α1-Acid glycoprotein fucosylation as a marker of carcinoma progression and prognosis

Authors


Abstract

BACKGROUND

Serum α1-acid glycoprotein (AGP), an acute-phase protein secreted by the liver, carries α(1,3)-fucosylated structures on its 5 highly branched, N-linked sugar chains.

METHODS

Serum AGP levels in patients with various types of malignancies (n = 214 patients) were measured using an enzyme-linked immunosorbent assay with anti-AGP antibody. To investigate glycoforms that differed in their degree of branching and extent of fucosylation, serum AGP samples were analyzed by crossed affinoimmunoelectrophoresis (CAIE) with concanavalin A, and Aleuria aurantia lectin (AAL), and anti-AGP antibody.

RESULTS

A significant difference (P < 0.001) in serum AGP levels was observed in preoperative patients compared with levels in the healthy control group, but the levels in individual patients did not reflect their clinical status. Conversely, it was found not only that the patterns of AGP glycoforms differed widely in the patient group compared with the healthy control group, but they also changed depending on each patient's clinical status. Furthermore, AGP glycoforms seemed to be appropriate markers of disease progression and prognosis according to follow-up studies of 45 patients during prolonged preoperative and postoperative periods.

CONCLUSIONS

Patients with advanced malignancies who had AGP glycoforms that contained highly fucosylated triantennary and tetraantennary sugar chains for long periods after surgery were likely to have a poor prognosis. However, patients who had AGP glycoforms without such changes were expected to have a good prognosis. Cancer 2004. © 2004 American Cancer Society.

α1-Acid glycoprotein (AGP; orosomucoid) is a major serum glycoprotein with a molecular weight of 41–43 kiloDaltons and with highly branched, N-linked glycans.1 Although various functions of the AGP molecule have been proposed, and its potential physiologic significance as an acute-phase protein with diverse immunomodulating effects has been investigated,2–5 the exact functions of AGP remain unknown. Previously, it was shown that the glycoforms of AGP, which consist of complex diantennary, triantennary, and tetraantennary glycan chains, changed during acute and chronic inflammation, pregnancy, estrogen treatment, cancer, liver diseases, and autoimmune diseases, like rheumatoid arthritis and Graves disease.1, 6–12 Accordingly, it has been demonstrated that the degree of branching and α(1,3)-fucosylation on the triantennary and tetraantennary chains on AGP has various biochemical functions, and the role of sialylated Lewis X (LeX) antigen expressed on AGP molecule also has been discussed.11, 13–18

The glycoforms of AGP have been analyzed using crossed affinoimmunoelectrophoresis (CAIE),19, 20 high pH anion-exchange chromatography,10 lectin enzyme-linked immunosorbent assay (ELISA),21 capillary electrophoresis,22 and lectin immunosensors.23 It has been demonstrated clearly with the CAIE technique in the presence of concanavalin A (Con A) that AGP can be subfractionated into at least three glycoforms that differ in their content of diantennary versus triantennary and tetraantennary, N-linked glycans based on the binding specificity of Con A for diantennary glycans.1, 24 When the fucose-binding Aleuria aurantia lectin (AAL)25 was used in the same CAIE technique instead of Con A lectin, the degrees of α(1,3)-fucosylation on AGP also could be estimated.10

In our previous studies, we found that serum α(1,3)-fucosyltransferase activity was elevated in patients with highly malignant or metastatic tumors,26–29 and an elevated level of this enzyme may be a sensitive and specific marker both for the diagnosis of malignancy and for follow-up studies during therapy in patients with advanced disease. Recently, it was determined that the major α(1,3)-fucosyltransferase activity in human serum is encoded by the FUT6 gene, and the missense mutation that occurs in the FUT6 gene results in the deficiency of serum α(1,3)-fucosyltransferase activity and the absence of α(1,3)-fucosylated AGP.30–32 Therefore, it is likely that serum α(1,3)-fucosyltransferase encoded by the FUT6 gene and originated in the liver is responsible for α(1,3)-fucosylation of AGP synthesized in the liver. More recently, it was suggested that the origin of elevated α(1,3)-fucosyltransferase in serum from patients with carcinoma is the tumor rather than the liver.33 Furthermore, it was demonstrated that AGP fucosylation is a useful marker for identifying patients with cirrhosis among patients with liver disease34 or rheumatoid arthritis.35

The objective of the current study was to investigate whether the AGP glycoforms in patients with carcinoma change with their clinical status and, if so, then whether these glycoforms may be used as markers of disease progression or prognosis. We assayed serum levels of AGP from patients with various types of carcinoma and determined the variations in their serum AGP glycoforms using the CAIE technique with Con A, AAL, and an anti-AGP antibody over a long postsurgical period.

