Anal. Chem. 2005, 77, 5615-5621
Strategies To Study Human Serum Transferrin Isoforms Using Integrated Liquid Chromatography ICPMS, MALDI-TOF, and ESI-Q-TOF Detection: Application to Chronic Alcohol Abuse M. Estela del Castillo Busto, Maria Montes-Bayo´n, Elisa Blanco-Gonza´lez, Juris Meija, and Alfredo Sanz-Medel*
Department of Physical and Analytical Chemistry. University of Oviedo, C/Julia´ n Claverı´a 8, 33006 Oviedo, Spain
Variations in the distribution of sialoforms of human serum transferrin (Tf) in correlation with pathological states, which are associated with abnormalities in glycosylation, is of great clinical interest. In such studies, the methodologies of analysis are required to be sensitive and selective for observing small variations among isoforms and able to characterize the molecular structure of such forms. Thus, the present work describes, in the first part, the separation of transferrin isoforms, after iron saturation of the protein, by high-performance liquid chromatography (HPLC) and the on-line specific atomic detection of the iron present on each of the separated isoforms by online coupling the HPLC system to an inductively coupled plasma mass spectrometer (ICPMS). This allowed low detection levels for the different isoforms (L.D. 0.03 µMTf). After screening of the isoforms containing iron by ICPMS, structural characterization of each isoform can be independently carried out. Thus, matrix-assisted laser desorption/ionization mass spectrometry (MALDI-TOF) and electrospray mass spectrometry (ESI-Q-TOF) are compared in the second part of this study. The different atomic and molecular MS methods revealed the presence of elevated carbohydrate-deficient transferrin (CDT) isoforms in human serum samples from chronic alcohol consumption patients. MALDI-TOF appeared to be sensitive to concentration levels of the analytes, and the observed mass accuracy was highly compromised by the protein heterogeneity (peak width at half-maximum ∼2000 Da for every fraction). On the other hand, ESI-Q-TOF allowed good mass accuracy (m e 0.05%) and peak width of 45 Da in the deconvoluted spectra; while ICPMS detection could be preferable for sensitive protein isoforms determinations, ESI-Q-TOF turns out to be an excellent “fingerprinting” technique for alcoholism diagnosis. Variations of the N-glycan moieties on glycoproteins lead to different glycoforms and affect protein activity and its biological function.1,2 The population of N-glycans attached to an individual * Corresponding author. E-mail:
[email protected]. Fax: +34 985 103125. (1) Dell, A.; Morris, H. R. Science 2001, 291, 2351-2356. (2) Helenius, A.; Aebi, M. Science 2001, 291, 2364-2369. 10.1021/ac050574s CCC: $30.25 Published on Web 07/23/2005
© 2005 American Chemical Society
protein will depend on the cell type in which the protein is expressed and on the physiological status of the cell and can be developmentally and disease regulated.3 Transferrins is the name for a family of monomeric glycoproteins with the main function of iron sequestration and transportation in vertebrates.4 Human serum transferrin (Tf) contains a single polypeptide chain with 679 amino acids and two structurally symmetrical lobes termed the C-terminal and N-terminal, respectively, with each lobe containing one metal binding center. These two metal binding domains can be free or occupied with iron and other metal ions. The two N-glysosylation sites are located in the C-terminal domain at the asparagines 413 and 611, and the carbohydrate moieties attached can have different structures.5 In general, they may show mono-, bi-, tri-, and tetraantennary structures, each antenna terminated with a negatively charged sialic acid (N-acetylneuraminic acid).6 Therefore, Tf can be considered microheterogeneous, attending to both the content of iron and the structure of the N-glycan chains. In this regard, carbohydrate-deficient glycoprotein syndromes (CDGSs), now called congenital disorders of glycosylation, are