Quantitative Comparison of Aflatoxin B₁ Serum Albumin Adducts in Humans by Isotope Dilution Mass Spectrometry and ELISA

Aflatoxins (AF) are a major risk factor for hepatocellular cancer in Asia and sub-Saharan Africa (1). AB₁ is the most carcinogenic AF, and an understanding of AFB₁ metabolism and subsequent development of blood and urinary biomarkers played an essential role in the molecular epidemiology studies that established this relationship (2, 3). AFB₁ is oxidized by cytochrome P450s to yield a variety of metabolites, and its toxicology has been reviewed (4). The major reactive metabolite is the exo-AFB₁ 8,9-epoxide, which if not detoxified, can bind to double-stranded DNA to form the promutagenic AFB₁-N⁷-guanine adduct or, following hydrolysis to the AFB₁-dihydrodiol, with proteins such as albumin (5-8). Both urinary AFB₁-N⁷-guanine and serum AF-albumin levels are correlated with dietary intake of AFs (8-13).

AF-albumin accounts for ∼2% of a single AFB₁ dose in rat (3-1,200 μg AFB₁/kg body weight) and human studies (8-10). However, because the half-life of human albumin is ∼20 days, AF-albumin may theoretically accumulate following chronic exposure to reach levels 30-fold higher than that found after a single dose (8, 14, 15). In contrast, urinary AFB₁-N⁷-guanine excretion will tend to parallel intake over the previous few days (16). Therefore, due to its correlation with dietary AF exposure, DNA adduct formation, and the integration of exposure over time, the AF-albumin adduct has been widely applied to assess exposure in epidemiologic studies, including intervention studies (17, 18). The only AFB₁ adduct structurally identified to date in enzymatically digested plasma albumin is AFB₁-lysine (AF-lysine; refs. 14, 19, 20). This adduct has been measured by ELISA, high-performance liquid chromatography (HPLC) with fluorescence detection and more recently by isotope dilution mass spectrometry (IDMS; refs. 10, 21-24). To date, the high-throughput, relatively inexpensive equipment needs and robust performance of the ELISA have made it the method of choice. The sensitivities of the ELISA (∼3 pg AF-lysine equivalents/mg albumin) and HPLC fluorescence assays are comparable, whereas the IDMS method is about 10-fold more sensitive. The ELISA, HPLC fluorescence, and IDMS methods all involve the release of AF residues from digested albumin, but whereas the latter two chromatographic approaches selectively detect AF-lysine, the ELISA will probably measure a broader range of AF adducts.

AF-albumin adduct levels measured by ELISA and AF-lysine adducts measured by HPLC fluorescence in the same human serum samples showed an excellent correlation, but the ELISA indicated ∼11-fold higher adduct burdens (9, 13). Because the ELISA uses AF-lysine to generate the standard curve in a competitive binding technique, the assay metric is pg AF-lysine equivalent/mg albumin and not pg AF-lysine/mg albumin, as used in the more chemically specific HPLC fluorescence and IDMS assays. These methods, therefore, measure closely related but analytically distinct biomarkers of AF and AFB₁ exposure. Here, we directly measure the concentration of AF-lysine and AF-albumin adducts in the same human samples by IDMS and ELISA. This provides an opportunity to compare these assays qualitatively and quantitatively. This is important both to permit comparisons between epidemiology studies using different methodologies and to investigate the contribution of the AF-lysine adduct to the total AF-albumin residues in human blood samples.

AFB₁, bromocreosol purple, human serum, and albumin were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO). The Coomassie protein assay was purchased from Bio-Rad (Hercules, CA). Pronase (120 KPUK/g, nuclease free) was purchased from Calbiochem-Novabiochem (La Jolla, CA). Sep-pak (C18) and mixed-mode, solid-phase extraction cartridges (Oasis MAX) were obtained from the Waters Corp. (Cambridge, MA).

AF-lysine and the tetra-deuterated (D4) internal standard AF-D4-lysine (ISTD) were prepared for mass spectrometric assays and chromatographically purified as previously described (25, 26).

Twenty serum samples from a previous study in West Africa (18) were used in the current study to compare the ELISA and IDMS methodologies. First, immunoglobulins were precipitated from serum using ammonium sulfate, and albumin was precipitated from the resulting supernatant using acetic acid (23). The concentrations of AF-albumin and AF-lysine in precipitated albumin were respectively measured by ELISA (University of Leeds) and IDMS (Johns Hopkins University). IDMS was additionally used to measure the concentration of AF-lysine in whole sera.

