Quantification and Utility of Monosialylated -Fetoprotein in the Diagnosis of Hepatocellular Carcinoma with Nondiagnostic Serum Total -Fetoprotein

Transcription

1 Clinical Chemistry 48: (2002) Enzymes and Protein Markers Quantification and Utility of Monosialylated -Fetoprotein in the Diagnosis of Hepatocellular Carcinoma with Nondiagnostic Serum Total -Fetoprotein Terence C.W. Poon, Tony S.K. Mok, Anthony T.C. Chan, Charles M.L. Chan, Veronica Leong, Steven H.T. Tsui, Thomas W.T. Leung, Herman T.M. Wong, Stephen K.W. Ho, and Philip J. Johnson * Background: At concentrations <500 g/l, serum fetoprotein (AFP) has low specificity in the diagnosis of hepatocellular carcinoma (HCC), but monosialylated AFP (msafp) is more specific for HCC. We describe two strategies for quantitative analysis of msafp and explore their diagnostic accuracy in cases of HCC with nondiagnostic serum total AFP concentrations. Methods: We first used isoelectric focusing, Western blot, and densitometry (IEF-Western blot assay). We then developed a second assay, a novel glycosylation immunosorbent assay (GISA), based on the specificity of sialyltransferase and immunosorbent technology. Both assays were used to measure msafp and msafp percentage relative to total AFP in sera with nondiagnostic AFP concentrations from 36 patients with newly diagnosed HCC and from 18 patients with liver cirrhosis. Results: The msafp percentages and concentrations were significantly higher in the HCC patient group regardless of the quantification methods. The msafp concentrations and msafp percentages obtained by the two assays were highly correlated (r 0.70 and 0.49, respectively). For discrimination of HCC with nondiagnostic serum total AFP from liver cirrhosis, the areas under the ROC curves were 0.81 (95% confidence interval, ) for msafp by IEF-Western blot assay, 0.73 ( ) for msafp by GISA, 0.89 ( ) for msafp percentage by IEF-Western blot assay, and 0.74 ( ) for msafp percentage by GISA. Conclusions: Both the serum concentration and percentage of msafp are potential diagnostic markers for HCC with nondiagnostic AFP. GISA can quantify a specific glycoform of a serologic marker American Association for Clinical Chemistry Measurement of serum -fetoprotein (AFP 1 ; reference values 10 g/l) provides a marker for the diagnosis and management of hepatocellular carcinoma (HCC) (1 3) and nonseminomatous germ cell tumors (4 7). Of patients with HCC, 80 90% will have concentrations above the upper reference limit (1 3). A serum concentration 500 g/l, in an area with high incidence of HCC and in the appropriate clinical setting, is usually considered diagnostic of HCC. However, modestly increased AFP ( g/l) is also common in nonmalignant chronic liver disease so that the specificity of the AFP test for HCC tends to be low (1, 8 10). This represents a serious clinical drawback for the test because most cases of HCC arise in patients with concurrent chronic liver disease (11, 12). Several attempts have been made to identify a HCCspecific glycoform, aiming to improve the specificity of AFP as a diagnostic test for HCC. The most successful approach has been based on the difference in the binding affinities of the AFP glycoforms to various lectins, partic- Department of Clinical Oncology, the Sir Y.K. Pao Centre for Cancer, The Chinese University of Hong Kong, Shatin, Hong Kong. *Author for correspondence. Fax ; pjjohnson@cuhk. edu.hk. Received November 28, 2001; accepted March 28, Nonstandard abbreviations: AFP, -fetoprotein; HCC, hepatocellular carcinoma; IEF, isoelectric focusing; dsafp and msafp, disialylated and monosialylated AFP, respectively; LC, liver cirrhosis; GISA, glycosylation immunosorbent assay; PBS, phosphate-buffered saline; and CI, confidence interval. 1021

