Comparative Proteomics of Glycoproteins Based on Lectin Selection and Isotope Coding Li Xiong,† Dina Andrews,‡ and Fred Regnier*,§ Department of Chemistry, Purdue University, West Lafayette, Indiana 47907 and Department of Veterinary Pathology, Purdue University, West Lafayette, Indiana 47907 Received April 2, 2003
Lectins have been widely used in glycan structure analysis. The studies described here exploit this fact to select glycopeptides carrying disease-associated modifications in their oligosaccharides. Coupling lectin affinity selection with recent advances in stable isotope coding for quantitative proteomics allowed a comparative proteomics method to be developed for examining aberrant glycosylation in cancer. Control and experimental samples were individually tryptic digested and differentially coded with stable isotope coding agents before they were mixed and affinity selected with a lectin affinity chromatography column. Glycopeptides carrying an R-L-fucose residue were selected with Lotus tetragonolobus agglutinin (LTA) immobilized on a chromatography matrix. Because the oligosaccharides of glycoproteins are generally heterogeneous and often of unknown structure, it was necessary to deglycosylate the selected peptides with PNGase F before they could be compared to sequences in DNA and protein databases. After deglycosylated peptides were transferred to a reversed phase chromatography (RPC) column and fractionated by gradient elution with increasing amounts of acetonitrile. The RPC fractions were then analyzed by both matrix-assisted laser desorption ionization mass spectrometry (MALDIMS) and electrospray ionization mass spectrometry (ESI-MS). When this method was applied to a study of lymphosarcoma in canines, it was found that during chemotherapy, a series of fucosylated proteins in the blood of patients decreased in concentration more than 2-fold. Two of the proteins identified, CD44 and E-selectin, are known to be involved in cell adhesion and cancer cell migration. The observed aberrant fucosylation of these proteins is consistent with the hypothesis that CD44 and E-selectin play a key role in metastasis and the spread of cancer cells to remote sites. Keywords: proteomics • cancer • glycoproteins • quantification • comparative proteomics • post-translational modification • glycosylation • GIST
Introduction Proteins are frequently altered after translation from mRNA on the ribosome. Among the many types of post-translational modification (PTM) reported in eukaryotic cells, glycosylation is one of the more common.1 Protein glycosylation is critical in many cellular processes and has been associated with diseases such as diabetes,2 cystic fibrosis,3 arthritis,4 Alzheimer’s disease,5 cancer,6 autoimmunity,7 renal function,8 heart disease,9 respiratory capacity diseases,10 and diseases involving stress.11 In fact, the new field of glycopathology is based on the recognition that aberrations in glycosylation can play a major role in disease.12 A critical component of glycopathology is the need to quantify the degree to which aberrant glycosylation of parent protein has occurred. Recent studies indicate that alterations in the glycosylation of cell surface proteins are a hallmark of cancer.13 Moreover, some of these tumor glycoproteins are shed from the cell surface into the circulatory system.14 There is the potential that †
Pfizer Global R&D, Ann Arbor, MI. Department of Veterinary Pathology, Purdue University, West Lafayette. § Department of Chemistry, Purdue University, West Lafayette. ‡
618
Journal of Proteome Research 2003, 2, 618-625
Published on Web 08/06/2003
monitoring glycoproteins with abnormal glycosylation could play a role in detecting and evaluating tumor progression in addition to assessing tumor load and therapy. This is a scientifically important issue because of the increasing incidences of cancer in the human population. Cancer is the second most common cause of death in America, being responsible for over 500 000 deaths annually. Statistics reveal that approximately 40% of the population will have some type of cancer during their life and that 20% of this group will die from the disease.15 Tumorigenesis is a multistep process, involving a series of mutations and other genetic events that are generally unique to the type of tumor and even the individual, to some extent.16 During the malignant progression of a tumor a selection process occurs in which the normal cell cycle is redirected toward unregulated growth, a decrease in natural programmed cell death (apoptosis), invasion of tumor cells into the surrounding tissue matrix increases and, eventually, tumor cell metastasis and migration to distant sites. A critical component of this process is the up-regulation of enabling enzymes.17 At present, the actual glycoproteins involved in cancer and the nature of their modification are not well understood. 10.1021/pr0340274 CCC: $25.00
2003 American Chemical Society
Comparative Proteomics of Glycoproteins
Elevation of N-linked oligosaccharide branching along with the requisite transferase enzymes has been found to be associated with increased metastatic potential of malignant cells.18-20 Transfection of the GlcNAc-TV gene into a lung epithelial cell line and overexpression of this enzyme increased migration of transfected cells and reduced adhesion to collagen and fibronectin.21 There is also a large body of evidence suggesting that the negative charge on sialic acid-containing glycoconjugates is important in cell-cell interaction and that the amount, type, distribution, and bonding of sialic acid to adjacent molecules is different in cancer tissue.22 An increase of Nacetylgalactosamine (N-GalNAc) terminated oligosaccharides has been reported to be yet another glycosylation phenotype in cancer.23 Finally, R(1f6) fucosylation has been associated with certain types of cancer.24 In the case of fucosylation, the literature indicates that (1) fucosylation of the GlcNAc residue attached to asparagine in complex