The concentration of AF-albumin adducts was measured as previously described (23). Albumin (2 mg) was digested overnight using Pronase, and AF-albumin adducts were isolated by solid-phase extraction using a Sep-pak cartridge. Samples were analyzed in quadruplicate by ELISA on two occasions on separate days.

The concentration of albumin in serum samples was measured using the bromocreosol purple assay (27). The average albumin concentration in serum samples (n = 20) was 52.5 ± 7.1 mg/mL (range, 42.4-72.3 mg/mL). Serum and albumin precipitated from the same serum sample were analyzed using a minor variation of the method reported by McCoy et al. (24). Serum (60 μL, ∼ 2 mg albumin) was mixed with PBS (149 μL), ISTD (100 μL × 2 ng AF-D4-lysine/mL), and Pronase solution (350 μL, 13 mg/mL PBS) and incubated for 4.5 hours at 37°C. Precipitated albumin solution volumes that provided 2 mg of albumin were mixed with PBS such that the combined volume was 209 μL. Precipitated albumin was otherwise processed in the same way as serum. Mixed-mode, solid-phase extracted samples were analyzed by HPLC with mass spectrometric detection using a ThermoElectron TSQ Quantum Ultra operated in the positive electrospray ionization selected reaction monitoring mode as previously described (24). The ISTD parent molecular ion [(M + H)⁺, m/z 461.3] underwent fragmentation to yield an ion at m/z 398.2. The AF-lysine molecular ion (m/z 457.2) fragmented to yield an ion at m/z 394.1. A 10-point isotopic dilution standard curve was generated by triplicate injection of AF-D4-lysine (200 pg) mixed with varying amounts of AF-lysine (0-29 ng/mL) prepared as 3-fold serial dilutions. The data were fitted using the method of least squares with a 1/x weighting factor.

The isotope dilution standard curve was linear over the range 1.3 to 2.9 ng injected onto the column in 100 μL (R² = 0.9998). The coefficient of variation was 3% when 1.3 pg AF-lysine was injected and increased to 45% for 0.5 pg AF-lysine.

Examples of chromatograms from the IDMS assay are presented in Fig. 1. AF-lysine concentrations measured by IDMS in serum and albumin precipitated from the same serum sample are presented in Fig. 2. AF-lysine was detected in all samples from Guinea but not in the negative control human serum. AF-lysine concentrations in precipitated albumin and serum samples in Fig. 2 are described by the equation y = 0.981x + 0.295 (R = 0.982, P < 0.001). The apparent recoveries of the ISTD in serum and precipitated albumin were respectively 90 ± 16% and 83 ± 16% (average ± σ, n = 20). Extensive hemolysis was indicated in one serum sample by its brick red to tan color that carried over into the precipitated albumin. This serum sample exhibited the highest AF-lysine concentration (see Fig. 3). The adduct concentration in this sample decreased by 19% (3.7 pg AF-lysine/mg albumin) in the corresponding precipitated albumin. The average difference between the precipitated albumin and serum sample concentration measured by IDMS was −0.14 ± 1.24 pg AF-lysine/mg albumin.

AF-albumin and AF-lysine adduct concentrations, measured respectively by ELISA and IDMS in precipitated albumin and serum from the same individual, are presented in Fig. 3. AF-albumin and AF-lysine concentration metrics in precipitated albumin samples are described by the equation y = 0.392x + 0.021 (R = 0.856, P < 0.0001). AF-albumin concentration metrics in precipitated albumin samples and AF-lysine concentrations in serum samples are described by the equation y = 0.379x + 0.426 (R = 0.885, P < 0.0001).

The ELISA for AF-albumin adduct is the most widely used assay for measuring human AF exposure in molecular epidemiology studies (23). The AF-albumin levels measured by ELISA correlate well with dietary AF intake and are reduced when intervention strategies are employed that diminish AF exposure (13, 18). Alternative assays for measuring AF exposure include HPLC with fluorescence detection and the recently developed IDMS assay (9, 10, 24). These chromatographic assays specifically measure the AF-lysine adduct; in the case of IDMS, the assay uses an AF-D4-lysine internal standard and detects a parent molecular fragment ion uniquely obtained from AF-lysine (24-26). This improved specificity over the HPLC fluorescence approach means that a more direct comparison with ELISA is now possible. This is important to be able to interpret data both in a qualitative and quantitative fashion across molecular epidemiology studies of AF exposure and disease.