MATERIALS AND METHODS

Materials

Con A lectin (type V), methyl-α-D-mannoside, methyl α-D-glucoside, α-L-fucose, human serum albumin, diethyl barbituric acid, calcium lactate, acrylamide, bovine serum albumin (BSA), Coomassie Brilliant Blue R250, sorbitol, and thimerosal were purchased from Sigma. Antihuman AGP rabbit serum and human serum protein calibrator were from DAKO (Carpinteria, CA). Antihuman AGP rabbit serum also was obtained by immunization of rabbits with purified human AGP, as described below. AAL was obtained and purified as described previously.25 Diethylaminoethanol (DEAE)-Sepharose CL-6B and SP-Sepharose CL-6B were obtained from Amersham (Uppsala, Sweden). GDP-fucose was from Calbiochem (San Diego, CA). Agarose (standard Low-m) and Affi-Gel protein A were from Bio-Rad (Hercules, CA) and agarose 361E was from Funakoshi (Tokyo, Japan). GelBond Film was from BioWittaker Molecular Applications (Rockland, MD). H type 2 and LeY Syntagen were from Chembiomed (Edmonton, Canada). Anti-LeY antibody (AH-6, mouse immunoglobulin M) was labeled with 125I, and anti-BSA was prepared as described previously.36

Study Population

Blood samples (n = 214 samples) were obtained from patients with various types of malignancies who were admitted to Gunma University Hospital (Maebashi, Japan) along with the guidelines for informed consent. Blood samples also were obtained from randomly selected volunteers as a healthy control group (n = 76 samples). Each serum sample was stored at − 80 °C until use. Samples from 45 patients with various malignancies (Table 1) were collected periodically for follow-up investigations during long preoperative and postoperative periods.

Table 1. Characteristics of Patients who were Followed for Crossed Affinoimmunoelectrophoresis Analysis
Type of carcinoma/stageAge (yrs) (mean ± SD)No. of patients
TotalMaleFemale
  • SD: standard deviation.

  • a

    Stage II breast carcinoma and recurrent splenic carcinoma.

Esophagus53.4 ± 9.8550
 Stage II 440
 Stage IV 110
Gastric64.5 ± 9.81899
 Stage I 330
 Stage II 514
 Stage III 532
 Stage IV 523
Colorectal65.0 ± 8.9871
 Stage I 431
 Stage III 220
 Recurrent 220
Lung68.9 ± 9.3550
 Stage I 330
 Stage III 110
 Recurrent 110
Hepatic65.5 ± 9.0431
 Stage II 110
 Stage III 321
Pancreatic70.3 ± 9.1312
 Stage II 202
 Stage IV 110
Othersa51.5 ± 3.5202
 Stage II 101
 Recurrent 101

ELISA of Serum AGP

The AGP levels in serum samples were measured by a sandwich-type ELISA using antihuman AGP and horseradish peroxidase-conjugated antihuman AGP. Each well of a 96-well microtiter plate (Nunc) was coated for 24 hours at 4 °C with 100 μL of antihuman AGP rabbit serum (DAKO) at a concentration of 2 μg/mL in the coating buffer, pH 9.6, containing 0.015 M Na2CO3, 0.035 M NaHCO3, and 0.02% NaN3. After washing with 0.01 M phosphate buffered saline (PBS), pH 7.4, containing 0.05% Tween 20 and 0.01% thimerosal (washing buffer), the wells were blocked overnight at 4 °C with 200 μL of PBS containing 25% BlockAce (Dainippon Pharmaceutical, Osaka, Japan), 10% sorbitol, and 0.05% thimerosal (blocking buffer) per well. The plates were dried at 25 °C after aspirating the blocking buffer and washing with the buffer and were stored at 4 °C until use. The plates were washed with the washing buffer before use; then, 100 μL PBS containing 10% BlockAce and 0.05% thimerosal (staining buffer) were added to each well. Ten microliters of sample or standard AGP were added to the staining buffer in each well and incubated for 2 hours at 25 °C. After washing the wells with the washing buffer, 100 μL of horseradish peroxidase-conjugated anti-AGP rabbit serum at a predetermined dilution were added to each well followed by incubation for 2 hours at 25 °C. The wells were washed with washing buffer. Next, 100 μL of coloring reagent, 0.1% o-phenylendiamine dihydrochloride, and 0.05% H2O2 in 0.1 M citrate phosphate buffer, pH 4.8, were added to each well for color development. After incubation for 10 minutes at room temperature, the color development was stopped by adding 100 μL of 2 N H2SO4 to the wells. The absorbance of each well at 490/650 nm was measured with a Well Reader SK603 (Seikagaku, Tokyo, Japan). The AGP concentration of samples was calculated with a second-order polynomial formula derived from the net absorbance of the standard AGP antigen. The cut-off value for serum AGP (mean + standard deviation [SD]) calculated from the healthy control group was 807 μg/mL.