N-linked glycans occurs with the formation of R-Fuc(1f6)-β-GlcNAc-Asn-25 and (2) glycoproteins with this moiety can be elevated in malignant carcinoma. Changes of this type have been detected in cultured neoplastic cells,26 human tumors,27 and in secreted host glycoproteins.28 Of particular relevance to invasion and metastasis are reports of increased expression of fucose-containing Lewis antigens on cancer cells. These antigens have been shown to interact with a class of adhesion molecules, called selectins, on vascular endothelium.29 In breast and ovarian cancer patients, an isoform of alpha-L-protease inhibitor with increased fucosylation was found to be a marker of unresponsiveness to chemotherapy.30 Taken together, these studies provide strong evidence that fucosylation of a variety of proteins may increase during malignant progression. The question is, which proteins are involved? The focus of the research reported here was to develop a proteomics method to quantify changes in the concentration of glycoproteins with specific types of glycosylation and apply it to the study of cancer in an animal model. The premise for this research was that (1) abnormal glycosylation of proteins is associated with cancer, (2) the degree to which aberrant glycosylation occurs can be an indicator of disease progression, (3) specific proteins must be involved, (4) some of these proteins are shed into blood, (5) affinity selection targeting a unique form of glycosylation would be the easiest way to recognize and quantify the proteins involved, and (6) analytical methods that allow searches for disease specific chemical modifications could provide a new way to search for disease markers. Glycosylation of proteins with R-L-fucose was chosen for study because of the availability of lectins that allow fucosylated peptides to be selected from blood samples.31
Experimental Section Materials. Serum samples from dogs with lymphosarcoma at various stages of treatment were obtained from the Purdue University School of Veterinary Medicine. TPCK-treated trypsin, Lotus tetragonolobus agglutinin (LTA), N-[2-hydroxyethyl]piperazine-N′-[2-ethanesulfonic acid] (HEPES), tris[hydroxymethyl]aminomethanehydrochloride (tris-acid), tris[hydroxymethyl]aminomethane (tris-base), sodium chloride, sodium azide, iodoacetic acid (IAA), urea, cysteine, dithiothreitol (DTT), N-tosyl-L-lysine chloromethyl ketone (TLCK), manganese chloride, calcium chloride, fucose, R-cyano-4-hydroxycinnamic acid, d0 and d6-acetic anhydride, N-hydroxysuccinimide, 3-aminopropyl triethoxysilane (APES), poly(acrylic acid) (PAA), and
research articles dicyclohexyl carbodiimide (DCC), hydroxylamine were purchased from Sigma-Aldrich (St. Louis, MO). Lichrospher Si-1000 (10 µm, 1000 Å) was obtained from E. Merck (Darmstadt, Germany). HPLC grade trifluoacetic acid was obtained from Pierce (Rockford, IL). Peptide-N-glycosidase F (PNGase F) was obtained from Calbiochem (San Diego, California). Peptides from human andrenocorticotropic hormone fragment (ACTH) 18-39 and human angiotensin I were purchased from Bachem (Torrance, CA). HPLC-grade acetonitrile, toluene, dioxane, hexane, and dimethyl sulfoxide (DMSO) were purchased from Mallinckrodt Baker (Phillipsburg, NJ). Double-deionized water (ddI H2O) was produced by a Milli-Q gradient A10 system from Millipore (Bedford, MA). All commercially available reagents were used directly without purification. Synthesis of Lectin Columns. The procedure used to immobilize the agglutinin (LTA) is a modification of a method in the literature.32-34 One gram of Lichrospher Si-1000 was activated with 40 mL of 6M HCl under N2 for 6 h at room temperature. After washing with water to neutral pH, the silica was filtered and dried overnight at 120 °C. The resulting particles were treated with 5% 3-amino-propyl triethoxysilane in 10 mL of sodium-dried toluene for 5 h in 90 °C to produce 3-aminopropyl silane (APS)-derivatized silica (3.44 × 10-4 mol APS/m2 silica surface). Elemental analysis of the APS product had a carbon content of 0.48%; hydrogen content of 0.12%; and nitrogen content of 0.13%. Poly(acrylic acid) (0.50 g; Mr 450 000) and N-hydroxyl succinimide (1.67 g) were dissolved in 50 mL DMSO, then 6.00 g DCC dissolved in 10 mL DMSO was added stepwise and stirred at room temperature for 3 h. The resulting mixture was filtered to eliminate dicyclohexylurea (DCU), a white precipitate. The filtrate, which contained succinimidyl polyacrylate, was added to the APS-modified silica. After shaking at room temperature for 12 h, the silica was filtered and washed with 50 mL of DMSO, dioxane, and deionized H2O, respectively. The polyacrylate-coated support with attached N-acyloxy-succinimide groups was designated as PAA-NAS silica (7.94 × 10-4 mol succinimidyl/m2 silica surface). [Elemental Analysis: C 2.48%; H 0.33%; N 0.43%.] Lotus tetragonolobus agglutinin (LTA) was immobilized on the silica surface through the following steps. LTA (100 mg) was dissolved in 20 mL of 0.1 M NaHCO3 buffer containing 0.2 M fucose and mixed with roughly 1 g of PAA-NHS silica. The reaction was allowed to proceed with vigorous shaking for 12 h, after which the sorbent with immobilized LTA was recovered by centrifugation. The remaining active succinimidyl ester sites were blocked with 0.1 M ethanolamine (pH 8.0) after agitating for 30 min. The silica particles were then washed with 1 M NaCl to remove the uncoupled ligands. [Elemental Analysis: C 4.06%; H 0.59%; N 0.94%.] Finally, the sorbent was washed with loading buffer (0.2 M Tris, 0.1 M NaCl, 1mM CaCl2, 1 mM MnCl2, 0.2% NaN3, pH 7.5) and packed into a 4.6 × 50 mm PEEK column. Synthesis of d0 and d6-Succinimidyl Acetate. A solution of 4.0 g (34.8 mmol) N-hydroxysuccinimide in 13.4 mL (104.2 mmol) acetic anhydride was stirred at room temperature for 15 h. The white crystals obtained were washed with hexane and dried under high vacuum. The yield was 3.58 g (60.2%). Product mp was 133-135 °C with diagnostic 1H NMR (CDCl3) lines at δ2.854 ppm (singlet, 4H, succinimide protons) and δ1.320, δ1.295, δ1.270 ppm (triplet, CH3). Proteolysis. One milliliter of dog serum from both treated or untreated animals was adjusted to pH 8.0 with 0.1 M HEPES Journal of Proteome Research • Vol. 2, No. 6, 2003 619
