Arsenite as the probable active species in the human carcinogenicity of arsenic: mouse micronucleus assays on Na and K arsenite, orpiment, and Fowler's solution.

Sodium arsenite, potassium arsenite, and Fowler's solution (arsenic trioxide dissolved in potassium bicarbonate) are equally active in the mouse bone marrow micronucleus assay (approximately 10 mg/kg by IP injection). The natural ore orpiment (principally As2S3) was inactive despite blood levels of arsenic of 300 to 900 ng/mL in treated mice at 24 hr. Sodium arsenite was active in three strains of mice. It is suggested that the human lung cancer observed among arsenic ore smelters and the skin cancer among people exposed therapeutically to Fowler's solution, have, as their common origin, the genotoxic arsenite ion AsO2-. The difficulty experienced when attempting to demonstrate rodent carcinogenicity for derivatives of arsenic suggests that the bone marrow micronucleus assay may act as a useful assay for potentially carcinogenic arsenic derivatives.


Introduction
Attention has recently been focused on the clastogenic activity of most of the human carcinogens (including arsenic) to the rodent bone marrow (1,2). The International Agency for Research on Cancer (IARC) (3,4) has classified arsenic and its derivatives as having sufficient evidence for carcinogenicity in man and limited evidence in animals, but the active species is not clear. Shelby (1) quoted the positive mouse bone marrow micronucleus assay data reported for sodium arsenite by Deknudt et al. (5), but an earlier negative bone marrow cytogenetic study on arsenic trioxide also exists (6). Given the range ofarsenic derivatives to which man has been exposed and the uncertainty surrounding the active carcinogenic species (7), we decided to conduct mouse bone marrow micronucleus assays on a range of derivatives of arsenic.
There are two major associations between human cancer and exposure to derivatives ofarsenic: lung cancer among arsenic ore smelters and skin cancer among people exposed therapeutically to Fowler's solution (potassium arsenite). Arsenic occurs in nature mainly in combination with sulfur, the main ores being arsenic trisulfide (orpiment, As2S3), arsenic disulfide (red orpiment or realgar, As4S4), and arsenopyrite (FeAsS). It also occurs in these forms and as arsenides (e.g., niccolite, NiAs) as an impurity (up to 3 %) in many nonferrous ores such as in those ofzinc and copper.
Natural weathering of sulfide ores can lead to natural deposits ofoxides ofarsenic. Orpiment (As2S3) was mentioned by Aristotie and was being roasted to yield arsenic trioxide (white arsenic; As203) as early as the first century A.D. (8). However, apart from the roasting of FeAsS in Japan, the specific roasting of arsenic ores is not now commercially practiced. Rather, the incidental roasting ofarsenic ores during the smelting ofother nonferrous ores provides the major source ofarsenic as arsenic trioxide trapped in flues. The main incidences ofhuman cancer appear to be associated with the culling and further purification ofthese flue deposits.
Fowler's solution is prepared by dissolving As203 in aqueous potassium bicarbonate and is formally aqueous potassium arsenite. Its medicinal use was mainly for the treatment ofa range of skin conditions, but it is no longer used therapeutically. There is therefore a possible link between the human lung cancer induced by the refining ofAs203 and the skin cancer induced by the application ofpotassium arsenite, which is prepared from As203. We therefore decided to repeat the positive mouse bone marrow micronucleus assay of Deknudt et al. (5) on sodium arsenite along with assays on potassium arsenite, Fowler's solution, and orpiment. selected determinations were made ofblood levels of arsenic in treated animals in order to relate genetic activity to bioavailability of arsenic.