CAIE

CAIE was performed according to previous methods13, 19 with a slight modification. For the 2-dimensional immunoelectrophoresis method, serum samples containing approximately 3 μg of AGP were subjected to electrophoresis through a first-dimension consisting of 7.9% (weight/volume) acrylamide gel containing 12 mg of Con A or 300 μL of AAL preparation (hemagglutination titer against O type blood cells was 1:512). Electrophoresis of the gel was carried out at 100 V for 3 hours. In the second dimension, 1% agarose gel containing 360 μL of anti-AGP antibody was separated from the first dimension gel by a 1-cm-wide intermediate gel containing 70 mM methyl α-D-mannoside and methyl α-D-glucoside (for Con A) or 70 mM methyl α-D-mannoside, methyl α-D-glucoside, and 8 mM α-L-fucose (for AAL) to decompose the complexes with the lectin. Electrophoresis in the first dimension was performed from right to left and, in the second dimension, from bottom to top. The electrophoresis was carried out for 18 hours at 2 V/cm and 5 °C. Then the resulting precipitation lines were visualized by Coomassie Brilliant Blue R250 staining.

The areas under the curve were determined with an area measurement program, NIH Image software (version 1.61; National Institutes of Health, Bethesda, MD), and the relative amount of each glycoform was expressed as a percentage of the total. In addition, all CAIE procedures were conducted with a standard AGP preparation (human serum protein calibrator; DAKO) in every electrophoresis as a control. In accordance with the patterns of the precipitation lines in CAIE both with Con A and with AAL, 3 glycoforms consisting of differently branched glycans (Fig. 1) were distinguished: not reactive with Con A (C0) (relative mobility, 0.53–1.00; no diantennary glycans and 5 triantennary or tetraantennary glycans), weakly reactive with Con A (Cw) (relative mobility, 0.32–0.53; 1 diantennary glycan and 4 triantennary or tetraantennary glycans), strongly reactive with Con A (Cs) (relative mobility, 0.0–0.32; ≥ 2 diantennary glycans and, maximally, 3 triantennary or tetraantennary glycans), not reactive with AAL (A0) (relative mobility, 0.68–1.00; not fucosylated), weakly reactive with AAL (Aw) (relative mobility, 0.40–0.68; 1 fucose residue per molecule), and strongly reactive with AAL (As) (relative mobility, 0.0–0.40; ≥ 2 fucose residues per molecule) (Fig. 2). In this study, the relative amounts of AGP glycoforms that were not retarded by Con A (%C0) were defined as the branching index. In addition, a fucosylation index was defined as the relative amounts of AGP glycoforms that were retarded by AAL, i.e., % (Aw + As). The cut-off values (mean + SD) of these 2 indices calculated in the healthy control group were 50.9 and 65.6, respectively (see below).

Figure 1.

Di-antennary, tri-antennary, and tetra-antennary glycans on serum α1-acid glycoprotein (AGP). All possibilities are shown in brackets. NeuAc: N-acetylneuraminic acid; Gal: galactose; GlcNAc: N-acetylglucosamine; Fuc: fucose; Man: mannose; Asn: asparagine.

Figure 2.

Subfractionations of α1-acid glycoprotein (AGP) present in healthy control serum into glycoforms differing in extent of branching (left) and fucosylation (right) of its glycans by crossed affinoimmunoelectrophoresis (CAIE) with concanavalin A (Con A) and AAL. Only the second dimension gels are shown; the application site was on the right-hand side. C0: AGP glycoproteins that did not react with Con A; A0: AGP glycoproteins that did not react with Aleuria aurantia lectin (AAL); Cw: AGP glycoproteins that reacted weakly to Con A; Cs: AGP glycoproteins that reacted strongly to Con A; Aw: AGP glycoproteins that reacted weakly to AAL; As: AGP glycoproteins that reacted strongly to AAL. For definitions of the electrophoretic mobility of the fractions and for further details, see Materials and Methods.