research articles buffer containing 8 M urea before proteolysis. Disulfide bonds in these samples were reduced by the addition of DTT to a final concentration of 20 mM. After incubation at 50 °C in a water bath for 1 h, iodoacetic acid was added to a concentration of 40 mM and the samples incubated in darkness on ice for two more hours. The alkylation reaction was then quenched for 30 min at room temperature by the addition of cysteine to a final concentration of 20 mM. After diluting the samples with 0.1 M HEPES to a final concentration of 0.8 M urea, TPCKtreated trypsin was added in digestion buffer (pH 8.0, containing 10 mM CaCl2) and incubated at 37 °C for 24 h. The trypsinto-protein ratio was 1:50 by mass. Proteolysis was terminated by adding tosyl-lysine chloromethyl ketone protease inhibitor at a molar concentration exceeding that of trypsin by 2-fold. Samples thus treated were stored in liquid nitrogen until analyzed. Acetylation of the Peptides. A 3-fold molar excess of d0-succinimidyl acetate and d6-succinimidyl acetate was added to the tryptic digests of the control and experimental samples in HEPES buffer (pH 7.5), respectively. The reaction was allowed to proceed for 5 h at room temperature. N-Hydroxylamine was then added to the mixture to adjust the pH to 10, and the mixture allowed to incubate for 1 h to hydrolyze esters formed during the acylation reaction. Chromatography. All chromatographic separations were performed using a Biocad Micro-Analytical Workstation from PE Biosystems (Framingham, MA). Tryptic digested human serum (0.5 mL) or human lactoferrin (0.2 mL) was applied to the LTA column after it had been equilibrated with 0.2 M Tris (pH 7.5) loading buffer containing 0.1 M NaCl, 1 mM CaCl2, and 0.1 M MnCl2. Unbound peptides were eluted with 10 mL of loading buffer. Bound glycopeptides were then eluted from the affinity column with a mobile phase containing 0.2 M fucose in a solution of 0.1 M Tris (pH 7.5) containing 0.1 M NaCl, 1 mM CaCl2, and 1 mM MnCl2. Glycopeptides selected by the LTA column were released with 0.1 M L-fucose and directly transferred to a 4.6 × 250 mm PepMap C18 column (Applied Biosystems). Further fractionation of the affinity selected glycopeptides on the C18 RPC column was achieved with a 120 min gradient from 5% acetonitrile in 0.1% aqueous trifluroacetic acid (buffer A) to 60% acetonitrile containing 0.1% aqueous TFA (buffer B). Eluted peptides were monitored at 214 nm and fractions were manually collected for MALDI-MS and ESI-MS analysis. Deglycosylation by N-glycosidase F (PNGase F). Glycopeptides were speed-dried under high vacuum and reconstituted in 80 µL 100 mM sodium phosphate, pH 7.2, containing 25 mM EDTA. The reaction mixture was incubated with 2 µL (10U) of PNGase-F for 12 h at 37 °C. Matrix-Assisted Laser Desorption Ionization Mass Spectrometry (MALDI-MS). MALDI-MS analyses were performed on a Voyager DE-RP BioSpectrometry workstation from PerSeptive Biosystems. The matrix solution was acetonitrile (ACN)water (50:50) containing 0.1% TFA saturated with R-cyano-4hydroxycinnamic acid. One microliter of matrix solution was spotted on top of the samples, and the solvent was allowed to evaporate in air before placing the sample plate in the mass spectrometer. Glycopeptides were analyzed in the linear, positive ion mode by delayed extraction using an accelerating voltage of 20 kV. Deglycosylated peptides were analyzed in the reflector mode with delayed extraction. External calibration was achieved using a mixture of standard peptides containing angiotensin I (Mr 1296.68) and ACTH 18-39 (Mr 2465.70). 620
Journal of Proteome Research • Vol. 2, No. 6, 2003
Xiong et al.
ESI-MS Analysis. Samples were analyzed with a QSTAR mass spectrometer (Applied Biosystems/PE SCIEX, Foster City, CA) fitted with an electrospray ionization (ESI) inlet. Samples were dissolved in ACN-H2O (50:50) containing 0.1% formic acid and infusion injected with a syringe pump flowing at 1030 µL/min. The QSTAR was operated above 8000 in resolution with a mass accuracy of 10-30 ppm using external calibration standards once a day. MS/MS sequencing was achieved by selecting the parent ion with the quadruple in the lowresolution mode to get the entire isotope cluster into the collision cell. The instrument was run in the positive ion TOF mode (m/z 300-2000). The typical ion spray voltage used was 5000 V, with a collision-energy of 30-60 eV to fragment peptides. Nitrogen was used as the collision gas with typical pressures in the collision cell during MS/MS being 4 to 6 ×10-6 Torr. Tandem Mass Spectrometry. Peptides observed to change more than 1.5-fold in concentration were analyzed by tandem mass spectrometry using collision-induced-dissociation (CID) to produce a sequence ladder. Spectra from the second dimension of mass spectrometry were analyzed both manually and with the Mascot software (Matrix Sciences, London, UK) interface for searching databases. Typically NCBI, SWISSPROT, and OWL databases were searched through Mascot. As there is often chymotrypsin activity in the commercially available TPCK-treated trypsin (data not shown), databases were searched for both tryptic and chymotryptic peptides with up to 9 miscleavages. Because databases contain few canine proteins, the search query was instructed to consider other mammalian proteins. All cysteine residues were treated as being carboxymethylated, lysine and N-terminal amino groups were considered to be acetylated, methionine oxidation, and sodium cationization was also included in searches. Deamidation of asparagine caused by deglycosylation was also considered. A sequence tag of several continuous amino acids (5-20 residues) and the peptide mass were generally sufficient to identify the protein parent of a peptide.