Nomenclature of Arsenic and Its Derivatives
Arsenic was known to the ancient Greeks in the form of the poisonous gold/red-colored sulfide ore (Zarnig, Iranian for gold; arrhen, Greek for virile, thus, arsenikon for the gold-hued poison, As2S3, the Latin auripigmentum led to the name orpiment). There is the potential for confusion with the terminology applied to arsenites. In chemistry books written at the turn ofthe century (9), arsenious acid is given as As(OH)3, thus, sodium arsenite is Na2AsO3. A dehydrated acid is also described as metaarsenious acid As(O)OH, whose sodium salt would be sodium meta-arsenite, NaAsO2. These separate terminologies are still used, but the separate existence ofarsenites and meta-arsenites is no longer considered valid. Thus, the 11th edition ofthe Merck Index (10) lists sodium meta-arsenite as an alternative name to sodium arsenite, providing an approximate structure of NaAsO2. Potassium arsenite is listed as being KAsO2:HAsO2 and the meta-arsenite is not mentioned. Likewise, the Merck Index names KAsO2 as the product formed from As203 in Fowler's solution, i.e., potassium arsenite. Confirming this confusion of terms, Scheele's green, copper hydrogen arsenite, is given by Merck in both of the above forms: CuHAsO3 and Cu(AsO2). In this paper the names and molecular species used are sodium arsenite (NaAsO2), potassium arsenite (KAsO2:HAsO2), arsenious trioxide (As203), Fowler's solution (regarded as potassium arsenite), and orpiment (As2S3). In the final experiment (Experiment 8 in Table 1) all dosed animals received equimolar arsenic, independent of the chemical species under test. Chemicals Cyclophosphamide (CP) was purchased from the Sigma Chemical Company (Poole, Dorset, U.K.), sodium and potassium arsenite were obtained from BDH Chemicals Ltd. (Poole, Dorset, U.K.). Fowler's solution was prepared by dissolving As203 (Ventron, Alfa, Coventry, U.K.) (1 g) in water (50 mL) and a solution of potassium bicarbonate (0.76 g) in water (50 mL). Ethanol was added to produce a 3 % solution. A sample of this solution (760 jtL) was added to water (10 mL) for the dosing solution (10 mL/kg gave equimolar arsenic to NaAsO2 [10 mg/kg]). Orpiment was purchased from a mineral shop in Brussels, but the address was not noted and its location cannot be found. The sample used was bright yellow with red streaks and metallic veins. The yellow regions are As2S3, and the orangecolored ones are realgar (As2S4). Leaves ofarsenic metal could be seen by their ability to reflect light. Inductively coupled plasma emission spectrometry revealed a composition ofapproximately 61.6% arsenic, 38.5% sulfur, and 0.5% antimony (theoretical values for As2S3 are 61.0% As and 39.0% S). The mineral was friable and was ground to a fine powder using a pestle and mortar. The fragments ofmetallic arsenic were removed. The yellow powder was added to corn oil (experiment 1). In experiments 2, 6, and 7, the corn oil suspension was ball-milled for either 24 hr (experiment 2) or 72 hr (experiments 6 and 7; Table 1). Finally, the suspensions were homogenized using an ultra-Turrax homogenizer (11) for 0.5 hr immediately before dosing. In the arsenic determinations, the following reagents were used: nickel nitrate hexahydrate, analar grade (BDH Chemicals, Ltd.) Triton X-100 (BDH Chemicals Ltd.), arsenic stock solution, 1000 /Lg/mL, spectrosol grade (BDH Chemicals Ltd.).

Animals and Dosing
Male CBA/CafAP/Alpk mice (experiments 1, 2, 5, 6, and 7), male BALB/cfBLIO/Alpk mice (experiments 3 and 4) and male C57BL/6JfBL1O/Alpk mice (experiment 8) age 8 to 12 weeks were used. Animals were housed and maintained as described previously (12) and were allowed Porton diet and water ad libitum. With the exception of experiment 1, each animal received a single oral or IP dose (10 mL/kg or as stated below) of the test chemical 24 hr before sacrifice. Animals used in experiment 1 received three daily oral gavages ofthe test compounds, and the bone marrow was sampled 24 hr after the final dose. The dose-level oforpiment (experiments 2, 6, 7) was determined by estimating the amount of material remaining from the original amount ball-milled, i.e., the suspension was removed and its volume measured and any remaining solid determined gravimetrically after washing with acetone and drying. Adjustments to the dosing volume were then made to give a maximum dose of 500 mg/kg orpiment. To ease dosing, orpiment was administered using a dosing volume of 20 mL/kg body weight in experiments 2, 6, and 7, as was corn oil in experiment 7. The dose level of500 mg/kg suspension used in experiment 7 (Table 1) was analyzed for arsenic (see below) and found to be within 95 % of theory (i.e., 1.2% arsenic). Initial dose levels of sodium arsenite (used as a positive control) were based on literature values (5). However, its initial inactivity in the mouse bone marrow (experiment 2) led to a more in-depth evaluation of its maximum lethal dose (MLD) and optimum route and vehicle (experiments 3, 4, and 5). A combination ofa single IP injection of 10 mg/kg (80 % 4-day MLD) sodium arsenite dissolved in distilled water gave optimum results (experiments 4 and 5). CP was used at a dose level previously used in this laboratory (13).