The direct correlation of AF-lysine concentrations measured by IDMS in serum and in albumin precipitated from the same serum sample indicated that precipitation does not degrade AF-lysine or affect the extent of protein digestion. The high recoveries of ISTD in precipitated albumin and serum samples support this interpretation and further indicate that Pronase does not degrade AF-lysine. Low sample recovery and degradation of AF-lysine during precipitation or digestion could have contributed to the lower adduct levels reported by HPLC fluorescence compared with ELISA (9, 10). It is difficult to generalize about the effect of hemolysis on IDMS measurements from the single observation of the 19% lower AF-lysine concentration in precipitated albumin relative to the matched serum sample in an extensively hemolyzed sample because other hemolyzed samples did not exhibit this effect. Thus, the concordance between the analyses of albumin in serum and precipitated albumin provides context for the comparison of ELISA and IDMS data.

The concentration of AF-albumin measured by ELISA was well correlated with the IDMS analysis of AF-lysine in paired samples but the former assay gave consistently higher levels than the latter, by on average 2.6-fold. The measurement of higher adduct concentrations by ELISA relative to IDMS is qualitatively consistent with the previous reports of rat and human ELISA and HPLC fluorescence data, in which the ELISA yielded ∼5-fold and 10-fold higher concentrations than the HPLC fluorescence method in matched rat and human plasma samples, respectively (9, 10, 13). The incomplete digestion of albumin leading to adducted peptides rather than the monoadduct, AF-lysine, was one suggested cause (9). The presence of AF adducts in addition to AF-lysine is an alternative potential cause of the difference in adduct levels measured by ELISA and IDMS.

Unlike laboratory animals treated only with AFB₁, humans can be simultaneously exposed to several AFs, including AFB₁, AFB₂, AFG₁, AFG₂, and AFM₁. Knowledge of AFB₁ metabolism suggests that additional protein adducts should be formed, and the differences in AF-albumin adduct concentration measured by ELISA and IDMS are therefore likely to be due to the presence of uncharacterized AF adducts. Spontaneous formation of the hemiacetal AFB_2a from AFB₁ or AFB₂ can be expected to yield protein adducts through reactions similar to that producing AF-lysine (19, 20, 28). Although significant human exposure to AFG₁ occurs, and this toxin can form both DNA and albumin adducts in rats (29), AFG₁-lysine adduct formation has not been shown in humans. AFM₁ is produced by the hydroxylation of AFB₁ and the terminal furofuran ring can be further oxidized to yield AFM₁-epoxide. AFM₁-guanine adducts have been detected in the liver DNA and urine of shrews treated with AFB₁ (30); however, AFM₁ protein adducts were not assayed, and AFM₁-lysine adduct formation has not been shown in humans. AFP₁, AFQ₁, and the series of AF toxicols could similarly undergo oxidation at the unsaturated 8,9-position to yield hemiacetals, epoxides, or dihydrodiols that could be further sources of protein adducts.

These hypothetical and known compounds would not be detected in the IDMS assay for AF-lysine but would likely be detected by the ELISA. It is notable that the HPLC fluorescence analysis of Pronase digested human serum revealed the presence of several uncharacterized fluorescent peaks in addition to AF-lysine (10). Chromatographic analysis of Pronase digests of serum protein from rats treated with ³H-AFB₁ revealed more than six fluorescent peaks, and AF-lysine accounts for <10% of the detected ³H (21).

Although HPLC fluorescence and IDMS data suggest that the ELISA is measuring other AF adducts in addition to AF-lysine, these closely related but analytically distinct approaches to assessment of AF exposure provide extremely similar classification of individuals with regard to their exposure status. This suggests that the additional adducts do not greatly skew the relationship between AF exposure and measured levels of the biomarkers AF-albumin and AF-lysine. Identification of unknown AF-albumin adducts would further clarify the mechanistic basis for using these biomarkers to assess exposure status in future epidemiologic studies. Because the sensitivity of the ELISA is good, and it requires less expensive instrumentation, it is more generally applicable than IDMS. However, the greater sensitivity and specificity of the IDMS assay can be used to detect AF-lysine adducts at lower levels and thereby potentially reduce exposure misclassification error. Ultimately, each investigator must evaluate their budget and instrumentation resources as well as perform statistical power calculations appropriate to their unique study population to determine which of these complementary analytic methods best serves their needs.