2 1022 Poon et al.: msafp in Diagnosis of Hepatoma ularly Lens culinaris agglutinin and concanavalin A (13 16). L. culinaris agglutinin-reactive AFP, named AFP-L3, has been identified as a potential marker for the detection of small HCC (13, 17, 18). Over the past few years, using isoelectric focusing (IEF) to directly identify isoforms of AFP, we have shown that a specific AFP band (designated Band II) appears to be relatively specific for HCC (19 21). Occasionally another tumor-specific isoform, Band III AFP, which is common in cases of nonseminomatous germ cell tumors (22), is also present in HCC cases (20). Preliminary studies have suggested that screening for the Band II isoform may allow early, even preclinical, diagnosis of HCC in high-risk patients (21). Recently we successfully determined the carbohydrate moieties of serum AFP (23) and have subsequently shown that Band I and Band II AFPs are composed of disialylated and monosialylated AFP (dsafp and msafp), respectively (24). Our previous studies have, however, been essentially qualitative in nature. In this report, we describe our initial attempts to develop new strategies for quantitative analysis of msafp and investigate the potential of such assays to permit discrimination between HCC and liver cirrhosis (LC) when the serum total AFP is at a nondiagnostic concentration. For the first assay, we constructed an assay based on IEF, Western blotting, and densitometry (IEF- Western blot assay). For the second assay, we developed a novel glycosylation immunosorbent assay (GISA), based on the specificity of sialyltransferase and immunosorbent technology. Materials and Methods patients Serum samples with an AFP concentration between 50 and 500 g/l from 36 patients with newly diagnosed HCC and from 18 patients with LC alone, attending our Joint Hepatoma Clinic, were used in the present study. The patients had been diagnosed according to standard clinical criteria as reported previously (21). All HCC cases were histologically confirmed. All patients in the LC group had been followed for at least 18 months (from the date of blood collection) for any sign of HCC, to exclude individuals with asymptomatic HCC. preparation of calibrators Serum containing 5 mg of total AFP was applied to a mixture of 50 g/l sorbitol, 100 ml/l glycerol, and ampholytes (6.25% Pharmalyte ph ; Pharmacia Biotech) that had been prefocused at constant power (15 W) for 30 min in a Rotofor cell (Bio-Rad) at 4 C. The sample was focused at constant power (15 W) until the voltage stabilized and then was focused for an additional 30 min before twenty 2-mL fractions were harvested. The ph and total AFP concentration in each fraction were measured. Fractions with 10 mg of total AFP were pooled and refractionated to improve separation of the pi variants. The ph and AFP concentration were again measured. Fractions containing only dsafp or msafp, free of other pi variants, were identified by the banding pattern in gel IEF. Finally, the separated dsafp or msafp in the pooled fractions was purified by affinity chromatography. ief-western blot assay The IEF-Western blot assay for quantitative analysis of msafp is a modification of the previously reported method for detection of Band II AFP (20 22). AFP isoforms in the serum samples or calibrators were separated by IEF on a polyacrylamide gel that had been preswollen with a solution containing 5 mol/l urea and 1:16 Pharmalyte (Pharmacia Biotech) in doubly distilled water. Calibrators were prepared by mixing the purified dsafp and msafp in appropriate proportions. One microliter of prediluted serum sample or calibrator containing 5 or 10 g/l total AFP was applied to the anode side of the gel after prefocusing (2000 V; 2.0 ma; 3.5 W; 10 C; 75 V-h). The sample was applied for 15 V-h (200 V; 2.0 ma; 3.5 W; 10 C), and the final isoelectric separation step was done for 450 V-h (2000 V; 5.0 ma; 3.5 W; 10 C). The focused proteins were transferred to a nitrocellulose membrane and then incubated with polyclonal rabbit anti-human AFP (Dako), followed by horseradish peroxidase-conjugated polyclonal swine anti-rabbit immunoglobulin (Dako). After the membrane was washed, the enhanced chemiluminescence detection system (ECL; Pharmacia Biotech) was used to visualize the AFP protein bands. The image of each band was scanned with a densitometer (GS-700; Bio-Rad), and the intensity of each band was expressed as a percentage of the total intensity of all AFP bands. The relative Band II intensity (Band II AFP%, as a percentage of total band intensity) was plotted against the percentage of msafp relative to the total AFP (the msafp%) to generate a calibration curve for determining the msafp% in each serum sample. The serum concentration of msafp was calculated using the formula: msafp concentration serum total AFP concentration msafp%. All patient samples were measured at least in quadruplicate. Samples were measured in a single assay with 11 replicates for determination of the intraassay CV and in quadruplicate in 7 separate assays for determination of the interassay CV. gisa GISA is a novel strategy that we developed to quantify a specific glycoform by use of the specific labeling activity of glycosyltransferase to recognize the specific terminal residue of the carbohydrate side chain. The predominant msafp molecules have a carbohydrate side chain that terminates with galactose, whereas dsafp molecules, which are fully sialylated, do not. In the presence of -galactoside- 2,3/6-sialyltransferase and CMP-5-fluoresceinyl-neuraminic acid, msafp molecules, but not dsafp molecules, will be specifically labeled at the terminal galactose residue. The labeled msafp was quantified

