6 | 35 |
Montoro-Leal20 | 2021 | Urine | AsB, cacodylate, As(III), As(V) | 4 | 8 |
Wegwerth25 | 2021 | Urine | AsB, As(III), DMA, AsC, MMA, As(V), Rox | 7 | 2 |
Rodriguez11 | 2021 | Urine | AsC, AsB, As(III), DMA, MMA, As(V) | 6 | 28 |
Song23 | 2021 | Urine | AsC/AsB, DMA, As(III), MMA, As(V) | 5 | 11 |
Herath16 | 2020 | Rice | As(III), DMA, MMA, As(V) | 4 | 4.5 |
Quarles27 | 2018 | Urine | AsB, DMA, As(III), MMA, As(V) | 5 | 5 |
Savage22 | 2017 | Water | AsB/TMAO, iAs | 2 | 10 |
Ciardullo15 | 2010 | Fish | As(III), As(V), MMA, DMA, AsB, TMAO, AsC | 7 | 25 |
Tian24 | 2009 | Plants | As(V), As(III), MMA, DMA, TMAO | 5 | 8.5 |
Ruiz-Chancho21 | 2008 | Plants | As, TMAO | 2 | 6 |
Zhao26 | 2006 | Plants, soils | As(III)/As(V), MMA, DMA, TMAO | 4 | 1.2 |
The one- and two-column methods were evaluated for recovery. Table 2 displays the results from the analysis of a urine sample that had been spiked with 10 μg L−1 of all 7 species being studied. TMAO was not included in the one-column method. Both methods had very good recovery that ranged from 94–107% (one-column) and from 97–105% (two-column). The precision ranged from 0.9 to 9.9% RSD and 2.1 to 8.0% RSD, for the one- and two-column methods, respectively (n = 3). While not shown here, urine was spiked individually with each species which resulted in recoveries that ranged from 94–105% for both methods.
As species | 6 species – one-column method | 7 species – two-column method |
---|---|---|
Measured value (μg L−1) | Measured value (μg L−1) | |
As(III) | 10.5 ± 0.4 | 9.7 ± 0.2 |
DMA | 9.4 ± 0.4 | 10.0 ± 0.8 |
AsC | 10.7 ± 0.1 | 10.2 ± 0.6 |
MMA | 10.0 ± 0.7 | 9.9 ± 0.3 |
As(V) | 10.1 ± 1.0 | 10.4 ± 0.3 |
TMAO | n/a | 9.8 ± 0.6 |
AsB | 9.9 ± 0.7 | 10.5 ± 0.6 |
The limits of detection (LOD) were calculated for both the one- and two-column methods (Table 3). The LODs were determined by analyzing urine blanks (n = 10) and applying a 3σ criteria.28 The LODs for these two methods are slightly higher than the previously reported method (0.3–1.7 ng L−1),27 however the values are comparable or lower than the published values using similar methods which range from 3–100 ng L−1.25,29,30 The LODs for the one- and two-column methods are comparable, with the average LOD for the one-column method (4 ng L−1) slightly lower than the two-column method (7 ng L−1). The limit of quantification (LOQ), using a 10σ criteria for these two methods, ranges from 15–30 ng L−1. The linearity of both methods was excellent (R2 ≥ 0.9995) and the slopes were comparable between methods.
One column method | Two-column method | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
Response function | R 2 | SESlope | SEInt | LOD (ng L−1) | Response function | R 2 | SESlope | SEInt | LOD (ng L−1) | |
a LOD = (3 × σblank)/m. m = slope. SE = standard error. | ||||||||||
AsB | y = 7453x − 139 | 0.9995 | 0.038 | 0.020 | 3 | y = 7267x − 247 | 0.9999 | 0.014 | 0.008 | 9 |
As(III) | y = 7499x − 326 | 0.9998 | 0.047 | 0.024 | 5 | y = 7418x − 256 | 0.9999 | 0.045 | 0.024 | 8 |
DMA | y = 7853x − 309 | 0.9998 | 0.068 | 0.036 | 5 | y = 7940x − 637 | 0.9996 | 0.038 | 0.021 | 5 |
AsC | y = 7271x − 236 | 0.9997 | 0.054 | 0.029 | 3 | y = 7503x − 277 | 0.9999 | 0.040 | 0.021 | 9 |
MMA | y = 8055x − 441 | 0.9997 | 0.066 | 0.035 | 5 | y = 7991x − 59 | 0.9998 | 0.072 | 0.044 | 6 |
As(V) | y = 8061x − 288 | 0.9996 | 0.050 | 0.027 | 6 | y = 8155x − 36 | 0.9999 | 0.057 | 0.031 | 4 |
TMAO | — | — | — | — | — | y = 8358x − 455 | 0.9999 | 0.042 | 0.023 | 7 |
These methods were validated by analyzing proficiency testing samples from the NYDOH (5) and CTQ (11) programs. The PT samples were analyzed for total arsenic first to ensure the correct value was obtained for each sample. Table S1† displays the total arsenic reference and measured values for the 16 proficiency samples. There was excellent correlation between the targeted and measured values (Fig. S1†), which is supported by the linear regression slope of 0.9938 (SEslope = 7.12), where a perfect correlation would be equal to 1.0000.
