Hemoglobin Adducts of Benzene Oxide in Neonatal and Adult Dried Blood Spots

Biomarkers of internal dose can be more accurate and precise surrogates for carcinogen exposures than environmental measurements per se (1). However, because chemical carcinogens are usually reactive electrophiles with very short life spans in vivo (e.g., alkylating and acylating agents, aldehydes, alkylnitrosamines, dialkylsulfates, oxiranes, quinones, reactive oxygen, and nitrogen species; ref. 2), it is rarely possible to measure them in target tissues. This has motivated the use of adducts of these electrophiles with abundant blood proteins, particularly hemoglobin (Hb) and human serum albumin, as measures of carcinogen dose (3, 4). Electrophiles enter the blood from absorption in the lungs or gut (e.g., inhalation of ethylene oxide) or, more typically, via the metabolism of procarcinogens in the liver or other tissues (e.g., production of benzene oxide by cytochrome P450 metabolism of benzene). Once in the blood, electrophiles react at varying rates with all available nucleophiles to form adducts by numerous mechanisms (5). Hb and human serum albumin contain a myriad of nucleophilic sites, namely the free thiol groups of Cys; amine groups of His, Trp, Lys, and the N-termini; hydroxyl groups of Ser and Tyr; and the carboxylic acid groups of Asp, Glu, and the C-termini. Because protein adducts are not repaired and are much more abundant than DNA adducts in blood (1 mL of blood contains about 150 mg Hb, 30 mg of human serum albumin, and 0.003-0.008 mg of DNA; ref. 5), they are potentially more useful measures of internal dose than DNA adducts, which have paradoxically received far more attention in this regard. Indeed, the kinetics of production and elimination of Hb and human serum albumin adducts are sufficiently simple to permit straightforward estimates of systemic doses of carcinogens over the mean residence time of these proteins (28 days for human serum albumin and 63 days for Hb in humans; refs. 6-8).

Levels of targeted Hb and/or human serum albumin adducts have been investigated in human blood for several environmental toxicants that are either electrophilic carcinogens or their precursors, that is, ethylene oxide, benzene, 1,3-butadiene, acrylamide, aflatoxin B1, a variety of aromatic amines, and polycyclic aromatic hydrocarbons (reviewed in ref. 5). However, the need to obtain venous blood samples has limited the utility of protein adducts (and other blood-based biomarkers) as measures of exposure in large epidemiologic studies. This has motivated investigators to use dried blood spots that can be obtained by a simple skin prick as sources of blood biomarkers. Guthrie and Susi (9) first used dried blood spots in the 1960s to screen newborn populations for hyperphenylalanine associated with the genetic disease phenylketonuria. Neonatal dried blood spots offer valuable opportunities for investigating chemical exposures in utero and their possible links to childhood cancers. Immunoassays have been applied to measure a variety of biomarkers in dried blood spots from adult populations, for example, folate (10), transferrin receptor (11, 12), immunoglobulin E (13), Epstein-Barr virus antibodies (14), leptin (15), and C-reactive protein (16). Thus, dried blood spots can potentially be used in both prospective and retrospective studies to process large numbers of blood specimens from human subjects.

A single dried blood spot contains about 50 μL of human blood (10). Assuming a protein concentration of 192 mg/mL (17), one dried blood spot should contain about 9.6 mg of protein, of which there should be about 7.7 mg of Hb (80% of total protein) and 1.2 mg human serum albumin (12% of total protein). Because most current assays for protein adducts typically require between 1 and 10 mg of globin (from Hb) or human serum albumin, a single dried blood spot should contain sufficient quantities of these proteins to measure adducts. Here, we describe experiments to isolate Hb in high purity from a single dried blood spot, to purify the resulting globin, and to measure cysteinyl adducts of benzene oxide (benzene oxide–Hb) in these proteins. We previously detected benzene oxide–Hb in globin isolated from conventional venous blood samples in both benzene-exposed and control subjects and showed that levels of benzene oxide–Hb increased with the level of benzene exposure (18). Detection of benzene oxide–Hb in control subjects points to production of adducts from environmental exposures to benzene and/or dietary and endogenous sources of benzene oxide or other precursor molecules that produce the same adduct (19). In the current study, we show that benzene oxide–Hb is present at comparable levels in adult globin, isolated either from dried blood spots or from conventional RBCs, and in globin obtained from neonatal dried blood spots.