Results and Discussion Animal Model. Dogs have proven to be a useful model for studying the etiology and therapy of spontaneous human cancers.35-40 The tumor class known as non-Hodgkin’s lymphomas (NHLs) in humans is similar in many ways to lymphosarcoma in dogs. There are even a number of similarities in cancer between the species.41 One is the high rate of incidence. A second is that NHL and lymphosarcomas are a disparate group of malignant diseases in both man and canines.42-44 They also range from some of the most indolent malignancies to the most rapidly growing and highly aggressive tumors in both species.45-46 Another similarity is in the clinical manifestations of these types of cancer in dogs and humans. High grade or aggressive tumors responds well to combination chemotherapy in both species. Although a small proportion of humans will be cured of NHL, the majority relapses within two years. The preponderance of dogs with lymphosarcoma also can expect to experience a relapse after remission of 12 to 14 months. Yet another similarity is that in both species, these tumors defy straightforward pathologic classification, which translates into meaningful information for the clinical oncologist faced with therapeutic decisions.47 Canines were used in this study because they provide an excellent, easily obtainable, large animal model for the discovery and profiling of tumor markers. A negative aspect of using canines is that the canine
Comparative Proteomics of Glycoproteins
genome has not been sequenced and many canine proteins are not in databases. Experimental Design. The goal of this study was to develop proteomics methods that recognize and quantify changes in the concentration of glycoproteins in correlation with cancer treatment. This was done in two ways. One was by simultaneously comparing serum samples from control and experimental animals. The other was by comparing samples from a single animal during the course of therapy. In the latter case, experimental samples were taken from canine cancer patients with multicentric lymphoma before treatment and during the course of chemotherapy with vincristine/pred/lspar.48 This protocol was chosen for several reasons. One was that an animal can be used as its own control. The potential for genetic variation between animals is eliminated in this way. Another is that it tests the efficacy of chemotherapy. A third is that recurrence of the disease can be monitored and associated with a specific protein or proteins. The analytical process selected involved the steps of (1) tryptic digesting equal amounts of control and experiment samples individually, (2) differentially acetylating the resulting tryptic peptides from these samples with d0- and d3-acetoxy succinimide, respectively, (3) mixing the two samples, (4) selecting R-L-fucose containing glycopeptides with a Lotus tetragonolobus lectin (LTA) column,49 (5) recovering glycopeptides from the column and deglycosylating the selected peptides with PNGase F, (6) fractionating the deglycosylated peptides with reversed-phase chromatography, (7) determining the isotope ratio of peptide isoforms by either MALDI-MS or ESI-MS, (8) MS/MS sequencing peptides that had changed substantially in concentration, and (9) identifying protein parents of sequenced peptides through database searches. This procedure is identical to the GIST protocol used in expression analysis50 with the exception of steps 4 and 5. The rationale for these steps will be developed below. Glycans are known to be heterogeneous.47 There is a high probability a given glycopeptide will occur in multiple glycoforms of unknown structure and unknown molecular weight. This means that the molecular weight of the peptide portion of the glycopeptide cannot be determined from a mass spectrum of the glycopeptide. Because peptide molecular weight is a prerequisite in database identification of a protein, this is a serious problem. Eliminating the unknown glycan component through PNGase deglycosylation circumvents this problem.47 PNGase F hydrolysis occurs at the Asn residue in peptides where the oligosaccharide is coupled to the polypeptide backbone. Although PNGase hydrolyzes glycopeptides with either R(1f6) fucosylation or R(1f3) fucosylation, those with R(1f6) fucosylation are cleaved much more rapidly. Deglycosylation also simplifies the reversed phase chromatogram of a peptide.47 Glycoforms can be partially or totally resolved during reversed phase chromatography. Deglycosylation eliminates this problem as well. But the deglycosylation solution has a price. Any possibility of determining the structure of the glycan on a specific peptide will be lost. Glycan characterization at specific sites in proteins will require another set of experiments where the identified protein is purified first, then tryptic digested, and the individual glycopeptides isolated. Although this work is based on the hypothesis that some tumor markers are over-fucosylated, most of the glycopeptides selected by the LTA column will not be tumor markers. Fucosyltransferases are found in many tissues. It would only be due to overexpression of these enzymes during tumorogenesis that fucosylation would occur at an abnormal level. The
research articles
Figure 1. Reversed phase chromatogram of deglycosylated peptides. These peptides were obtained by (1) selecting fucosecontaining glycopeptides from a tryptic digest of a serum sample with an LTA affinity column and (2) deglycosylating all the selected peptides simultaneously with PNGase in a single treatment. The serum sample being examined was prepared by combining equal parts of 0.5 mL tryptic digests of serum from dogs with lymphoma before and after receiving chemotherapy. Elution of the reversed phase columns was achieved in two hr linear gradient with mobile phases ranging from 5% acetonitrile, 95% H2O, with 0.1% TFA to a final mobile phase containing 60% acetonitrile, 40% H2O, with 0.1% TFA.
LTA selection process simply enriches the sample in fucosylated peptides, while greatly reducing the number of nonglycosylated peptides. LTA will not select tumor markers specifically. Differential Changes in Concentration. After proteolysis, differential labeling according to sample origin, sample mixing, affinity selection, and deglycosylation, the sample was fractionated by reversed phase chromatography. It is seen from the reversed phase chromatogram (Figure 1) that the mixture of peptides selected by the LTA affinity column is relatively simple. Although serum is thought to contain at least 10 000 proteins,51 the number containing R-L-fucosylation selected by the LTA affinity column was much smaller than the 300 000 to 500 000 peptides a tryptic digest of serum would potentially contain. Sample simplification is one of the very positive attributes of affinity selecting substances with a low abundance structural feature. Another is that analytes can be concentrated orders of magnitude. Fractions collected from the reversed-phase column (Figure 1) were analyzed by MALDI-MS. Typical mass spectra from collected fractions are seen in Figures 2 and 3. The deuterated isoforms of peptides are from an animal with untreated lymphosarcoma whereas the nondeuterated isoforms were taken from the same animal after chemotherapy. The ion pairs of peptide isoforms differed by 3 and 6 amu, depending on whether the C-terminal amino acid in the peptide isoforms was arginine or lysine, respectively. Isotope ratios used to assess changes in protein concentration were measured in terms of the relative peak height of the deuterated to nondeuterated isoforms. It is seen in the mass spectrum of the peptide ion pairs at m/z 1215.81 and 1218.85 (Figure 2) obtained from RPC fraction 15 that the concentration of this peptide decreased 5.2-fold in concentration in the treated animal. By contrast, little change in concentration was seen in the other peptide ion pair in the spectrum. The opposite effect was seen in the mass spectra in Figures 3 obtained from chromatography fraction 35. The peptide seen in isoforms at m/z 1355.86 and 1358.89 increased 1.9-fold with treatment. Journal of Proteome Research • Vol. 2, No. 6, 2003 621
research articles
Xiong et al.
Figure 2. MALDI-MS spectrum of acetylated isoforms of peptides isolated from differentially coded serum tryptic digests. These peptides were derived from fucose-containing glycopeptides. Glycopeptides containing fucose from tryptic digests of serum were affinity selected with LTA, deglycosylation with PGNase F, and fractionated with reversed phase chromatography prior to MS analysis. The relative peak height of the peptide isoforms shows little change the relative concentration of one peptide while the other decreased 5.2-fold during chemotherapy. Figure 4. Differential display of chemotherapy-induced changes in the relative concentration of fucose-containing glycopeptides selected from canine serum. A total of 57 peptide pairs were observed to change in concentration. Among these, one-third changed more than 2-fold in concentration after chemotherapy. Concentration changes are represented as percentage change when concentrations before and after chemotherapy are compared. Peptides with a positive change decreased in concentration during therapy where those with negative values increased.
