Sodium hydroxide, phosphate buffered saline (PBS) tablets, GSH, sodium 2,3-dimercapto-1-propanesulfonate (DMPS), blue dextran, oxidized glutathione (GSSG), and glycine (all > 95%) were purchased from Sigma-Aldrich (St.Louis, MO, USA). Sephadex G-15 (120 μm mean spherical particles of dextran cross-linked with epichlorohydrin, exclusion limit: molecular mass (MM) < 1500 Da) was purchased from GE Healthcare. Sodium arsenite (NaAsO₂) was obtained from GFS Chemicals (> 99%). The source of MMA(III) and DMA(III) was the solid methyldiiodoarsine (CH₃AsI₂) and the liquid dimethyliodoarsine [(CH₃)₂AsI], which were synthesized according to previously established procedures [51, 52]. To verify the purity of these trivalent organoarsenicals, ¹H NMR spectra were obtained for these compounds (chemical shifts: 1.28 ppm for MMA(III) and 1.53 ppm for DMA(III)) and the starting material (chemical shifts: 1.44 ppm for MMA(V) and 1.97 ppm for DMA(V)). The obtained chemical shifts were found to be akin to those previously reported for these compounds [48, 53]. Arsenobetaine bromide (AsB) was synthesized according to a published procedure and its purity was verified by its melting point (experimental: 225°C versus reported: 227°C) [54]. All solutions, including mobile phases, were prepared with water from a simplicity water purification system (resistivity 18.2 MΩ⋅cm, Millipore).

To avoid oxidation of GSH in aqueous mobile phases, all GSH-containing solutions were prepared fresh prior to each chromatographic run and used within 4 hours. PBS-buffer was prepared by dissolving PBS tablets in the appropriate volume of water. After the dissolution of a monothiol (5 or 10 mM GSH) and/or a dithiol (1 mM DMPS) in 400 mL PBS, the pH of the mobile phase was adjusted to 7.4 with sodium hydroxide (4.0 M) using a VWR Symphony SB20 pH meter and filtered through a 0.45 μm Nylon membrane (Alltech).

Aqueous solutions of As(III), MMA(III), DMA(III), and AsB were prepared by dissolving NaAsO₂, CH₃AsI₂, (CH₃)₂AsI, and AsB in water to obtain a concentration of 10 μg As in 20 μL. It is generally accepted that MMA(III) and DMA(III) are the only species that are formed upon hydrolysis of CH₃AsI₂ and (CH₃)₂AsI in water [18, 53]. Even though all chromatograms were generated with the acidic solutions of MMA(III) and DMA(III) (hydroiodic acid is formed during the hydrolysis of CH₃AsI₂ and (CH₃)₂AsI in water), the retention times (t_r) of MMA(III) and DMA(III) were reduced by only ∼ 20 seconds when neutralized solutions were injected. To mitigate oxidation to the pentavalent state, As(III) and MMA(III) solutions were prepared every 14 days while DMA(III) was prepared fresh every 2-3 days [55–57]. All solutions were stored in septum glass vials at 4°C.

The liquid chromatographic (LC) system consisted of a Waters 510 high-performance LC isocratic dual-piston pump, a Rheodyne six-port injection valve (20 μL sample loop), a glass thermostatable SEC column packed with Sephadex G-15 (31×1.0 cm, GE Healthcare), and a cellulose filter on the column head (1.0 cm × 1.0 mm, GE Healthcare). After thermostating the packed column at the desired temperature of 4, 25, or 37°C (NESLAB RTE-7 digital one refrigerated bath, Thermo Scientific), it was equilibrated with at least 60 mL of mobile phase before the As compounds were injected. A flow rate of 1.0 mL/min was used throughout this study and each retention time was determined in triplicate (RSD < 1.0%) after the individual injection of each arsenical.

As-specific detection was achieved with a Prodigy, high-dispersion, radial-view ICP-AES (Teledyne Leeman Labs, Hudson, NH, USA) by monitoring the As atomic emission line at 189.042 nm. Hyphenation of the LC system to the ICP-AES was accomplished by connecting the LC column exit to the concentric glass tube nebulizer with a polyethylene tube (38 cm, 0.13 mm I.D.). The plasma Ar gas-flow rate and the nebulizer gas pressure were 19 L/min and the radiofrequency power output was 1.3 kW. Time scans were performed using the time-resolved analysis mode (Salsa software version 3.0) with a data acquisition rate of one data point every 1.5 seconds. The data were exported and smoothed (bisquares weighting) using commercially available software (SigmaPlot 9.0). The chromatographic window of the packed Sephadex G-15 column was determined by injecting aqueous solutions of blue dextran (MM 2 MDa, t_r~6 minutes), to define the exclusion volume (V₀), and glycine (MM 75 Da, t_r~15 minutes), to define the inclusion volume (V_i). The column was size calibrated with aqueous solutions of GSSG (MM ∼ 600 Da, t_r~10 minutes) and GSH (MM ∼ 300 Da, t_r~13 minutes). All calibration experiments were performed with a PBS mobile phase (pH 7.4 at 25°C), while the C atomic emission line at (193.091 nm) was monitored (Figure 2). Previous studies have demonstrated that the pore size of Sephadex stationary phases does not change appreciably in the 4–37°C temperature range [22].