S-phenylcysteine and [²H₅]S-phenylcysteine were kindly provided by Dr. A. Gold of the Chemistry Core of the Superfund Basic Research Program at the University of North Carolina, Chapel Hill. Hydrochloric acid (concentrated), acetone (nanograde), hexanes (pesticide grade), and ethanol (100%) were purchased from Fisher Scientific. Methanesulfonic acid was obtained from Fluka Chemical Company. Trifluoroaceticanhydride was from Pierce and was distilled once before use. Human Hb was purchased from Sigma Chemical Company. ProteinSaver 903 specimen collection cards were purchased from Whatman. Safety lancets for preparation of fresh adult dried blood spots were obtained from Fisher Scientific. Bio-Rad protein assay reagents were purchased from Bio-Rad Laboratories.

Nine newborn dried blood spots, collected via heel lancet, were obtained from the North Carolina Laboratory of Public Health. These single dried blood spots, which were collected on Whatman ProteinSaver 903 specimen collection cards, were obtained within 30 days of collection. Nine adult dried blood spots were prepared using archived frozen whole blood, randomly selected from 191 adult volunteer subjects collected in the state of North Carolina in 1998 and stored at −80°C (20). After thawing to room temperature, 50-μL aliquots of these blood specimens were spotted on specimen collection cards to produce adult dried blood spot; these dried blood spots were stored at room temperature for 2 weeks before processing. Five of the nine adult dried blood spots were spotted in duplicate to assess assay precision. Fresh adult dried blood spots, required for experiments to develop the method, were prepared on specimen collection cards with blood obtained by finger lancet from laboratory volunteer subjects.

To determine optimal conditions for precipitating Hb, while leaving other blood proteins in solution, dried blood spot specimens were precipitated from varying mixtures of ethanol or water. Eighteen individual dried blood spots from a single volunteer subject were dried overnight and then excised with scissors and extracted with 2 mL of deionized water by agitation on a shaker table (Lab-Line 4626, Barnstead) at 160 rpm for 90 min. The specimen collection paper was removed using forceps, and the eluted blood was concentrated to ∼200 μL using a SpeedVac system (Savant SC110, ThermoFisher Scientific). To determine the optimum concentration of ethanol or water to precipitate Hb, the eluted blood was added dropwise to aqueous solutions containing between 15% and 66% ethanol in increments of 3% ethanol. All samples were incubated at 4°C for 1 h and were centrifuged at 30,000 × g to pellet the precipitated protein. The supernatant fraction, containing soluble proteins, was decanted and discarded. Total protein concentration was measured in the precipitated protein fraction using a variation of the Bradford assay (Bio-Rad Protein Assay, Bio-Rad Laboratories) based upon absorbance at 595 nm. Hb concentrations were estimated by selective absorbance at 523 nm.

Dried blood spots were excised from specimen collection cards with scissors and placed in 4-mL glass vials with Teflon-lined caps. Deionized water (2 mL) was added, and the vials were briefly agitated with a vortex mixer and then gently mixed on a rotary shaker (Lab-Line 4626, Barnstead) at 160 rpm for 90 min. The specimen collection paper was removed with forceps, and the eluted blood was dried with a SpeedVac. After dissolving the dried proteins in 570 μL of deionized water, 430 μL of ethanol (43% v/v, which was found to be optimal in the preliminary experiment described above) was added dropwise to selectively precipitate Hb. Samples were incubated at 4°C for 1 h, transferred to 1.5-mL plastic vials, and centrifuged at 30,000 × g to pellet the Hb. The precipitated Hb was reconstituted in 200 μL of deionized water and then added dropwise to preweighed 4-mL glass vials containing 2 mL of 0.1% HCl in acetone (−20°C) to precipitate globin. Vials were incubated for 4 h at −20°C, centrifuged, and the supernatant was decanted. The globin was then washed thrice with ice-cold acetone and was dried in a vacuum oven overnight at 37°C and 15 mmol/L Hg. The amount of isolated globin was determined gravimetrically.