α(1,3)-Fucosyltransferase Assay

α(1,3)-Fucosyltransferase activity was measured by a sandwich-type immunoradiometric assay, as describe previously.36 The reaction mixture in a tube contained 50 μL of serum sample; 200 μL of buffer containing 16 μmol N-2-hydroxyethyl-piperazine-N′-2-ethane sulphonate-NaOH, pH 7.0; 2 μmol MnCl2; 0.4 μmol adenosine triphosphate; 4 μmol NaN3; 0.5 μg ovalbumin; 15 nmol GDP-fucose; and 0.1 nmol H type 2 (Fucα1,2Galβ1,4GlcNAcβ)-conjugated BSA (H type 2; Syntagen). After incubation at 37 °C overnight, an anti-BSA, antibody-coated bead was added to each reaction tube, and the mixture was placed on a shaker at room temperature for 2 hours. Then, the bead was washed 3 times with distilled water, and a 200 μL solution of 125I-labeled anti-LeY antibody containing 10 μmol phosphate, pH 7.0; 1 μg ovalbumin; and 0.1 μg NaN3 was added to the tube. The mixture was placed on a shaker at room temperature for 2 hours. The beads were washed three more times with distilled water, and the radioactivity remaining on the beads was measured with a γ-counter. Enzyme activity was expressed in arbitrary units (U/mL) compared with the standard enzyme prepared from human plasma and LeY hapten (Fucα1,2Galβ1,4[Fucα1,3]GlcNAcβ)-conjugated BSA (0.0–0.2 μmol LeY hapten per gram of BSA) as a standard product.

Statistical Analysis

The significance level of the correlation coefficient was evaluated with the t statistic by using StatView software (version 5.0; SAS Institute Inc., Cary, NC). Multivariate analyses for predicting patient prognoses were done by the discrimination analyses of patients' clinicopathologic background factors together with AGP levels and AGP glycoforms in serum samples by using the Excel software (version 11.0; Microsoft Inc., Redmond, WA). Regression analyses also were performed for the same samples between patients with a good prognosis and patients with a poor prognosis.

Immunization

Purification of AGP from pooled human plasma was conducted by a sequential chromatography using DEAE-Sepharose CL-6B and SP-Sepharose CL-6B ion-exchange columns (details of the purification procedure will be published elsewhere). Purified human AGP was emulsified with an equal volume of Freunds' complete adjuvant, and the mixture was injected subcutaneously into rabbits every 2 weeks. The rabbits were bled 3 weeks after the third injection. The antiserum was purified by affinity chromatography on a column of Affi-Gel protein A according to the manufacturer's instruction.

RESULTS

Serum AGP Levels in Patients with Malignancies

Serum samples from preoperative patients with various malignancies (n = 214 samples) and from the healthy control group (n = 74 samples) were assayed to determine serum AGP levels (Fig. 3). AGP levels in the patient group were significantly high compared with AGP levels in the healthy control group (mean ± SD: 888.3 μg/mL ± 558.8 μg/mL vs. 569.4 μg/mL ± 238.0 μg/mL; P < 0.001). Very low AGP levels (< 200 μg/mL) and high AGP levels (> 1000 μg/mL) were found in some patients and in some healthy controls, respectively. When AGP levels in the patient group were compared with tumor size, clinical stage, and status according to the tumor-lymph node-metastasis (TNM) classification system, no significant difference was seen, even though patients with advanced or recurrent disease and/or metastasis seemed to secrete high levels of AGP (data not shown). These observations seem to suggest that, although AGP is known as a marker of inflammatory state, its serum levels per se may not necessarily relate to disease progression.

Figure 3.

Levels of serum α1-acid glycoprotein (AGP) in patients with various types of carcinoma (ca.). Each vertical bar indicates the mean value of AGP levels.