Validation of the one-column and two-column methods were then performed following confirmation that the total arsenic values were correct. The proficiency samples can be separated into three groups: CTQ PC samples which provide a range of inorganic arsenic and total arsenic, CTQ QM samples which provide target values for each arsenic species, and the NYDOH UE samples which only provide target values for total arsenic. There was a fourth group included which were in-house spiked urine samples to ensure that there were target values for each species in the method since none of the proficiency testing samples provided values for AsC or TMAO.
Table S2† displays the reference values for total arsenic and inorganic arsenic for a direct comparison to the measured values for the six arsenic species via the one-column method. The measured values are reported by species, sum of the species, and total inorganic arsenic per sample. Fig. S2† displays a linear regression for the measured sum of arsenic species to the reference values. The correlation is very good (m = 1.0109, SEslope = 7.54) over a fairly wide range (4–631 μg L−1 total arsenic) of arsenic samples. Fig. S3† displays the inorganic reference values reported to the measured values. The slope for this linear regression is 0.9548 (SEslope = 1.70) which is being lowered by the highest concentration point. This sample had a reference value of 153 μg L−1 inorganic arsenic and a measured value of 146 μg L−1 inorganic arsenic, which equates to a −4.5% BIAS which is acceptable. If this point is removed the slope would be 0.9754 further supporting the method has good correlation. The final comparison was done between the reference and measured values for each arsenic species (Fig. S4†), which revealed a correlation that was close to perfect (m = 1.0008, SEslope = 1.35).
Table 4 displays the reference values for total arsenic and inorganic arsenic providing a direct comparison to the measured values for the seven arsenic species, two-column method. The measured values are reported by species, sum of the species, and total inorganic arsenic per sample, however in this experiment TMAO was also included. One noticeable difference from the previous study is that six of the proficiency testing samples had detectable levels of TMAO. Fig. S5† displays the comparison between the sum of the arsenic species measured and the reference total arsenic values. The correlation is very good (m = 0.9954, SEslope = 12.38) with only one data point that appears clearly off of the trend line. That data point was from PC-U-S2008 which had a reference value of 378 μg L−1 total arsenic and a measured value of 338 μg L−1 sum of arsenic species, which equates to a % BIAS of −10.6. When removing this data point the standard error of the slope goes from 12.38 to 4.96. Fig. S6† displays the comparison between the sum of the inorganic species measured to the reference inorganic arsenic values. The correlation between the inorganic species is excellent (m = 1.0018, SEslope = 2.98). Fig. S7† displays the comparison between the individual arsenic species measured using the two-column method and the reported values, with a slope of 0.9988 (SEslope = 1.20). The correlation between the measured values and the reference values were very good for both the one- and two-column methods, suggesting that either method can be used for reliable and accurate arsenic species measurements. Table S3† displays the CTQ QM reference value for each arsenic species and how it compares to the one- and two-column measured values. Two samples had reportable amounts of As(V) and MMA that were not on the provided reference values. The QM-U-Q2013 proficiency testing sample had 2.56 ± 0.19 μg L−1 As(V) and 2.91 ± 0.31 μg L−1 As(V) measured by the one- and two-column methods. The QM-U-Q2005 proficiency testing sample had 2.64 ± 0.23 μg L−1 MMA and 1.17 ± 0.09 μg L−1 MMA by the one- and two-column methods. These two samples were produced in 2005 and 2013, so it is not unreasonable to have some species interconversion over time which may be the cause for these species being measured. The fact that both methods detected As(V) and MMA further confirms the existence of each species in these samples.