Globin had been isolated from RBCs from the same nine subjects who provided dried blood spots according to our conventional method (21). Portions of these conventional globin specimens were processed in parallel with globin derived from dried blood spots.

Two-dimensional electrophoresis was done using the carrier ampholine method of isoelectric focusing by Kendrick Laboratories. Nonequilibrium pH gradient electrophoresis was carried out in a glass tube with an inner diameter of 2.0 mm at 200 V for 13 h using 1.5% (pH 3.5-10) and 0.25% (pH 8 to 10.5) ampholines (GE Healthcare). Tropomyosin (1 μg; molecular weight, 33,000; pI 5.2) and 1 μg of lysozyme (molecular weight, 14,000; pI 11.5) were added as internal markers. After equilibrium in SDS sample buffer (10% glycerol; 50 mmol/L dichlorodiphenyltrichloroethane; 2.3% SDS; 0.0625 mol/L Tris; pH 6.8), each tube gel was sealed to the top of a stacking gel overlying a 12% acrylamide slab gel (0.75 mmol/L thick), and SDS slab gel electrophoresis was carried out for 4 h at 15 mA per gel. The following proteins were added as molecular weight standards: myosin (molecular weight, 220,000), phosphorylase A (molecular weight, 94,000), catalase (molecular weight, 60,000), actin (molecular weight, 43,000) carbonic anhydrase (molecular weight, 29,000), and lysozyme (molecular weight, 14,000; Sigma Chemical Co). Gels were scanned with a laser densitometer (model PDSI, Molecular Dynamics Inc.), which had been checked for linearity with a calibrated neutral density filter set (Melles Griot). The images were analyzed using Progenesis Discovery software (version 2005, Nonlinear Dynamics). The general method of computerized analysis included automatic spot finding, background subtraction (Progenesis mathematical model), quantification, and matching with detailed manual checking. Purity of the globin derived from dried blood spot was assessed by comparing two-dimensional gels of globin from dried blood spot with globin derived from RBCs, assuming the latter to be 100% pure.

The cysteinyl adduct of benzene oxide–Hb was assayed using the procedure of Yeowell-O'Connell et al. (21) with minor modifications. After protein isolation, 4.5 mg of globin (from either a dried blood spot or a conventional RBC specimen) and 1 pmol of [²H₅]S-phenylcysteine (internal standard) were placed in 4-mL glass vials and dried in a vacuum oven (70-80°C; 15 mmol/L Hg). Trifluoroaceticanhydride (750 μL) and methanesulfonic acid (20 μL) were added, and the vials were capped tightly with Teflon-lined caps. This reaction cleaves the cysteinyl adduct of benzene oxide and converts it to the volatile derivative phenyltrifluorothioacetate. The reaction mixture was heated at 100°C for 40 min, cooled to room temperature, and the excess trifluoroaceticanhydride was removed under a stream of nitrogen. Hexane (1 mL) was added, and the organic extract (containing phenyltrifluorothioacetate) was washed with 1 mL of 0.1 mol/L Tris buffer (pH 7.3) and twice with 1 mL of deionized water. The hexane layer was then transferred to 1.5-mL high-recovery autosampler vials (Agilent Technologies) and was reduced to a final volume of about 50 μL under nitrogen for analysis of phenyltrifluorothioacetate by gas chromatography–mass spectrometry in negative ion chemical ionization mode. A Hewlett-Packard 5890 series II plus gas chromatograph and a Hewlett-Packard 5989B MS engine were used with a DB-5 fused silica capillary column (60 m; 0.25-mm internal diameter; 0.25-μm phase thickness; J and W Scientific, Inc.) operating a helium carrier gas at a flow rate of 1 mL/min. The injection port and source temperatures were 250°C and 100°C, respectively. The oven temperature was held at 50°C for 3 min and then ramped at 5°C/min to 140°C. Late-eluting compounds were removed by increasing the oven temperature to 250°C at 50°C/min, where it was held for 10 min. Injections (3 μL) were made in the splitless mode. Ions of phenyltrifluorothioacetate (m/z 206) and [²H₅]phenyltrifluorothioacetate (m/z 211) were monitored using selective ion monitoring. To estimate the precision associated with the gas chromatography–mass spectrometry of the benzene oxide–Hb analyte, triplicate injections were done for four dried blood spots (three adult and one newborn), and duplicate injections were done for one adult dried blood spot.