multisystemic diseases that are typically associated with major nervous system impairment.3,7 In particular, alterations in the normal distribution of Tf glycoforms in human serum of affected individuals serve as a convenient marker of such glycosylation abnormalities. In this vein, carbohydrate-deficient transferrin (CDT) is currently the most sensitive and specific marker for chronic alcohol abuse, exhibiting higher selectivity than other classical biomarkers (e.g., activity of γ-glutamyl transferase).8,9 In this case, a noticeable increase of the Tf isoforms with low content in sialic acid (asialo-, monosialo-, and disialo-Tf) has been observed, even if results regarding accurate molecular mass and structures of those Tf isoforms are still lacking.10 (3) Durand, G.; Seta, N. Clin. Chem. 2000, 46, 795-805. (4) Gumerov, D. R.; Mason, A. B.; Kaltashov, I. A. Biochemistry 2003, 42, 54215428. (5) De Jong G.; van Dijk, J. P.; van Eijk, H. G. Clin. Chim. Acta 1989, 190, 1-46. (6) Kleinert, P.; Kuster, T.; Durka, S.; Ballhausen, D.; Bosshard, N. U., Steinmann, B.; Ha¨nseler, E.; Jaeken, J.; Heizmann, C. W.; Troxler H. Clin. Chem. Lab. Med. 2003, 41, 1580-1588. (7) Keir, G.; Winchester, B. G.; Clayton, P. Ann. Clin. Biochem. 1999, 36, 2025. (8) Arndt, T. Clin. Chem. 2001, 47, 13-27. (9) Musshoff, F. J. Chromatogr., B 2002, 781, 457-480.
Analytical Chemistry, Vol. 77, No. 17, September 1, 2005 5615
Figure 1. Schematic representation of the transferrin molecule carrying two complete biantennary N-glycans. AD1-AD3 N-terminal domain; CD1-CD3 C-terminal domain. Iron binding sites are also depicted.
In the serum of healthy individuals, transferrin has two fully occupied N-glycosylation sites and it is a relatively homogeneous glycoprotein with more than 80% of each oligosaccharide structure in the form of double antennae terminated in one sialic acid each (see Figure 1).11 This high level of homogeneity permits meaningful analysis of the intact transferrin by mass spectrometric (MS) techniques introducing the sample by electrospray (ESI) or by matrix-assisted laser desorption/ionization (MALDI).1 In both techniques, variations in the oligosaccharide content are readily revealed by shifts in the molecular weight that can be ascribed to the differences on the sialic acid content (292 Da per acid molecule) or on the degree and level of sugar branching (2208 Da per one complete biantennary N-glycan). MS techniques are normally used after several separation/purification steps (e.g., 1-D gel electrophoresis, isoelectric focusing (IEF), or after complementary liquid chromatographic approaches).12,13 Traditionally Tf isoforms have been separated by IEF, as the isoforms have different negative charge depending mainly on their content of sialic acid residues.14,15 Another well-established method makes use commercial immunoassays kits based on the removal of nonCDT isoforms by microcolum ion exchange chromatography followed by immunochemical determination of the remaining Tf isoforms (all the forms together). Therefore, it is not possible to detect variations among the different CDT isoforms or the ratios between CDT and non-CDT forms.16,17 More recent studies have also applied anion exchange liquid chromatography to the successful separation of Tf isoforms18 before ESI-MS analysis.6 However, to be able to determine Tf microheterogeneity and the low amounts of existing CDT in the plasma of healthy controls, (10) Lacey, J. M.; Bergen H. R.; Magera, M. J.; Naylor S.; O’Brien, J. Clin. Chem. 2001, 47, 513-516. (11) Oda, R. P.; Prasad, R.; Scout, R. L.; Coffin, D.; Patton, W. P.; Kraft, D. L.; O’Brien, J. F.; Landers, J. P. Electrophoresis 1997, 18, 1819-1826. (12) Peter, J.; Unverzagt, C.; Engel, W. D.; Renaur, D.; Seidel, C.; Ho ¨sel, W. Biochim. Biophys. Acta 1998, 1380, 93-101. (13) Nakanishi, T.; Shimizu, A.; Okamoto, N.; Ingengoh, A.; Kanai, M. J. Am. Soc. Mass Spectrom. 1995, 6, 854-859. (14) Richards, M. P.; Huang, T. L. J. Chromatogr., B 1997, 690, 43-54. (15) Yang, L.; Tang, Q.; Harrata, A. K.; Lee, C. S. Anal. Biochem. 1996, 243, 140-149. (16) Martello, S.; Trettene, M.; Cittadini, F.; Bortolotti, F.; De Giorgio, F.; Chiarotti, M.; Tagliaro, F. Forensic Sci. Int. 2004, 141, 153-157. (17) Arndt, T.; Kropf, J.; Brandt, R.; Gressner, A. M.; Hackler, R.; Herold, M.; Van Pelt, J.; Martensson, O.; Salzmann, K.; Velmans, M. H. Alcohol Alcohol. 1998, 33, 639-645. (18) Arizaga-Rodrı´guez, S.; Blanco Gonza´lez, E.; Alvarez Llamas, G.; MontesBayo´n, M.; Sanz-Medel, A. Anal. Bioanal. Chem. In press.