Micronucleus Assays
Bone marrow smears were prepared as described previously (14) with a minor alteration. Briefly, the femoral cells were flushed out with fetal calf serum (FCS; Northumbria Biologicals, Northumberland, U. K.) containing 25 mM EDTA, resuspended and filtered through bolting cloth (150 ,m; Henry Simon Ltd., Stockport, U.K.). After spinning (2000 rpm for 5 min), the resultant marrow button was resuspended in a small volume of supernatant, and smears were prepared by placing a drop of this suspension onto the center of a methanol-cleaned slide. A second slide placed on top of the first spread this drop by capilliary action. The two resulting smears were then slid apart, air dried, and fixed in methanol for 5 min. These were stained with acridine orange as described previously (14). A total of2000 polychromatic erythrcytes (PE) per animal were scored for micronucleated PE (MPE). The erythropoietic ratio (poly-

Determination of Arsenic in Blood
A method was developed to determine arsenic in mouse blood by electrothermal atomic absorption spectroscopy (ETA-AAS). Typical sample preparation procedures involve digestion of the blood with acids (15); however, these methods are time consuming and risk contamination. A simple, rapid method has been developed that involves diluting the blood with deionized water and Triton X-100. The direct injection ofblood into the graphite furnace results in a buildup of carbonaceous residue and nonspecific absorption during the atomization cycle, and precautions must be taken to efficiently ash the sample without the volatilization ofarsenic. The chemical stabilization ofarsenic was achieved using nickel nitrate as a matrix modifier (16). However, it has been found that the addition of Ni to blood results in coagulation/precipitation of proteins (17), therefore a "within-tube" standards-addition procedure was used by adding the As stan-dard and Ni to the sample before the dry/ash/atomization cycle. A L'Vov platform was used (18). The introduction ofoxygen during the ash stage was evaluated and rejected (19)(20)(21).
Instrumentation A Perkin Elmer model 3030 atomic absorption spectrometer equipped with an HGA-500 graphite furnace and an AS-40 autosampler were used. Deuterium arc background correction was used. Absorbance measurements were at 193.7 nm using an argon hollow cathode lamp (Perkin Elmer) and a band width of 0.7 nm. Pyrolytically coated graphite tubes with totally pyrolytic L'Vov platforms were used, and measurements were based on peak height, using a read time of 2.0 sec. Some blood samples contained lumps and these were dispersed using an ultrasonic processor, model W-370 (Heat Systems Ultrasonic Inc., Farmingdale, New York) in two 10-sec sessions.

Sample Preparation
Blood samples from each test group were pooled and diluted 1:4 by pipetting 500 Ztg ofblood and diluting with 500 1AL of0.5% m/V Triton X-100 and 1 mL ofdeionized water in 30 mL Sterilin Universal plastic containers. Blood samples taken from mice treated with orpiment required a further 10 to 20 x dilution.

Arsenic Assays
The Perkin Elmer AS-40 autosampler was used to perform standard additions to the blood within the atomizer tube. A 10 1zL volume of blood was dispensed onto the L'Vov platform in the graphite tube, the capillary was automatically washed, and 10 AtL of the As standard containing 0.2% m/V nickel nitrate was deposited onto the droplet of blood (21). To determine the As content of the blood samples, matrix-matched calibration graphs were prepared using a 1:4 dilution of blood from mice treated with corn oil. The blood from mice treated with orpiment contained higher levels of As and was diluted to achieve an absorbance of0.1 to 0.2. A standard-additions procedure was used to calculate the As concentration in the blood. The furnace program is shown in Table 2.

Optimum Assay Conditions
A two-stage temperature program was required to dry the sample and to prevent boiling and sputtering of the blood in the graphite tube. Ash/atomization curves for 1.5 ng As in a 1:4 dilution ofblood using an Ar ash revealed that addition of nickel nitrate allowed ash temperatures of up to 1200°C without the volatilization ofarsenic. Increasing the ashtime from 30 sec to 60 sec improved the analyte peak shape and assay sensitivity. The concentration of nickel nitrate was optimized. A standardadditions graph was prepared in a 1:4 dilution of mouse blood to observe any enhancement or suppression ofthe analyte signal compared with aqueous arsenic standards. It became evident from these graphs that severe suppression of the As signal occurs in the blood matrix. Hence, standard additions or matrixmatched standards were necessary to determine As in the blood samples.