Figure 3. MALDI-MS spectrum of acetylated isoforms of peptides isolated from differentially coded serum tryptic digests. These peptides were derived from fucose-containing glycopeptides. Glycopeptides containing fucose from tryptic digests of serum were affinity selected with LTA, deglycosylation with PGNase F, and fractionated with reversed phase chromatography prior to MS analysis.
Sixty fractions from the reversed phase column were analyzed for chemotherapy-induced changes in peptide concentration (Figure 4). Proteins both decrease and increased in concentration during chemotherapy. Change in this Figure is represented in percentage. Peptides of higher concentration initially than after chemotherapy were plotted as having undergone a positive percentage change in the Figure. In contrast, the percentage of those peptides of lower concentration initially than post-treatment is indicated as a negative. This 622
Journal of Proteome Research • Vol. 2, No. 6, 2003
Figure 5. Tandem mass spectrum derived by collision-induced dissociation of the (M+2H)2+ precursor, m/z 1018. The amino acids are represented in their single-letter code. The sequence identification of the peptide and database search showed that this peptide was derived from a sodium channel alpha-subunit, with the sequence RQNNVKNISSIYIKEGDK.
manner of representing change as being either positive or negative was done for ease of graphical representation, not for any scientific reason. A total of 57 peptides were observed to change in concentration. Among these, one-third changed more than 2-fold in
research articles
Comparative Proteomics of Glycoproteins Table 1. Sequence Identification and in the Relative Concentration of Peptides during Chemotherapy pep MW (Da)
sequence identified
parent protein
2662.91 2821.45 3404.50 3205.38 2450.27 3034.22 3324.21 3882.58 1037.37 1654.8 1635.78
QTTRMTDVDRSGTANGENWTR NALSMCISVVAPSAEGGLNLTTTFLR TAACTPTSCSGHGECIETINSSTCQCYPGF EVPREINMSCSGEPVFGAVCTFACPEGW SSMANKTCSMNEIKKSTLKDF WLWINSTDPAGQLQWLVGELQAAEDR GAIVAVTGDDVNDSPALKKADIGVAMGIAGSDAAK MAPWPHGNGSVASWPAAPTPTPDAANTSGLPGAPWAVAL TVQYQNEL DVFLGTFLYEYSR DVFLGTFLYEYAR
CD44 variant prostaglandin-D synthase E-selectin precursor E-selectin interleukin-4 acid sphingomyelinase H+/K+-exchanging ATPase alpha chain beta-3-adrenergic receptor Rab22a protein albumin-(Canis familaries) albumin-(Canis familaries)
concentration after chemotherapy. Peptides showing large changes in concentration are identified in Table 1. Three assumptions were made relative to the data in this table. One was that changes in the concentration of peptides corresponded directly to changes in the concentration of their protein parent between the treated and untreated states. A second was that no peptide that changed less than 2-fold would be considered a potential cancer marker. This decision was arbitrary and may be shown after more extensive studies not to be valid. The third was that chemotherapy itself could induce changes in protein concentration during treatment and that only those proteins which had previously been found in some type of cancer would be considered as potential cancer markers. This decision was based on the fact that none of the peptides, or their protein parents, which increased in concentration during chemotherapy has ever been described in the literature as being associated with cancer. It will be noted in the “isotope ratio” column of Table 1 that several of the peptides were given the designation “singlet”. This means that in the animal being studied the peptide appeared as a single cluster of ions instead of the doublet cluster separated by 3 or 6 amu and there can be no computation of relative change. These singlet clusters can originate in several ways. One would be due to a very large change in concentration of the parent protein between the control and experimental samples. Another would be from amino acid variations in the primary structure of proteins when samples from two individuals are being compared. This was the case with the two albumin peptides at m/z 1654.8 and 1635.8 in Table 1. The sequences of the two peptides are DVFLGTFLYEYSR and DVGLGTEFLYEYAR, respectively. They vary at a single site. The samples that produced these peptides came from different animals and will be the subject of a future paper on single amino acid polymorphism in canine proteins. Protein Identification. Peptide identification was based on tandem mass spectrometry. The parent ion of peptides observed to change more than 1.5-fold in concentration was selected with a quadrapole and subjected to collision-induceddissociation (CID). Product ions formed from the collision cell were then analyzed in the reflectron time-of-flight portion of the mass spectrometer. Spectra from the second dimension of mass spectrometry were analyzed both manually and with the Mascot software interface for searching databases. Typically the NCBI, SWISS-PROT, and OWL databases were searched through Mascot. Databases were searched for both tryptic and chymotryptic peptides with up to 9 miscleavages since proteolysis is often incomplete and chymotrypsin activity can occasionally be found in the commercially available TPCK-treated trypsin. The search query was instructed to consider other mammalian
isotope ratio (d3:d0)
+8.02 +7.26 +2.68 +2.66 singlet -2.83 singlet singlet singlet singlet
proteins because the set of canine proteins in databases is incomplete. A number of amino acid modifications were considered in database searches such as the fact that all cysteine residues were carboxymethylated, lysine and Nterminal amino groups were acetylated, and methionine could be oxidized. The possibility of sodium cationization and asparagine deamidation was also included in searches. Sequences of the peptides identified are given in the Table 1. The glycosylation sequences Asn-X-Ser or Asn-X-Thr expected in N-linked glycoproteins are highlighted with bold underscoring in Table 1. It is important to note that a few proteins in Figure 4 that changed significantly in concentration are not seen in Table 1. The peptide pair at m/z 1215.81 and 1218.85 in Figure 2 is an example. Although good MS/MS sequence data was obtained on this peptide, it could not be found in any of the known databases. This could be for several reasons. One would be that the parent canine protein has not previously been sequenced and differs in sequence from other mammalian proteins of the same function in databases. A second reason could be that a mutation in the animal from which this sample was taken caused the peptide to be structurally polymorphic. It is frequently the case in the proteomics of blood samples that abundant proteins interfere and are removed before beginning a search for low abundance markers. This increases the complexity of the analysis and takes additional time. Because a relatively specific affinity chromatography selector was used in these studies and none of the abundant proteins in serum were thought to contain the structural feature being selected, abundant proteins were not removed before the analysis. It will be noted that some of the peptides in Table 1 contain neither the glycosylation sequence Asn-X-Ser nor AsnX-Thr. This indicates that they could not have been derived from N-linked glycopeptides. Moreover, most of them seem to have come from high abundance proteins, such as serum albumin, hemoglobin, and cytochrome P-450. Obviously, they were nonspecifically selected by the lectin column. This was probably due to their inordinately high concentration in samples. Although nonspecific binding did not seem to interfere with glycopeptide identification in these studies, high abundance, nonspecifically bound peptides could be a problem if they coelute with peptides of interest and either suppressed their ionization or are of the same mass and interfere with MS/ MS sequencing. Selecting with the LTA column at the glycoprotein level followed by glycopeptide selection again at the peptide level after trypsin digestion would probably have circumvented the selection of these nonglycopeptides. A number of proteins were identified in cancer patients that changed in concentration less than 2-fold with treatment. These Journal of Proteome Research • Vol. 2, No. 6, 2003 623