5616 Analytical Chemistry, Vol. 77, No. 17, September 1, 2005
extremely high sensitivity and specificity are required. In this regard, the coupling of liquid chromatography to a specific (elemental) detector as the ICPMS seems to provide the most sensitive and selective approach to detect protein isoforms if a heteroatom is firmly associated with the protein (e.g., iron bound to Tf)18,19 The charge (and so the separation) of the isoforms can be affected both by sialic acid contents and by possible iron content (the isoelectric point of Tf decreases by ∼0.2 with each FeIII ion bound). For this reason, iron saturation of the protein is required20 previous to Tf isoforms’ separation based on their charge (only 30% of the Tf present in serum is loaded naturally with FeIII). This fact adds an interesting dimension to the analysis of Tf isoforms at extremely low levels of concentration, as required in many clinical issues. In this work, the advantageous use of an octapole reaction system (ICP-(ORS)-MS) as iron-specific detector for screening Tf isoforms present in human serum samples separated by HPLC is illustrated. Once separated, MALDI-TOF and ESI-Q-TOF identification techniques are compared in each HPLC-ICPMS fraction, to obtain the most adequate strategies for the molecular characterization of each separated Tf isoform at very low concentrations and to better evaluate spectral differences among Tf isoforms in alcoholic patients and healthy individual. Additionally, calculations on same-mass possible glycoisomers are performed (considering variations in the degree of branching of the sugar chain and possible additions of other saccharydes, e.g., fucose, to the carbohydrate moiety) to assess the validity of Tf isoform characterization based on molecular mass determinations. EXPERIMENTAL SECTION Materials. Human serum transferrin (Tf) and bovine serum albumin were purchased from Sigma-Aldrich (St. Louis, MO). Mobile phases for HPLC containing (A) 25 mM trisaminomethane (Merck, Darmstadt, Germany)/acetic acid (Merck), pH 6.0, and (B) A + 250 mM ammonium acetate (Merck) were prepared by dilution of the solid salts with the 18 MΩ cm distilled deionized water. Ultrapure 18 MΩ cm distilled deionized water was obtained by means of a Milli-Q system (Millipore, Bedford, MA). The transferrin fractions, collected after HPLC separation, were desalted and preconcentrated in Centricon YM-50 (50 000 Da cutoff) centrifugal filter device (Millipore). For MALDI experiments, a saturated solution of sinapinic acid (Sigma-Aldrich) in 30% acetonitrile (Merck)/0.1% trifluoroacetic acid (TFA, Sigma-Aldrich) was used as matrix to immobilize the sample. In the case of ESI-QTOF, the cleaned fraction was reconstituted in 95% acetonitrile/ 0.2% formic acid (Merck). Human serum samples from healthy voluntaries were kindly provided by Hospital Central of Asturias (Asturias, Spain) and Laboratory for Clinical Analysis (Asturias, Spain), and those from alcoholic patients were obtained from Hospital de San Agustin of Avile´s (Asturias, Spain). Serum samples were separated into 1-mL aliquots and then frozen (-20 °C) until analysis. The remaining sample (not used in the analysis) was discarded after defrosting. (19) Sanz-Medel, A.; Montes-Bayo´n, M.; Ferna´ndez Sa´nchez, M. Anal. Bioanal. Chem. 2003, 377, 236-247. (20) de la Calle Guntin ˜as, M. B.; Bordin, G.; Rodrı´guez, A. R. Anal. Bioanal. Chem. 2004, 378, 383-387.