Detection Limit Calculations
A theoretical detection limit was calculated for As in the 1:4 dilution ofblood as 35 ng/g. A more realistic detection limit was determined by relating an absorbance, which was just visible above the baseline noise, to the spiked blood arsenic concentration. This was calculated at approximately 100 ng/g and was used as the detection limit for the analytical procedure.

Results
The results of eight micronucleus experiments, together with selected determinations ofarsenic in blood, are shown in Table   1 in the order in which they were conducted. The sequence ofexperiments illustrates the problems that can beset genotoxicity testing. In experiment 1, a negative test response was observed for orpiment. The relatively low dose level used was caused by problems encountered with dosing the suspension, but was offset by the use of a triple-dose protocol. Although the experiment had a positive control (CP), we decided to repeat the experiment using the more appropriate positive control, sodium arsenite (5). Confident of a positive response for sodium arsenite, we omitted CP from experiment 2, and this gave completely negative data. Nonetheless, we established that orp.iment was absorbed, albeit the level of arsenic in the arsenite-treated animals was below the level of detection of the arsenic assay. In experiment 3, we repeated exactly the experiment reported by Deknudt et al. (IP injection of an aqueous solution using male BALB/c mice) (5). The positive responses observed involved changing two variables at once, so in experiment 4 we established that IP injection of an aqueous solution was critical for a positive test response (IP injection in corn oil giving negative results). Experiment 5 confirmed that the strain ofmouse was not important, so in experiment 6 we reverted to male CBA animals. In experiment 6, we observed a positive response for sodium arsenite and proof again that orpiment was absorbed into the bloodstream. The orpiment data from experiment 6 was positive when assessed by statistical analysis. However, the magnitude of this response lay within the historical control group mean range, and no such positive effect was observed in experiments 1, 2, or 7. Experiment 7 also confirmed that sodium arsenite was only positive following injection in water. The orpiment response in experiment 7 was elevated but was not statistically significant. In experiment 8, sodium and potassium arsenite and Fowler's solution, each gave a clear positive response; use ofa third strain of mouse in this experiment (C57BL6) confirmed that the genotoxicity of arsenite is strain independent. Fromtheseexperimentsweconcludethatsodiumandpotassium arsenite are clearly positive in the micronucleus assay. Fowler's solution gavea positive response ofthe same magnitude as that of the equimolar arsenites (experiment 8). Orpiment was absorbed as evidencedby theelevated levelsofarsenic inthebloodofmice, but an overall negative micronucleus test response was observed. Arsenite is rapidly eliminated when administered by IP injection ( -75 % in 24 hr) (4), thus the failure to detect it in blood at 24 hr indicates a different pharmacokinetic profile from that oforpiment. The elevated blood levels of arsenic in orpiment-treated animals, coupled to the lack ofa depression in erythopoiesis (as seen for arsenite) are consistent with a genetically inactive form of arsenic being present in the blood oforpiment-treated mice.