research articles included regulatory proteins such as ATPase, membrane receptors, and channel proteins. It is difficult to know in an animal receiving chemotherapy whether changes of this magnitude are closely connected to the disease. Proteins that change more than 2-fold in concentration during therapy would be more desirable disease marker candidates. The five proteins in Table 1 that decreased more than 2-fold in concentration during chemotherapy were endothelial selectin (E-selectin), E-selectin precursor, a CD44 variant, prostaglandin D synthase, and interleukin-4 (singlet in heavy isotopic form). Again, it is noted that some proteins also increased in concentration during chemotherapy. These changes are thought to be triggered by the chemotherapy itself and to have nothing to do with cancer. Although there is no direct proof for this hypothesis in these studies, none of the proteins observed to increase in concentration during chemotherapy have been associated with cancer in the literature. Biological Function of the Identified Glycoproteins. CD44 plays a known role in cancer. Although CD44 is constitutively expressed in canine lymphatic tissue, expression is reportedly enhanced in activated lymph nodes.52 Invasion of endothelial cells by other cells carrying CD44 has been correlated with glycosylation of the protein.53 CD44 is a glycoprotein with a single polypeptide chain composed of a distal extracellular domain, a membrane-proximal region, a transmembranespanning domain, and a cytoplasmic tail. The primary structure (with the exception of the membrane-proximal region) displays a high degree of interspecies homology. It has been assigned many different functions relating to cell adhesion, cell migration, lymphocyte activation, and metastatic behavior in cancer.54 Genomic organization of CD44 is based on twenty exons,55 the first five and last five of which are constant. In contrast, the 10 exons located between these regions can be alternatively spliced during synthesis, resulting in the generation of a variable region. Differential splicing in the 10 variable exon regions has been reported in the case of cancer,56 as was found in the studies reported here. Increased sialylation of CD44 in cancer cells seems to reduce its hyaluronate-binding capacity, and may together with expression of the isoform that contains exons 4-7, lead to a metastasizing phenotype.47 It has been reported that for some human neoplasms (carcinomas of the digestive tract, non-Hodgkin’s lymphomas, thyroid carcinomas, and others) there is a correlation between the particular pattern of CD44 variants produced and clinicopathological parameters of the tumor, such as grade, stage, presence of metastases, and potential for survival.57 The role of aberrant fucosylation of CD44 in cancer progression and treatment observed here is yet to be determined. E-Selectin is known to be a fucosylated glycoprotein58 as found in this study. Overexpression of E-selection has been associated with multiple types of cancer.59 A question arising from this study is whether the decrease in fucosylated Eselection during therapy was due to a decrease in expression or a down-regulation of fucosylation. The natural role of E-selectin is to mediate the rolling of circulating leukocytes on vascular endothelial cells.60 Up-regulation of both fucosylated selectin and fucosylated CD44 (Table 1) along with the fact that E-selectin is a fucose-specific lectin that binds to fucosylated glycoforms of CD4461 suggests that aberrant fucosylation of CD44 provides an essential structural element that enables these two proteins in concert to enhance tumor cell migration. 624
Journal of Proteome Research • Vol. 2, No. 6, 2003
Xiong et al.
Prostaglandin D synthase (PGDS) was another protein observed in these studies to be substantially down-regulated during chemotherapy. It has never been reported in either nonHodgkin’s lymphoma or canine lymphosarcoma, although it is overexpressed in human brain tumors and breast tumors,62 in addition to ovarian tumors63 and has been suggested as a general cancer marker.64 Again, the question raised in this study is whether the changes in fucosylated PGDS concentration are the result of altered expression or glycosylation. PGDS is a 30 kDa glycoprotein found in many body fluids and tissues and is responsible for the biosynthesis of prostaglandin D2.65 N-Glycosylation has been observed at Asn29 and Asn56 with almost quantitative R(1-6) fucosylation at the proximal Nacetylglucosamine on at least one glycosylation site.66 Prostaglandin D analogues have antitumor activity, increase the number of white blood cells in vivo, and synchronize the cell cycle of tumor cells.67 The mechanism of PGDS in cancer progression is not yet clear. Interleukin-4 (IL-4) is a similar case. It has never been association with the various types of lymphomas grouped within Non-Hodgkin lymphoma or with canine B- or T-cell lymphomas, but IL-4 production increases in malignant human T cells.68 IL-4 is a T cell-derived glycoprotein of 129 amino acids in humans that is glycosylated on Asn38 and is a B cell stimulatory factor.69-70 Although it is clear that stimulating B cell growth would be fortuitous in the progression of a lymphosarcoma, it is not apparent how fucosylation could play a role.
Conclusion It may be concluded from these studies of fucosylated glycoproteins that when the GIST coding protocol for comparative proteomics and quantification is coupled with (1) a lectin affinity selector specific for fucosylation, (2) the selected peptides are deglycosylated, and (3) the deglycosylated peptides are analyzed by mass spectrometry it is possible to recognize, identify, and create a quantitative differential display of proteins associated with disease remission, in this case cancer. It is further concluded that application of this method to blood samples allows multiple disease markers to be recognized and quantitatively compared simultaneously as a function of disease progression or treatment. Although all of the glycoproteins that decreased in concentration during chemotherapy had been previously associated with some type of cancer, this is the first time they had all been simultaneously recognized and displayed in a single type of cancer and associated with aberrant fucosylation. Although the studies reported here were directed toward the quantification of fucosylation, it is concluded that the coupled GIST-lectin-deglycosylation protocol may be generalized and used for comparative proteomics of many types of glycoproteins. It is further concluded that the procedures described here have the potential to (1) aid disease detection in many types of glycopathology, (2) monitor disease progression and treatment, (3) signal the reoccurrence of disease when therapy fails, and (4) even identify patterns of disease markers that are unique to individuals. Elements essential for this to occur are (1) the availability of lectins for targeting multiple types of glycosylation and (2) hydrolytic enzymes that cleave either individual carbohydrate residues from the glycan or the whole oligosaccharide from the polypeptide. The fact that a wide variety of lectins and glycosidases have been described in the
research articles
Comparative Proteomics of Glycoproteins
literature and are commercially available will make future studies in lectin based comparative proteomics relatively easy.