Table 1. Operating Conditions for FPLC, ICPMS, MALDI-MS, and ESI-MS
anion exchange column mobile phases gradient injection volume flow rate detection forward power external flow carrier gas flow isotope monitored collision/reaction gas flow QP bias octapole bias extraction ion mode instrument mode external calibration matrix laser spectrum acquisition shots/spectrum mass range voltages accelerating grid delay time
FPLC Parameters Mono Q HR 5/5 (50 × 5 mm i.d., Pharmacia) (A) 25 mM Tris-HAc buffer (pH 6.0) (B) A + 250 mM ammonium acetate 0-75% B in 45 min 100 µL 1 mL min-1 UV at 280 nm
ICPMS Parameters 1500 W 15 L min-1 1.1 L min-1 56Fe H2 2.5 mL min-1 -11 V -13 V -3.5 V
MALDI-MS Parameters positive linear 10 µM bovine serum albumin sinapinic acid 2800 V 100 5000-200 000 Da 25 000 V 90% 750 ns
ESI-MS Parameters scan type positive TOF MS ion spray voltage 5.5 kV nebulization gas N2 injection rate 5 µL min-1 external calibration poly(propylene glycol) scan range m/z 500-4000 spectrum deconvolution Bayesian protein reconstruction
HPLC-UV/ICPMS. HPLC separations were carried out using a dual-piston liquid chromatographic pump (Shimadzu LC-10AD, Shimadzu Corp., Kyoto, Japan) equipped with a sample injection valve (Rheodyne, model 7125; Cotati, CA), fitted with a 100-µL injection loop, and an anion-exchange column, Mono-Q HR 5/5 (50 × 5 mm i.d., Pharmacia, Amersham Bioscience). A UV-visible variable-wavelength detector (LKB, model 2151, Berlin, Germany), fitted with a 10-µL flow cell, was used for absorption measurements at 280 nm. Specific atomic detection of Fe in the column effluent was performed using an inductively coupled plasma mass spectrometer model 7500 from Agilent Technologies (Agilent, Tokyo, Japan) equipped with a collision cell system (ICP-(ORS)-MS). H2 has been used as collision gas at a flow of 2.5 mL min-1 in order to reduce the 40Ar16O+ interference on 56Fe+. Operating conditions are showed in Table 1. MALDI-TOF. The instrument used in the study was a Voyager-DE STR Biospectrometry Workstation (Applied Biosytems, Langen, Germany) equipped with a nitrogen pulsed laser (337 nm) and operating in positive mode. Typically mass spectra were acquired by averaging 10 accumulated spectra of 100 single laser shots. External calibration was performed for molecular assignments using a standard of bovine serum albumin. All
experimental and calculated mass values given refer to the average molecular mass and are the mean of three to five different samples. Sample preparation for MALDI-TOF measurements was performed on a stainless steel hydrophobic target (Voyager 96 × 2 Sample Plate P/N V700813) using the dried-droplet technique. For that, an aliquot (0.5 µL) of the sample solution and an equal aliquot of the matrix solution were mixed on the target in the given order and dried at room temperature. The matrix solution used was prepared by dissolving 5 mg of sinapinic acid (Sigma) in 1 mL of 30% acetonitrile, 0.1% TFA. Experimental conditions are given in Table 1. ESI-Q-TOF. The instrument used for this study was a QStar XL model (Applied Biosystems) equipped with the ion spray source and using N2 as nebulization gas. It uses a high-pressure collision cell (LINAC). The instrument was calibrated every day using a standard solution of poly(propylene glycol). Tf fractions from the HPLC