Discussion
The ability of the rodent bone marrow micronucleus assay to detect as positive the large majority ofthe established human carcinogens qualifies it for use in studying human carcinogens whose exact causative agent is not known. Among the established carcinogenic human environments, the causative agent in cancer induced by derivatives ofarsenic is probably the least well defined. Smelting of metal arsenides or arsenic sulfide ores, or purification ofarsenic trioxide derived as a byproduct from the smelting of other nonferrous ores, and therapeutic exposure to Fowler's solution (potassium arsenite) has adequate evidence of human carcinogenicity (3,4,7,22). There is, however, only very limited evidence for the carcinogenicity of arsenic derivatives to experimental animals (4). Neither is it at all clear how derivatives of arsenic could cause cancer. The ability of some derivatives of arsenic to form pseudo disulfide bonds (As-S) is responsible for a range of enzyme toxicities, and this activity may also lead to the limited evidence for genotoxic activity, most of which is associated with sodium arsenite and its ability to transform cells and damage chromosomes (in vitro) (23,24). The only positive rodent genotoxicity data reported in the literature for derivatives of arsenic is for sodium arsenite in the mouse bone marrow micronucleus assay (5).
The present experiments have confirmed the data of Deknudt et al. (5) indicating that sodium arsenite is genotoxic to the bone marrow cells of mice. Similar activity was observed for potassium arsenite, either as the pure chemical or as formed from As203 in Fowler's solution (Table 1). In contrast, orpiment was concluded to be inactive in this assay despite blood levels of arsenic being at least 10-fold higher in orpiment-treated mice than in those exposed to sodium arsenite. These data suggest that oxygen derivatives ofAs(III) (25) are responsible for the cancer induced in humans. Thus, arsenic trioxide (As203), which is associated with lung cancer in smelters, is soluble in potassium bicarbonate to yield potassium arsenite, which, in Fowler's solution, induces skin cancer in treated humans. The possibility exists that As203 absorbed in man forms arsenite by reaction with bicarbonate in blood.
The inactivity of orpiment in the micronucleus assay strengthens the idea that the naturally occurring sulfides ofarsenic do not contribute directly to smelters' lung cancer. A corollary of this inactivity is that the relatively high levels of arsenic absorbed by the mice was not transformed into the appropriate As(EI) oxide species. This observation is interesting within the context of the closely allied experiments of Arrouijal et al. (26) and Hildebrand et al. (27) on nickel subsulfide (a-Ni3S2), the presumed carcinogenic species in nasal cancer induced among nickel ore roasters (28). These investigators demonstrated solubility for a-Ni3S2 in assay medium and located nickelcontaining deposits within cells treated in vitro. This cellular bioavailability of nickel was probably responsible for Arrouijal et al. (26) also being able to demonstrate a positive mouse bone marrow micronucleus assay response for nickel subsulfide.
It is proposed that arsenite (AsO2 ) is the common carcinogenic species in both the lung cancer among smelters and the skin cancers among those exposed to Fowler's solution. The common activity of these separate arsenic derivatives in the mouse micronucleus assay (experiment 8; Table 1) suggests that this test system may be more effective than animal carcinogenicity bioassays for determining the potential carcinogenic haiard of other derivatives ofarsenic. The general insensitivity of rodents to the carcinogenicity ofderivatives ofarsenic remains a mystery. The extent of human exposure to arsenic trioxide was substantial in the last century, the sublimation chambers being referred to as the poison tower (Fig. 1). Obviously, under those conditions ofexposure, cancer did not become apparent due to early deaths, as evidenced by the following quotation: "Its manufacture [As203] has been chiefly confined to Bohemia and Hungary. Persons brought up from their youth in the works live not longer than  (36) describes the As203 isolation toweras follows; a sectionofthe "poison tower," where the arsenious acid is condensed inthe spaces M, N, 0, P, Q, R; the fumes enterthe lowercompatment atL, andthen follow the direction of the arrows, until finally, the uncondensable vapors depart through the opening S. In the flues and first chambers ofthe tower, the purest arsenic is to be found, that which is deposited in the upper ones being impregnated with sulfur. At the termination ofthe working ofeach charge, the covers, T, are taken away for the purpose ofcollecting the whole ofthe condensed arsenious acid into the lower chamber, which is emptied only about once every 2 months and contains at that period about 25 tons ofthe impure compound. The arsenious acid procured by the foregoing mode is purified by sublimation before it is sent to market.
to the age of 30 or 35 years. Knowing the deleterious nature of their occupation, they are careless" (29). Obviously, as industrial hygiene increased, a wave of cancer deaths occurred (4) and then abated as hygiene improved even further and as usage waned in the 1960s forward (30). Thus, there may be a narrow window between toxicity and tolerance where cancer is evidenced, and this may contribute to the difficulty of showing a carcinogenic effect in animals. Some negative rodent cancer bioassay data may, however, be genuine, as forAs2S3 (31 ), given its inactivity asa genotoxininthepresent experiments, butthe carcinogenic inactivity ofarseniteto rodents is more mysterious (4). A further contribution to the problem of demonstrating an effect in animals may be statistical and related to the heterogeneous effects seen for skin cancers induced by arsenite (as Fowler's solution) inman. Thus, the usual doseofO.2 to 0.5 mL of a 1% solution gives rise to skin cancers in only a susceptible subgroup ofhumans (4). Cuzicketal. (32) reported no overall increase in (internal) cancer incidence among a cohort ofpatientstreatedorally with Fowler's solution ( 1.2 mg/kg/day for 9 months), but an increase in skin cancer was seen among a subgroup showing signs of arsenical poisoning and dermatological changes such as hyperkeratosis. It would obviously be ofgreat valueto resolve the carcinogenicity ofarsenic trioxide or sodium/potassium arsenite to rodentsby theconductofa full National Toxicology Program-type rodent bioassay. To date, the only derivative ofarsenic tested by the National Toxicology Program is roxarsone (a nitrophenyl arsonic acid), and that gave only equivocal evidence ofpancreatic cancer in the male rat (33).
H. T. acknowledges financial support by the Laboratory of the Government Chemist.