Adduct data were transformed to natural logarithms before statistical analyses to satisfy assumptions about homogeneity of variance and normality. Sources of random variation of adduct levels, from errors associated with the assay and gas chromatography–mass spectrometry injections, were estimated using a nested random-effects model where injections were nested within assays and assays within dried blood spot specimens (Proc NESTED of SAS). The corresponding coefficients of variation (CV) were estimated as follows: CV_assay = [exp(σ̂_a²) − 1]^1/2, CV_injection = [exp(σ̂_i²) − 1]^1/2, and CV_method = [exp(σ̂_a² + σ̂_i²) − 1]^1/2, where σ̂_a² and σ̂_i² represent the estimated variance components (of logged data) for assays and injections, respectively, from the random-effects model (18). Comparisons of mean (logged) adduct levels between sources of nine blood specimens, namely, adult conventional versus adult dried blood spot and adult dried blood spot versus newborn dried blood spot, used two-tailed Student's t tests with unequal variances and a significance level of α = 0.05. Pearson r was estimated for (logged) adduct levels from adult-conventional and adult–dried blood spot specimens. Blood specimens with multiple assays and/or injections were aggregated (logged levels: first by injection and then by assay) before testing of mean values and estimation of r. Statistical analyses used SAS software (version 9.1 for Windows, SAS System Software) for random-effects models and Microsoft Excel for t tests and correlation.

As shown in Fig. 1, Hb in blood from extracted dried blood spots began to precipitate at an ethanol concentration of ∼20% and reached a plateau at ∼37%. Human serum albumin and most of the other plasma proteins remained soluble until an ethanol concentration of ∼47% was reached. Because Hb and total protein were measured using different methods (Bio-Rad assay for total protein and absorbance at 523 nm for Hb), and because a combination of Hb and human serum albumin was used as a standard in the Bio-Rad assay to approximate all of the proteins in blood, total precipitated protein was normalized to the Hb measurements. This adjustment was based on two-dimensional gel electrophoresis results (described hereafter), which showed that, after precipitation with 43% ethanol, Hb represented 96% of the total protein. Based on these results, an ethanol percentage of 43% was chosen to optimize Hb recovery from the dried blood spot while preventing other proteins from precipitating.

The estimated mean globins isolated from adult and newborn dried blood spots were 6.4 mg (SD, 0.96) and 7.2 mg (SD, 2.20) of globin, respectively. This corresponds to a recovery of >80% based on previously reported Hb levels in whole blood (adults, 140 mg/mL; newborns, 193 mg/mL; ref. 22). The purity of the isolated globin, assessed by two-dimensional gel electrophoresis, was determined by laser densitometry to be ∼96%.

The estimated components of variance of (logged) benzene oxide–Hb levels for assays and gas chromatography–mass spectrometry injections were σ̂_a² = 1.62 × 10⁻² and σ̂_i² and 7.25 × 10⁻³, respectively. The corresponding CV values are as follows: CV_assay = 0.128, CV_injection = 0.085, and CV_method = 0.154. Using conventional samples of globin derived from RBCs, we previously reported values of CV_assay = 0.28 and CV_injection = 0.10 (18).

Table 1 lists the estimated means, SDs, and ranges of logged benzene oxide–Hb levels and the corresponding geometric mean benzene oxide–Hb levels for samples of nine globin specimens from different sources. The estimated statistics were very similar across sources. The geometric mean benzene oxide–Hb levels ranged from 27.7 to 33.1 pmol/g globin. Neither of the comparisons of mean (logged) benzene oxide–Hb levels between sources (adult conventional versus adult dried blood spot and adult dried blood spot versus newborn dried blood spot) was significant at a P value of 0.05.