Follow-up Study of Serum AGP Levels

The level of serum AGP, as a major acute-phase glycoprotein, in patients with malignant disease is expected to change both after surgery and with the patient's clinical status. Forty-five patients (Table 1) were followed to measure their AGP levels during the preoperative and postoperative periods. They were classified into 2 groups: patients with a good prognosis (34 patients) and patients with a poor prognosis (11 patients). No clinical recurrences were observed in the former group at the time of this report, although 15 of 34 patients were classified with advanced or metastatic disease at the time of surgery. Serum AGP levels in these patients with a good prognosis changed somewhat: In 21 of 34 patients (> 60%) who had a good prognosis, AGP levels increased shortly after surgery and then decreased slowly to below the cut-off level, which was set equal to the mean + SD (807 μg/mL) level in the healthy control group (Fig. 4Aa). However, serum AGP levels in the other patients (13 of 34) who had a good prognosis increased continuously to above the cut-off value (Fig. 4Ab). In contrast, 11 patients who underwent noncurative surgery also showed different changes in serum AGP levels, even though all 11 patients had recurrences and subsequently died during the observed period. In 9 of 11 patients who had a poor prognosis, serum AGP levels did not fall to below the cut-off level (Fig. 4Bb); however, in 2 patients, AGP levels decreased to below the cut-off level (Fig. 4Ba). Therefore, no obvious difference was seen between patients with respect to AGP levels and their clinicopathologic background after surgery. Furthermore, as described below, the fate of each patient could not be predicted by the serum AGP level per se. During treatment, elevated levels of AGP persisted for a long period in some patients and for a short period in other patients, perhaps indicating that an acute inflammatory condition similar to that seen during the period shortly after surgery prevailed.

Figure 4.

(A) Changes in serum α1-acid glycoprotein (AGP) levels in patients who had a good prognosis, including surgically achieved tumor-free patients with Stage II gastric carcinoma (a) and with Stage IIA esophageal carcinoma (b). (B) Among patients who had a poor prognosis, clinical recurrences occurred in patients with Stage IIIB colon carcinoma (a) and in patients with rectal carcinoma (b). The horizontal bars indicate the cut-off serum AGP level (807 μg/mL).

CAIE Patterns of AGP with AAL and Con A Lectin

Glycoforms of AGP were analyzed with a CAIE technique using Con A lectin or AAL as affinity molecules in the first-dimension gel and using anti-AGP in the second-dimension gel (Fig. 2). CAIE with Con A or AAL, as described previously,24 resulted in the detection of AGP glycoforms that differed in the degree of branching (C0, Cw, and Cs) and fucosylation (A0. Aw. and As) of their N-linked glycans. Each percentage of C0, Cw, Cs, A0, Aw, and As was calculated to determine the reproducibility of CAIE using results from the standard serum (human serum protein calibrator; DAKO) (n = 76 samples) and were determined as follows: 46.7% ± 3.2% (C0), 39.7% ± 3.6% (Cw), 13.6% ± 3.1% (Cs), 40.1% ± 3.9% (A0), 41.7% ± 3.9% (Aw), and 18.2% ± 3.2% (As). The relative amounts of glycoforms on the AGP from healthy controls (n = 74 samples) also were determined as follows: 45.7% ± 5.2% (C0), 39.0% ± 3.1% (Cw), 15.3% ± 5.7% (Cs), 41.6% ± 5.0% (A0), 37.9% ± 3.7% (Aw), and 20.4% ± 3.6% (As). These results from CAIE of standard serum and sera from healthy controls suggest that the current method allows quantitative and reproducible determination of AGP glycoforms.

Types of AGP Glycoforms Based on CAIE Patterns of AGP

In our previous studies,1, 24 we demonstrated that relative amounts of C0 and Aw + As, defined as the branching index and the fucosylation index, consisted of triantennary and tetraantennary glycans and α(1,3)-fucosylated glycans, respectively. To characterize glycoforms of AGP determined by CAIE with a pair of Con A and AAL, four types indicating the degrees of branching and fucosylation of each glycoform were set up with the aid of two indices. Those 4 types were characterized according to the levels of their branching and fucosylation indices using 2 cut-off values (50.9% and 65.6%), which were calculated from the healthy control group (Fig. 5) (type A: branching index, < 50.9%; fucosylation index, < 65.6%; type B: branching index, < 50.9%; fucosylation index, ≥ 65.6%; type C: branching index, ≥ 50.9%; fucosylation index, < 65.6%; type D: branching index, ≥ 50.9%; fucosylation index, ≥ 65.6%). All samples (n = 76 samples) from the healthy control group were categorized as type A irrespective of their serum AGP levels. Samples from the patients group were categorized from type A to type D, which seemed to depend on the clinical status of each patient after surgery. Thirty-four patients who had no clinical evidence of recurrence or metastasis at ≈ 5 months after surgery had type A, B, or C, irrespective of the extent of disease progression, including patients with advanced disease and patients with metastasis; however, without exception, all 11 patients with recurrent or metastatic disease after surgery who died had type D. Typical patterns from CAIE with Con A and AAL in each type are shown in Figure 5. Hence, maintaining type D may suggest lack of good prognosis after surgery.