Acknowledgment. The authors gratefully acknowledge financial support from the National Institute of Health (Grant No. GM-59996). References (1) Rademacher, T. W.; Parekh, R. B.; Dwek, R. A. Annu. Rev. Biochem. 1988, 57, 785-838. (2) Vlassara, H.; Palace, M. R. J. Intern. Med. 2002, 251, 87-101. (3) Kirkham, S.; Sheehan, J. K.; Knight, D.; Richardson, P. S.; Thornton, D. J. Biochem. J. 2002, 361, 537-546. (4) Reddy, G. K.; Dhar, S. C. Agents Actions 1988, 25, 63-70. (5) Ahmed, N.; Argirov, O. K.; Minhas, H. S.; Cordeiro, C. A.; Thornalley, P. J. Biochem. J. 2002, 364, 1-14. (6) Seberger, P. J.; Chaney, W. G. Glycobiology 1999, 9, 235-241. (7) Sato, M.; Toyozaki, T.; Odaka, K.; Uehara, Tomoya; A.; Yasushi, H. H.; Yoshida, K.; Imanaka-Y. K.; Yoshida, T.; Hiroe, M.; Tadokoro, H.; Irie, T.; Tanada, S.; Komuro, I. Circulation. 2002, 106, 1397-1402. (8) Sebekova, K.; Blazicek, P.; Syrova, D.; Krivosikova, Z.; Spustova, V.; Heidland, A.; Schinzel, R. Kidney Int. 2001, 78, S58-S62. (9) Spinale, F. G.; Coker, M. L.; Heung, L. J.; Bond, B. R.; Gunasinghe, H. R.; Etoh, T.; Goldberg, A. T.; Zellner, J. L.; Crumbley, A. J. Circulation 2000, 102, 1944-1949. (10) Puchelle, E.; de Bentzmann, S.; Zahm, J. M. Respiration 1995, 62, 2-12. (11) Matis, W. L.; Lavker, R. M.; Murphy, G. F. J. Invest. Derm. 1990, 94, 492-5. (12) Kobata, A. Glycoconjugate J. 1998, 15, 323-331. (13) Springer, G. F. J. Mol. Med. 1997, 75, 594-602. (14) Orntoft, T. F.; Vestergaard, E. M. Electrophoresis 1999, 20, 362. (15) Howe, H. L.; Wingo, P. A.; Thun, M. J.; Ries, L. A. G.; Rosenberg, H. M.; Feigal, E. G.; Edwards, B. K. J. Nat. Cancer Inst. 2001, 93, 824-842. (16) Kinzler, K. W.; Vogelstein, B. Cell, 1996, 87, 159-170. (17) Jessani, N.; Liu, Y.; Mark, H.; Cravatt, B. F. Proc. Natl. Acad. Sci. USA 2002, 99, 10 335-10 340. (18) Dennis, J. W. In Fukuda, M., Ed., CRC Press: London, 1995, pp 161-194. (19) Dennis, J. W.; Laferte, S.; Waghorne, C.; breitman, M. L.; Kerbel, R. S. Science 1987, 236, 236-239. (20) Seberger, P. J.; Chaney, W. G. Glycobiology 1999, 9, 235-241. (21) Demetriou, M.; Nabi, I. R.; Coppolino, M.; Dehar, S.; Dennis, J. W. J. Cell Biol. 1995, 130, 383-392. (22) Bhavandandan, V. P.; Furukawa, K. In Biology of the Sialic Acids; Rosenberg, L. A., Ed.; Plenum Press: New York, 1995, pp 145. (23) Dwek, M. V.; Ross, H. A.; Leathem, A. J. Proteomics 2001, 1, 756762. (24) Turner, G. A. Adv. Exp. Med. Biol. 1995, 276, 231. (25) Haselhorst, T.; Weimar, T.; Peters, T. J. Am. Chem. Soc. 2001, 123, 10 705. (26) Smets, L. A.; Van Beek, W. P. Biochim. Biophys. Acta, 1984, 738, 237-249. (27) Hakomori, S. I. Adv. Cancer Res. 1989, 52, 257-231. (28) Turner, G. A. Clin. Chim. Acta, 1992, 298, 149-171. (29) Takada, A.; Ohmori, K.; Yoneda, T.; Tsuyuoka, K.; Hasegawa, A.; Kiso, M.; Kannagi, R. Cancer Res. 1993, 53 (2), 354-361. (30) Goodarzi, M. T.; Turner, G. A. Clin. Chim. Acta 1995, 236, 161. (31) Sato, M.; Yonezawa, S.; Uehara, H.; Arita, Y.; Naramats, T. Differentiation, 1986, 30, 211. (32) Tan, L. Ph.D. Thesis, Purdue University, Lafayette, IN, 1998, p 78. (33) Geng, M.; Ji, J.; Regnier, F. E. J. Chromatogr. A, 2000, 870, 295. (34) Xiong, L.; Regnier, F. E. J. Chromatogr. B, 2002, 782, 405-418 (35) MacEwen, E. G. Cancer Met. Rev. 1990, 9, 125-135. (36) Strandstrom, H. V.; Bowen, J. M. Advances in Comparative Leukemic Research. New York, Elsevier 1998, 447-450. (37) Knapp, D. W.; Waters, D. J. Mol. Med. Today. 1997, 3, 8-11.