Figure 2 shows a scatter plot of the (logged) benzene oxide–Hb levels in globin derived from adult conventional RBCs and dried blood spots in single specimens of blood obtained from the same subjects. The Pearson r for the nine data pairs was 0.732.

Although blood protein adducts of targeted electrophiles have been measured in humans exposed to prominent genotoxins (see review by Tornqvist et al.; ref. 5), all previous investigations used venous blood samples. The purpose of this investigation was to prove the concept that protein adducts can be measured in dried blood spot-derived globin from both newborn and adult subjects. We focused upon one adduct, namely, benzene oxide–Hb, that we had previously measured in globin isolated from benzene-exposed workers and control subjects from Shanghai, China (18).

An important innovation of our assay involves the isolation of relatively pure Hb from a dried blood spot based upon precipitation in 43% ethanol. Conventional Hb adduct assays have relied on first separating RBCs (which are composed almost exclusively of Hb) from the plasma by centrifugation. This is not possible with dried blood spots because RBCs are lysed during the blood-drying process, giving rise to a more complex whole blood matrix. Precipitating Hb from dried blood spots via addition of ethanol provides a relatively simple and inexpensive means of obtaining globin for use in molecular epidemiologic studies. Although our focus here concerned isolation of Hb, it should be noted that, after precipitation of Hb, the supernatant fraction could also be used to provide additional proteins, notably human serum albumin.

We successfully measured benzene oxide–Hb in all dried blood spot specimens from nine newborns and nine adults in North Carolina and also in globin isolated from conventional RBCs for the same adult subjects. No significant differences in mean (logged) benzene oxide–Hb concentrations were detected between samples of globin isolated from the different sources (adult dried blood spot versus newborn dried blood spot and adult dried blood spot versus adult conventional). Because these t tests were conducted with the logged adduct levels, in a natural scale, they tested for differences between geometric mean benzene oxide–Hb concentrations in the different sources of Hb. The lack of significant differences between geometric mean levels indicates that the typical blood concentrations of benzene oxide–Hb were comparable across sources. This indicates that the typical benzene oxide–Hb level in adult dried blood spots was similar to that measured in conventionally isolated adult globin and that the typical benzene oxide–Hb level in a North Carolina newborn was similar to that measured in a North Carolina adult. Given a pooled estimate of the variance of 0.144 for logged benzene oxide–Hb concentrations, our tests (with nine specimens per group) had a power to detect a 1.7-fold difference in geometric mean values.

Our ability to measure benzene oxide–Hb in dried blood spots proves the concept that a single dried blood spot provides suitable Hb for determinations of a prominent adduct in human populations. This conclusion is reinforced by the high pairwise correlation of benzene oxide–Hb levels measured in globin from dried blood spot and conventional RBCs from the same nine adult subjects (Fig. 2). Because dried blood spots can be collected much more simply and with greater subject acceptance than venous blood samples, this opens the door to the use of protein adducts as biomarkers of human exposure to benzene and to a host of genotoxic and carcinogenic substances in either adults or newborns. With the relentless improvements in analytic sensitivity that we are witnessing, it is reasonable to expect that it will soon be possible to conduct assays of numerous protein adducts with a small portion of a dried blood spot, rather than an entire dried blood spot as reported here.

Given the relatively long residence time of 63 days for chemically stable Hb adducts in human blood (5), Hb adducts provide rather steady measures of human exposure either in utero or in adult populations. Such steady biomarkers provide less biasing surrogates of the true long-term exposure levels that give rise to human diseases and are therefore preferable to short-term biomarkers such as urinary metabolites, whose levels vary greatly from day to day (23).

It is also interesting to note that the geometric mean concentration of benzene oxide–Hb, which in our adult subjects (dried blood spot and conventional globin; n = 18) was e^{(3.32+3.50)/2} = e^3.41 = 30.3 pmol/g globin, is quite similar to the median value of 37.1 pmol/g globin reported in 44 control subjects from Shanghai, China (18). This suggests that the magnitudes of environmental, dietary, and endogenous sources of the precursors of benzene oxide–Hb, which include environmental benzene and other (unknown) contributors, are similar in different parts of the world.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.