Figure 5.

Classification of four types of α1-acid glycoprotein (AGP) glycoforms according to the fucosylation index (FI) and the branching index (BI) for various disease states. The FI and the BI were defined by the percentages of glycoforms that did not react to concanavalin A (Con A) and glycoforms that reacted strongly and weakly to Aleuria aurantia lectin (AAL), as determined by crossed affinoimmunoelectrophoresis (CAIE) with Con A and AAL, respectively. Four quadrants (A–D) were defined by the cut-off values 65.6% and 50.9%, which were calculated as the FI and the BI (mean + standard deviation) in healthy controls. Typical CAIE patterns using Con A and AAL as affinity molecules in the first-dimension gel were inset into each type.

Follow-Up Studies on AGP Glycoforms

Follow-up studies were conducted on 45 patients for a long period after surgery, and each type of glycoform was determined. They were classified into patients with a good prognosis (34 patients) and patients with a poor prognosis (11 patients), as described above. Figure 6 shows changes in the types of glycoforms during follow-up studies in eight patients with advanced malignancies who were selected from patients who had different prognoses. Four patients who had hepatic (Fig. 6Aa), gastric (Fig. 6Ab), lung (Fig. 6Ac), or rectal (Fig. 6Ad) malignancies in advanced stages (except for the patient in Fig. 6Aa) but did not have recurrences had type A, B, or C glycoforms at the end of the 5-month period (Fig. 6Aa–d). It is noteworthy that, after possessing type D for > 2 months, dramatic drops were seen in the relative branching and fucosylation indices in the patient with rectal carcinoma (Fig. 6Ad) 5 months after surgery. Four patients who had pancreatic (Fig. 6Ba), esophageal (Fig. 6Bb), pancreatic (Fig. 6Bc) or colon (Fig. 6Bd) malignancies, and who had recurrences after surgery and subsequently died had type D glycoforms for > 2 months, even though they had low indices during a short period after surgery. It was of particular interest that changes into type D were observed in advance of a clinical recurrence in these 11 patients (data not shown), who had serum AGP levels (1430.7 μg/mL ± 728.6 μg/mL) determined several months after surgery, did not always have levels greater than the cut-off value (807 μg/mL). During the analyses of AGP glycoforms in 45 patients over long periods by CAIE, all but 4 patients showed a typical change of glycoforms shortly after surgery together with a rapid decrease in the relative C0 value for a short period, indicating an increase in the Cs value due to the elevated synthesis of diantennary chains on AGP (Figs. 1, 6). The aforementioned 4 patients did not show an increase in the relative Cs value, even though they secreted high levels of AGP into serum.

Figure 6.

Changes in the fucosylation and branching indices in patients with a good prognosis (A) and patients with a poor prognosis (B). (A) Patients with Stage II hepatic carcinoma (a), Stage IIIA gastric carcinoma (b), Stage IIIA lung carcinoma (c), and Stage IIIB rectal carcinoma (d). (B) Patients with Stage IIA pancreatic carcinoma (a), Stage IIB esophageal carcinoma (b), Stage IV pancreatic carcinoma (c), and recurrent colon carcinoma (d). Solid and dotted lines indicate the cut-off values for the fucosylation index (65.6%) and the branching index (50.9%), respectively. Thick bars also indicate the period (days after surgery) during which patients had type D glycoforms.

Factors that Affect Patient Prognosis

Discrimination analyses between clinicopathologic background factors, including TNM status, tumor size, AGP level, and fucosylation and branching indices, were performed to identify factors associated with the length of patients' survival. Recurrence and a poor prognosis could not be predicted by AGP levels or clinicopathologic background factors, except for tumor status (P < 0.001). However, an individual patient's chance of survival could be predicted with the lowest misclassification rate (0.0666; P < 0.0001) using the fucosylation and branching indices. When the cut-off values for the classification of AGP glycoforms (Fig. 5) were determined with 2 values (fucosylation index, 69.9; branching index, 49.7), which were obtained in the regression analysis of the 2 indices from 45 patients with malignancies (risk rate, < 0.05), AGP glycoforms in all 11 patients who had suffered recurrences and died showed indices above these values at time points several months after surgery, but patients who had no recurrences had indices below these 2 values. These results indicate that the type of AGP glycoform may be a good prognostic marker for malignant disease, and particularly for predicting the fate of patients with advanced malignancies; however, the serum levels of AGP and the other clinical background factors (except tumor status) in the same patients were hardly useful for predicting prognosis.