To determine the influence of temperature on the on-column formation and stability of complexes that were formed between the injected trivalent As compounds and the mobile phase thiol (10 mM GSH), As(III), MMA(III), and DMA(III) were chromatographed on a Sephadex G-15 column at 4, 25, and 37°C. An additional arsenical, AsB (Figure 1(d)), that does not interact with GSH, was also chromatographed under these conditions as an internal standard. The observed retention behavior of As(III), MMA(III), DMA(III), and AsB is depicted in Figure 3.

In general, all four arsenicals eluted within the chromatographic window. As expected, AsB (MM ∼ 178 Da) eluted in the small-molecular-mass region between GSH (MM ∼ 300 Da) and glycine (MM 75 Da), irrespective of the column temperature (t_r~14 minutes). In addition, the gradual reduction in peak width at baseline (w_b) with an increase in temperature from 4 to 37°C can be rationalized by the faster rate of diffusion of AsB into and out of the pores. Based on the retention times of the peaks corresponding to As(III), MMA(III), and DMA(III) (t_r range ∼9–13.5 minutes) compared to glycine (V_i, t_r~15 minutes) and the fact that a single chromatographic peak was obtained for each arsenical, they each must have eluted from the column in the form of their respective (GS)_x-complexes at all investigated temperatures. This interpretation is further substantiated by the observation that As(III), MMA(III), and DMA(III) each eluted at ∼20 minutes, which is 5 minutes after the V_i, when PBS-containing mobile phases without GSH were employed (data not shown). An unspecified chemical interaction between these trivalent arsenicals and the Sephadex G-15 matrix is most likely the cause of this behavior and has been previously observed for As(III) on a similar stationary phase material [22]. Presumably, the free hydroxyl groups of the Sephadex G-15 matrix (stemming from the dextran groups) interacted with the hydroxyl group(s) of the arsenicals via hydrogen bonding; hence, retarding their migration through the column. It is chemically improbable that the observed retention time changes between 4 and 37°C with the GSH-containing mobile phases were caused by the aforementioned unspecified chemical interaction because all injected arsenicals reacted first with GSH in the interstitial volume of the filter (∼80 μL) prior to encountering the stationary phase pores.

The elution order of the injected trivalent arsenicals at 4°C was As(III) (t_r~9 minutes), MMA(III) (t_r~10 minutes), and DMA(III) (t_r~13 minutes), which implies that the hydrodynamic radii of the on-column formed trivalent As-(GS)_x complexes decreased in this order. Based on the known chemical affinity of As(III), MMA(III), and DMA(III) for GSH [22, 23, 62] (1)–(3), this order of elution strongly indicates that As(SG)₃, CH₃As(SG)₂, and (CH₃)₂AsSG had formed on the column head. Evidence in favor of the on-column formation of As(SG)₃ (MM ∼ 900 Da) comes from the observation that As(III) eluted between V ₀ and the MM 600 calibration standard (GSSG), which indicates that a complex with an MM between 1500 and 600 Da had formed (Figure 3). Similarly, CH₃As(SG)₂ and (CH₃)₂AsSG were most likely formed on the column head because the retention times corresponding to MMA(III) and DMA(III) were identical to those of GSSG (MM ∼600 Da) and GSH (MM ∼300 Da; Figure 3, vertical-dashed lines). These trivalent As-(GS)_x complexes could not be structurally characterized in the column effluent by electrospray ionization mass spectrometry (ESI-MS) because the temperature of the electrospray ion source chamber (120–365°C, Esquire 3000 ESI-quadrupole ion trap mass spectrometer) would have dissociated these thermally labile complexes (CH₃As(SG)₂ decomposes at 180°C, while (CH₃)₂AsSG decomposes at 100°C [23]) during the ionization process, prior to mass analysis. Moreover, as advised by the instrument supplier (Bruker Daltonics, Billerica, MA, USA), the salt concentration of mobile phase buffers that can be analyzed by ESI-MS must be below 10 mM (total salt in the utilized PBS buffer was ∼164 mM). Nevertheless, the alignment of the on-column formed complexes with the MM standards (Figure 3) in conjunction with the propensity of trivalent As compounds to react with soft thiol ligands in aqueous solution to form As(SG)₃, CH₃As(SG)₂, and (CH₃)₂AsSG [20–24] strongly suggests that the latter species were formed on the column.