Total As target (ref. range) | iAs trget (ref. range) | Two-column method (μg L−1 As) | Sum | iAs | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|
As III | DMA | AsC | MMA | As V | TMAO | AsB | |||||
PC-U-S1907 | 128 (108–148) | 2.44 (0.475–4.41) | 0.628 ± 0.035 | 1.92 ± 0.12 | 0 | 0 | 0.503 ± 0.025 | 1.67 ± 0.02 | 134 ± 11 | 139 | 1.13 |
PC-U-S1908 | 172 (145–199) | 153 (122–184) | 0 | 2.05 ± 0.27 | 0 | 0.68 ± 0.07 | 148 ± 7 | 0 | 12.2 ± 0.6 | 163 | 148 |
PC-U-S1912 | 631 (535–726) | 2.08 (0.123–4.03) | 2.32 ± 0.12 | 2.01 ± 0.08 | 0 | 0 | 0 | 2.35 ± 0.15 | 652 ± 32 | 659 | 2.32 |
PC-U-S1913 | 26.2 (20.6–31.8) | 23.9 (17.5–30.3) | 19.4 ± 1.1 | 5.85 ± 0.43 | 0 | 0 | 1.42 ± 0.15 | 0 | 0 | 26.7 | 20.8 |
PC-U-S2008 | 378 (326–430) | 2.73 (1.08–4.38) | 0.397 ± 0.047 | 6.04 ± 0.11 | 0 | 0.25 ± 0.03 | 1.40 ± 0.06 | 8.39 ± 0.55 | 322 ± 26 | 338 | 1.80 |
QM-U-Q2004 | 93.3 (92.2–94.4) | 80.4 | 82.7 ± 4.9 | 2.39 ± 0.19 | 0 | 0 | 5.15 ± 0.61 | 0 | 6.87 ± 0.52 | 97.1 | 87.9 |
QM-U-Q2005 | 49.9 (49.4–50.4) | 40.2 | 0 | 1.78 ± 0.07 | 0 | 1.17 ± 0.09 | 46.1 ± 6.0 | 0 | 3.20 ± 0.19 | 52.5 | 46.1 |
QM-U-Q2006 | 32.4 (32.0–32.8) | 28.7 | 27.8 ± 2.4 | 1.89 ± 0.03 | 0 | 0 | 1.89 ± 0.21 | 0 | 1.49 ± 0.08 | 33.1 | 29.7 |
QM-U-Q2013 | 42.3 (41.8–42.8) | 6.12 | 5.37 ± 0.25 | 22.0 ± 2.9 | 0 | 3.8 ± 0.4 | 2.91 ± 0.31 | 0 | 5.71 ± 0.34 | 39.8 | 8.28 |
QM-U-Q2014 | 378 (374–382) | 0 | 0 | 2.95 ± 0.21 | 0 | 0 | 0 | 2.37 ± 0.49 | 374 ± 21 | 380 | 0 |
QM-U-Q2015 | 86.0 (85.0–87.0) | 0 | 0 | 73.5 ± 5.1 | 0 | 0 | 0 | 0 | 9.88 ± 0.74 | 83.4 | 0 |
UE19-10 | 188 (150–226) | — | 121 ± 8 | 1.88 ± 0.21 | 0.845 ± 0.100 | 0 | 55.7 ± 2.7 | 0 | 7.03 ± 0.38 | 186 | 177 |
UE19-11 | 3.70 (0.0–9.7) | — | 0.308 ± 0.025 | 0.487 ± 0.033 | 1.28 ± 0.14 | 0 | 0.764 ± 0.076 | 0 | 3.01 ± 0.28 | 5.85 | 1.07 |
UE20-06 | 61.0 (49–73) | — | 18.9 ± 1.8 | 7.75 ± 0.66 | 8.96 ± 0.79 | 3.24 ± 0.33 | 6.59 ± 0.52 | 2.28 ± 0.40 | 23.5 ± 2.3 | 71.2 | 25.5 |
UE20-08 | 21.0 (15–27) | — | 19.7 ± 0.6 | 0.415 ± 0.040 | 1.92 ± 0.08 | 0 | 1.48 ± 0.01 | 0 | 2.11 ± 0.14 | 25.6 | 21.2 |
UE20-10 | 101 (81–121) | — | 46.9 ± 4.1 | 9.94 ± 0.46 | 5.75 ± 0.55 | 0 | 17.6 ± 0.6 | 2.40 ± 0.20 | 22.7 ± 1.7 | 105 | 64.5 |
Urine spike 1 | 70 | 20 | 9.35 ± 0.45 | 9.21 ± 0.51 | 10.5 ± 0.4 | 10.2 ± 0.1 | 10.3 ± 0.5 | 9.41 ± 0.33 | 9.99 ± 0.41 | 69.0 | 19.7 |
Urine spike 2 | 70 | 20 | 9.59 ± 0.27 | 10.1 ± 0.4 | 10.2 ± 0.2 | 10.5 ± 0.6 | 10.8 ± 0.4 | 11.2 ± 0.8 | 10.7 ± 0.7 | 73.1 | 20.4 |
Footnote |
† Electronic supplementary information (ESI) available. See https://doi.org/10.1039/d2ja00055e |
This journal is © The Royal Society of Chemistry 2022 |