Correlation between Fucosylated AGP and Serum α(1,3)-Fucosyltransferase Activity

Figure 7 shows serum α(1,3)-fucosyltransferase activity assayed in a large number of patients (n = 66 patients) in association with their levels of α(1,3)-fucosylation on the AGP. The fucosylation index showed a strong correlation with serum α(1,3)-fucosyltransferase activity (r = 0.772; P < 0.0001).

Figure 7.

Correlation between serum α(1,3)-fucosyltransferase activity and fucosylated α1-acid glycoprotein in patients with various types of carcinoma. As: glycoforms that reacted strongly to Aleuria aurantia lectin; Aw: glycoforms that reacted weakly to Aleuria aurantia lectin.

DISCUSSION

Alterations in glycoconjugates expressed on the cell surface have been observed widely in malignant tumor tissues37, 38 together with highly elevated expression of α-fucosylated antigens and aberrant distribution of such antigens. These antigens have been detected in plasma and serum originated and eluted from tumor tissues.37–43 With the aid of monospecific antibodies, glycoconjugates expressed on these tumor-related antigens have been recognized as clinically useful markers for diagnosis and prognosis in patients with malignancies; however, more sensitive and specific markers are needed for detecting clinical recurrences or metastasis during the treatment and follow-up of patients with malignancies after surgery. Many reports have indicated that the determination of changes in glycosylation as well as in serum levels of AGP may be useful for the diagnosis and management of various diseases, including malignant diseases.6, 9, 11, 44–46 Measurement of the amount of AGP in a large numbers of serum samples from patients with preoperative malignancies in this study showed that levels of AGP were increased significantly (P < 0.001) compared with the levels in the healthy control group, although no correlation was detected with clinical background factors, including clinical stage or tumor size. Follow-up studies of serum AGP levels during a long period after surgery revealed that AGP levels both in patients with a poor prognosis and in patients with a good prognosis changed indistinguishably, except that, in most patients, a significant elevation in the AGP level was observed during the period shortly after surgery.

Changes in AGP glycoforms in patients with malignancies during preoperative and postoperative periods could be analyzed easily in this study with the aid of CAIE using Con A and AAL as affinity components in the first-dimension gel. Accordingly, distribution of the AGP glycoforms was assayed in 45 patients with various types of malignancies during a long period after surgery. In most of our patients, postoperative changes in AGP glycoforms that occurred during the short period after surgery were associated not only with an increase in the % Cs but also with a decrease in the % C0, as determined by CAIE with Con A, in addition to the large increase in total AGP concentration. The changes in AGP glycoforms suggests that the increase in relative amounts of diantennary glycans, rather than triantennary or tetraantennary species, occurred along with the new synthesis of AGP as a consequence of acute-phase reaction. It was found that some patients with relatively high AGP levels had poor expression of such changes in glycosylation during the observed period. The acute phase-related changes diminished at ≈ 2 weeks after surgery, and it was assumed that subsequent variations in glycoform distribution reflected changes in the clinicopathologic status of the patients.

Indeed, the results showed that all 11 patients with a poor prognosis who had recurrent disease or metastasis had type D AGP glycoform patterns at 5 months after surgery. In contrast, 34 patients with a good prognosis had type A, B, or C AGP glycoform patterns during the same period. However, a similar correlation between AGP glycoform type and patients' prognosis was not seen when the AGP type was determined during the preoperative period or shortly after surgery. The significance of AGP glycoforms for predicting prognosis in patients with malignant disease was also found when the cut-off values for the fucosylation and branching indices were determined in a regression analysis of samples from the 45 patients in place of samples from normal controls. Greater than one-third of patients with type A or B AGP glycoforms had a good prognosis after surgery, even though they had advanced disease (Stages III and IV). Hence, our investigations convincingly demonstrated that the combination of branching and fucosylation indices was a very significant marker for prognosis compared with other clinical factors, including tumor size, TNM classification, and serum AGP levels. Therefore, changes in the glycosylation of AGP after surgery can indeed be used as a parameter for monitoring and predicting the fate of tumor-bearing hosts.