With regard to the retention behavior of As(III), the peak that was observed at 4°C showed considerable tailing, which suggests that As(SG)₃ and an additional complex with a smaller hydrodynamic radius, possibly HOAs(SG)₂, was formed on the column. An increase of the column temperature from 4 to 37°C resulted in a 4.6 minute retention shift of the As(III) peak toward the small-molecular-mass region (Figure 3), which can be interpreted in two different ways. Firstly, the alignment of this peak with GSSG at 25°C and with GSH at 37°C implies similar hydrodynamic radii, which indicates that HOAs(SG)₂ and (HO)₂AsSG were likely formed on the column under these conditions. Even though the peak corresponding to As(III) at 37°C did not perfectly align with the MM 300 calibration standard (GSH), it eluted closer to this standard than to glycine, which suggests that the As(III) was loosely bound to one GS-moiety. Alternatively, the three GS-moieties that are bound to As(III) at 4°C could undergo faster exchange upon increase of the temperature from 4 to 37°C (1). Nonetheless, these results strongly confirm the previously demonstrated lability of the As-S bonds in As(SG)₃ [20, 22, 47–49].

The temperature-dependent retention behavior of MMA(III) and DMA(III) was significantly different from that of As(III) because the retention time of the peaks corresponding to MMA(III) and DMA(III) shifted only marginally (∼40 seconds for both versus 4.6 minutes for As(III)) upon an increase in the column temperature from 4 to 37°C (Figure 3). Hence the on-column formation of the corresponding trivalent As-(GS)_x-complexes (where x = 1 or 2) was only minimally affected by the column temperature. This, in turn, implies that the As-S bonds in CH₃As(SG)₂ and (CH₃)₂AsSG are more stable than those in As(SG)₃. Overall, these results, which were obtained under simulated physiological conditions, are in excellent accord with previous observations, which were conducted under nonphysiological conditions. In particular, these latter studies revealed that CH₃As(SG)₂ and (CH₃)₂AsSG could be synthesized in aqueous solution [23], whereas As(SG)₃ could only be synthesized in alcoholic solutions (alcohol apparently stabilizes the hydrolytically labile As-S bonds) [20, 49].

The entire temperature-dependent retention behavior of As(III), MMA(III), DMA(III), and AsB was repeated with a 5 mM GSH-containing mobile phase (PBS, pH 7.4) on the same SEC column. These results were identical to those illustrated in Figure 3 (data not shown), which makes the observed comparative stability of the on-column formed complexes relevant to mammalian, protein-free, hepatocyte cytosol.

ATP-driven GS-X conjugate export pumps, which shuttle xenobiotic-GS-conjugates across phospholipid bilayer membranes via multidrug resistance proteins 1 and 2, are known to exist at the basolateral and apical hepatocyte membrane, respectively [42–45]. Therefore, our findings raise the possibility that the biomethylation of As(III) to MMA(III) and DMA(III) in the liver of mammals may have evolved simply to export As(III) into the bloodstream for subsequent urinary excretion via the kidney. This rather simplistic hypothesis could explain why DMA(III) has been detected in rat erythrocytes bound to hemoglobin [63] and why MMA(III) and DMA(III) have been detected in mammalian urine [12–16]. The exact mechanism by which these trivalent arsenicals are exported from the liver to the bloodstream, however, needs to be further investigated.

To substantiate that the retention shift of As(III) to a smaller retention time upon the addition of GSH to the mobile phase was caused by the on-column formation of As(SG)₃, a mobile phase containing both GSH (5 mM) and DMPS (1 mM, Figure 1(f)), a chelating agent that forms a stronger complex with As(III) than GSH, was investigated. This resulted in the expected increase in the retention time of As(III) to V_i(t _r ≈ 17 minutes), which is inaccord with previous findings [22]. In contrast, no As peaks for MMA(III) and DMA(III) were detected in the column effluent when a PBS mobile phase containing GSH (5 mM) and DMPS (1 mM) was employed. The cause of this reproducible behavior is not presently understood.