(38) Waters, D. J.; Coakley, F. V.; Cohen, M. D.; Davis, M. M.; Karmazyn, B.; Gonin, R.; Hanna, M. P.; Knapp, D. W.; Heifetz, S. A. J. Comput. Assist. Tomogr. 1998, 22, 236-240. (39) Hahn, K. A.; Bravo, L.; Adams, W. H.; Frazier, D. L. In Vivo 1994, 8, 133-143. (40) Withrow, S. J.; Powers, B. E.; Straw, R. C.; Willkins, R. M. Clin. Orthop. 1991, 270, 159-168. (41) Teske, E. Vet. Quart. 1994, 4, 209. (42) Howe, H. L.; Wingo, P. A.; Thun, M. J.; Ries, L. A. G.; Rosenberg, H. M.; Feigal, E. G.; Edwards, B. K. J. Nat. Cancer. Inst. 2001, 93, 824-842. (43) Cartwright, R.; Brincker, H.; Carli, P. M.; Clayden, D.; Coebergh, J. W.; Jack, A.; MaNally, R.; Morgan, G.; De Sanjose, S.; Tumino, R.; Vornanen, M. Eur. J. Cancer. 1999, 35, 627-633. (44) Morgan, G.; Vornanen, M.; Puitinen, J.; Naukkarinen, A.; Brincker, H.; Olsen, J.; Coeburgh, J. W.; Vrints, L. W.; Clayden, D.; McNally, R.; Jack, A.; Carli, P. M.; Petrella, T.; Tomino, R.; D’Lollo, S.; Barchielli, A.; Cartwright, R. Ann. Oncol. 1997, 8, S49-S54. (45) Vonderharr, M. A.; Morrison, W. B. Lymphosarcoma. Morrison, W. B. In Caner in Dogs and Cats; Medical and Surgical Management, 1st ed; Lippincott, Williams & Wilkins: Maryland, 1998, 667-695. (46) Armitage, J. O.; Non-Hodgkin’s lymphomas. Handin, R. I.; Lux, S. E.; Stossel, T. P. Blood: Principles and Practice of Hematology, 1st ed; J. B. Lippincott: Philadelphia, 1995, 851-884. (47) Harris, N. L.; Jaffe, E. S.; Diebold, J.; Flandrin, G.; MullerHermelink, H. K.; Vardiman, J.; Lister, T. A.; Bloomfield, C. D. Histopathology 2000, 36, 69-87. (48) Onions, D. E.; Owen, L. N.; Bostock D E Internat. J. Cancer. 1978, 22, 503-7. (49) Xiong, L. Regnier, F. E. J. Chromatogr., in press, 2003. (50) Regnier, F. E.; Riggs, L.; Zhang, R.; Xiong, L.; Liu, P.; Chakraborty, A.; Seeley, E.; Sioma, C. J. Mass Spectrom. 2002, 37, 133-145. (51) Anderson, N. L.; Anderson, N. G. Mol. Cell. Proteom. 2002, 1, 845867. (52) Katoh, S.; Zhang, Z.; Oritani, K.; Shimozato, T.; Kincade, P. W. J. Exp. Med. 1995, 182, 419-429. (53) Goupille, C.; Halloouin, F.; Meflah, K.; Le, P. J. Glycobiology 1997, 7, 221-229. (54) Gunthert, U.; Stander, R.; Mayer, B.; Terpe, H. J. Finke, L.; Friedrichs, K. Cancer. Surv. 1995, 24, 19-42. (55) Naor, D.; Sionov, R. V.; Ish-Shalom, D. Adv. Cancer Res. 1997, 71, 241-319. (56) Guenthert, U. Stauder, R. Mayer, B. Terpe, H.-J. Finke, L. Friedrichs, K. Cancer Surv. 1995, 24, 19-42. (57) Rudzki, Z.; Jothy, S. Mol. Pathol. 1997, 50, 57-71. (58) Tyrrell, D.; James, P.; Rao, N.; Foxall, C.; Abbas, S.; Dasgupta, F.; Nashed, M.; Hasegawa, A.; Kiso, M.; Asa, D. Proc. Natl. Acad. Sci. (USA) 1991, 88, 10 372-10 376. (59) Larkin, M.; Ahern, T. J.; Stoll, M. S.; Shaffer, M.; Sako, D.; O’Brien, J.; Yuen, C. T.; Lawson, A. M.; Childs, R. A.; Barone, K. M. J. Biol. Chem. 1992, 267, 13 661-13 668. (60) Listinsky, J. J.; Siegal, G. P.; Listinsky, C. M. Am. J. Clin. Pathol. 1998, 110, 425-440. (61) Dimitroff, C. J.; Lee, J. Y.; Rafii, S.; Fuhlbrigge, R. C.; Sackstein, R. J. Cell Biol. 2001, 153, 1277-1286. (62) Su, Bing; Guan, Ming; Zhao, Ruijiao; Lu, Yuan. Clin. Chem., & Lab. Med. 2001, 39, 1198-1203. (63) Su, B.; Guan, M.; Zhao, R.; Lu, Y. Clin. Chem. and Lab. Med. 2001, 39, 1198-1203. (64) Xu, S.; Venge, P. Biochim. Biophys. Acta 2000, 1482, 298-307. (65) Saso, L.; Leone, M. G.; Sorrentino, C.; Giacomelli, S.; Silvestrini, B.; Grima, J.; Li, J. C.; Samy, E.; Mruk, D.; Cheng, C. Y. Biochem. & Mol. Biol. Intl. 1998, 46, 643-656. (66) Hoffmann, A.; Conradt, H. S.; Gross, G.; Nimtz, M.; Lottspeich, F.; Wurster, U. J. Neurochem. 1993, 61, 451-456. (67) Arai, Y.; Wakatsuka, H.; Sasaki, Y. Eur. Pat. Appl. 1986, EP172963. (68) Vowels, B. R.; Cassin, M.; Vonderheid, E. C.; Rook, A. H. J. Invest. Dermat. 1992, 99, 90-4. (69) Defrance, T.; Vanbervliet, B.; Aubry, J. P.; Takebe, Y.; Arai, N.; Miyajima, A.; Yokota, T.; Lee, F.; Arai, K.; de Vries, J. E.; J. Immunol. 1987, 139, 1135-1141. (70) Tsarbopoulos, A.; Birendra, N.; Nagabhushan, T.; Covey, T. R. J. Mass Spectrom. 1995, 30, 1752-1763.
PR0340274
Journal of Proteome Research • Vol. 2, No. 6, 2003 625