Until now, fucose residues on AGP have been found only in α1,3-linkages, which are believed to be synthesized by the hepatic α(1,3)-fucosyltransferase VI.47, 48 This was also concluded from our earlier studies, in which healthy individuals who had lethal homozygous mutations in the FUT6 gene did not express AAL-reactive fucosylated AGP and had markedly diminished levels of plasma α(1,3)-fucosyltransferase activity compared with the levels in control plasma.31, 32 Furthermore, a high correlation was obtained between the fucosylation of AGP and the plasma α(1,3)-fucosyltransferase activities in various patients as well as in individuals with heterozygous mutations in the FUT6 gene. In the current study, we also observed a high correlation between the fucosylation index and serum α(1,3)-fucosyltranferase activity. These results strongly suggest that the tumor-dependent increases in AGP fucosylation result from changes in the hepatic glycosylation process of AGP, because fucosylation cannot take place in plasma due to the absence of GDP-fucose as a sugar donor for α-fucosyltransferase. In addition, these studies support a substantial hepatic contribution to the tumor-associated increase in plasma α(1,3)-fucosyltransferase activity. Because the variations in branching of the glycans are tumor-related, it can be assumed that activities of hepatic glycosyltransferases, such as N-acetyl glucosaminyltransferase IV and/or V, α(1,4)galactosyltransferase, α(2,3)-sialyltransfersase, and α(2,6)-sialyltransferase, which determine the branching and terminal glycosylation, are affected by the tumor. Apparently, these effects may reflect a systemic reaction of the liver in response to the tumor like its response during both acute and chronic inflammation.47

The possibility cannot be excluded that the tumor itself also contributes to the increased plasma enzyme level, because such a contribution was found recently in patients with colorectal carcinoma.33 Along with α(1,3)-fucosyltransferase, tumor-associated elevations in the activities of α(1,2)-fucosyltransferase, α(1,4)-fucosyltransferase, and α(1,6)-fucosyltransferase have been reported.26–29, 42, 49 This emphasizes the specific role that the fucosylated antigens appear to play in carcinogenesis.37, 38, 40, 43

Recently, in patients with rheumatoid arthritis and patients with liver disease, strong correlations were reported between the clinical background and the amount of fucosylation in AGP, which was determined by an ELISA using anti-AGP antibody and AAL.34, 35 However, among our patients, 5 of 34 patients who had AGP with highly fucosylated but less branched glycans had a good prognosis. Furthermore, increased fucosylation that was also reported to occur after surgery in patients with benign diseases13 was detected in ≈ 50% of patients.

AGP glycoforms classified as type D, as discussed above, were associated with a poor prognosis in patients with carcinoma, whereas glycoforms classified as types A–C were associated with a good prognosis. According to their reactivities with Con A and AAL, the type D AGP glycoforms contain mainly five triantennary and/or tetraantennary glycans that are substantially fucosylated, whereas the majority of type B and C AGP glycoforms are fucosylated to various extents, and one or more of the five triantennary or tetraantennary glycans have been replaced by diantennary glycan(s). In these fucosylated glycans, part of the fucosyl residues are present in a sialyl LeX configuration (NeuAcα2,3Galβ1,4[Fucα1,3]GlcNAc) because of the generally high degrees of sialylation in AGP. Antiinflammatory properties have been described for such substituted AGP molecules.13, 17, 47

Essentially, such properties were found for AGP glycoforms that also contained one or more diantennary glycans, like type A.47 Therefore, it is tempting to speculate that the association of a good prognosis with these types of AGP glycoforms is more than a coincidence. It has been reported that repeated lactosamine units ([Galβ1,4GlcNAcβ]2) occur on some of the glycans of normal AGP.50 Therefore, it is possible that part of the fucose residues on highly fucosylated AGP may be present on these extensions of the glycans as dimeric or trimeric LeX determinants. Indeed, it was found that the ratio of the branching and fucosylation indices was low in some patients in the current study, indicating that > 2 α(1,3)-fucosylated residues may be synthesized on a single N-linked glycan of AGP. Determination of highly fucosylated glycans found in type D AGP glycoforms is in progress.

Various glycoforms are being isolated for future studies to enable elucidation of the fine structure of fucosylated sugar chains in patients with malignant disease. Even though the current CAIE technique is not simple or convenient for assaying large numbers of samples at once, it may be useful for identifying AGP glycoforms without providing an analysis of their detailed structures. Furthermore, assay reproducibility using standard samples indicates that this technique must be suitable for clinical samples, particularly for determining AGP glycoforms as novel tumor markers. In conclusion, the results of this study show that variations in the relative distribution of AGP glycoforms in serum, as determined by the CAIE technique using Con A and AAL, are appropriate markers of disease progression and prognosis in patients with malignancies.

Ancillary