column were reconstituted in 95% acetonitrile/ 0.2% formic acid and injected at a flow of 5 µL min-1. The scanned range goes from m/z 500 to 4000, and the applied voltage is 5500 V. The Bayesian deconvolution algorithm available in the Analyst software is applied to the intact Tf spectrum. Experimental working conditions are summarized in Table 1. Iron Saturation and Separation of Tf Isoforms. For iron saturation of human serum transferrin, 0.5 mL of the serum sample was diluted (1 + 1) in 25 mM Tris-acetic acid buffer and then incubated with 25 µL of a 10 mM FeIII solution (as FeCl3, Merck) and 25 µL of a 500 mM sodium bicarbonate (Merck) solution for 3 h at room temperature. The samples previously saturated with iron were subsequently injected in the HPLC system (100 µL), and the Tf isoforms were separated by means of a linear gradient of ammonium acetate (0-250 mM in 45 min) buffered by 25 mM Tris-acetic acid (pH 6.0) solution. The eluate from the HPLC column was passed through a UV detector (set at 280 nm for protein monitoring) for whole protein monitoring and on-line to the ICP-(ORS)-MS detector for specific iron detection. Thus, the UV molecular and atomic corresponding chromatograms were recorded sequentially. Sample Cleaning/Preconcentration. The HPLC fractions monitored by ICPMS were collected for further MALDI-TOF and ESI-Q-TOF experiments. For this purpose, six fractions were collected and individually processed through a Centricon YM-50 (50 000 Da cutoff) centrifugal filter device (5000g, 10 min at 4 °C) in order to exchange the elution buffer by an aqueous solution and to reduce the final sample volume to 0.1 mL. The fractions corresponding to one injection (MALDI-TOF) or five injections (ESI-Q-TOF) were collected and processed as described before. Calculation of Possible Glycan Isobars. A system of biantennary N-glycan chain (A), sialic acid (B), β-Gal-β-GlcNAc sugar side chain (C), and additional fucose (D) was considered. Within this system, existence of the N-glycan isobars was investigated by the brute-force Diophantine analysis using the structural constraints shown in Figure 2. The following variations in mass were adapted: ∆(A; B; C; D) ) 1664; 292; 365; 146 u accordingly. All the structural permutations (consistent with the chemical constraints as shown in Figure 2) were generated and their mass was analyzed. After FeIII saturation, just those Tf species with different number of sialic acid residues should be separated by ion Analytical Chemistry, Vol. 77, No. 17, September 1, 2005
5617
Figure 2. Modular structures of N-glycans and their structural constraints. Symbols used are described in Figure 1.
exchange. Thus, this calculation would illustrate any possible same-mass sugar isomers coeluting in a single chromatographic peak and the expected mass differences among them. RESULTS AND DISCUSSION HPLC-UV/ICPMS Separation of Tf Isoforms in human Serum. Five sera from healthy individuals and two sera from alcoholic patients were processed for Tf isoforms separation using a Mono-Q HR 5/5 (anion exchange) column and on-line UV and ICPMS detection. The isoelectric point of the most abundant Tf isoform, tetrasialo-Tf, S4 (about 65-80% of the total Tf in normal serum samples) is 5.4 while the less sialylated forms have pI values of 5.9, 5.8, and 5.7 for asialo, S0 (