3 Clinical Chemistry 48, No. 7, by the conventional ELISA. The major steps are shown in Fig. 1. In the assay, patient serum samples were first diluted at least threefold with phosphate-buffered saline (PBS), ph 7.4, and supplemented with 100 g/l milk powder. Duplicate 30- L aliquots of the calibrators and the diluted samples were added to the wells of a 96-well microplate, precoated with 50 L/well of polyclonal rabbit antihuman AFP (1:200 dilution; Dako), and incubated at 4 C overnight to capture the AFP molecules. The next day the microplate was washed four times with PBS containing 5 ml/l Tween 20 (Amresco), and once with labeling buffer containing 45 mmol/l sodium chloride, 1 g/l bovine serum albumin, 1 ml/l Triton X-100, and 15 mmol/l Tris-HCl, ph 7.2. To each well we added 45 L of labeling buffer containing 2.5 U/L purified rat liver sialyltransferase (Sigma) and 0.75 mg/l CMP-5-fluoresceinylneuraminic acid; we then incubated the plate for 2 h at 37 C to label the msafp molecules. The microplate was then washed three times with washing buffer (PBS, ph 7.4, containing 5 ml/l Tween 20) and twice with alkaline washing buffer (5 ml/l Tween 20, 2.5 g/l bovine serum albumin, 0.1 mol/l sodium carbonate, ph 9.0) to further reduce the background noise. Fig. 2. Quantitative analysis of msafp by IEF-Western blot with densitometry, using calibrators with known msafp percentages. Calibrators containing different percentages of msafp were separated by IEF and visualized by Western immunoblotting (A). A calibration curve (y x x 3.6; R ) was then generated by plotting the relative Band II intensity against the percentages of msafp in the calibrators (B). Similar calibration curves were obtained when the calibrators were prepared at a final concentration of 5 or 10 g/l. The labeled msafp molecules were released from each well by the addition of 60 L of releasing buffer (PBS containing 2.5 g/l bovine serum albumin, 0.5 ml/l Tween 20, 0.1 mol/l NaOH, and 5 mmol/l sodium phosphate), transferred to a microcentrifuge tube, and neutralized with a predetermined amount of 0.1 mol/l HCl solution. The neutralized samples were added to a new microplate precoated with 100 L of a 1:500 dilution of polyclonal rabbit anti-human AFP per well and incubated at 37 C for 90 min. After the plates were washed four times with washing buffer, 100 L of a 1:500 dilution of horseradish peroxidase-conjugated polyclonal sheep anti-fluorescein (Roche Diagnostics) in PBS (ph 7.4) con- Fig. 1. Major steps in the GISA for quantitative measurement of msafp. (1), after captured to a microplate precoated with anti-afp, the msafp molecules, which carry a carbohydrate side chain with a terminal galactose, but not dsafp molecules are specifically labeled with fluoresceinyl-neuraminic acid (CMP-Sialic acid-fitc) under the catalysis of sialyltransferase. (2), the final reaction mixtures are then transferred to another microplate precoated with anti-afp. (3), the captured msafp molecules, which carry the FITC labels, are recognized by an anti-fitc-peroxidase conjugate. The msafp concentration is directly proportional to the peroxidase (POD) activity in each well. Fig. 3. Typical calibration curve of the GISA for quantitative analysis of msafp. Equation for the curve: y x x 0.33 (R ).

4 1024 Poon et al.: msafp in Diagnosis of Hepatoma Table 1. Total AFP concentrations and msafp concentrations and percentages measured by IEF-Western blot analysis and GISA in HCC patients and LC patients with nondiagnostic serum total AFP. Study groups Total AFP, g/l msafp, g/l msafp% msafp, g/l msafp% HCC patients (n 36) Median (25th 75th percentile) 140 ( ) 33 (21 53) a 22 (18 30) a 15 (5 32) 9 (4 15) Mean LC patients (n 18) Median (25th 75th percentile) 146 (81 245) 17 (10 25) 11 (8 14) 4 (3 9) 4 (2 6) Mean HCC vs LC NS b P P P P a P 0.001, compared with the readings obtained by the GISA. b NS, not significant. IEF-Western blot analysis GISA taining 100 g/l milk powder, 2.5 g/l bovine serum albumin, and 0.5 ml/l Tween 20 was added to each well and incubated at 37 C for 1 h. The microplate was then washed five times with washing buffer. Finally, 100 L of BM blue, a peroxidase substrate (Roche Diagnostics), was added to each well to develop the color. After the reaction was stopped with 150 L of 2 mol/l H 2 SO 4, the absorbance of each well was measured at 450 nm (against a reference wavelength of 690 nm). The serum msafp% was calculated using the formula: msafp% msafp concentration serum total AFP concentration 100%. Quality-control samples were measured in a single assay with nine replicates for the intraassay CV and measured in duplicate in seven separate assays for the interassay CV. statistical analysis The Mann Whitney U-test was used to compare differences between the study groups. The 2 test with Yates correction was used to compare frequency data between the groups. Correlations between the markers were analyzed by the Spearman rank-order correlation test. Statistical software Analyze-it TM ( was used to construct ROC curves and to calculate their areas and confidence intervals (CIs). Results quantification of msafp by ief-western blot assay As shown in Fig. 2, the relative Band II intensity was directly proportional to the msafp% in the calibrators. Using the calibration curve, we were able to quantify the msafp% in the serum samples. The intraassay CV (n 11) was 10% at a msafp percentage value of 38%, whereas the interassay CV (n 7) was 16% and 6% at msafp percentage values of 12% and 24%, respectively. quantification of msafp by gisa As shown in Fig. 3, the absorbance was directly proportional to msafp concentration in the calibrators. When 800 g/l dsafp and 40 g/l msafp were added to an AFP-negative serum, the measured concentration of the mixture was 49 g/l. The result indicated that the GISA was capable in differentiating msafp from dsafp and giving an accurate measurement of msafp. The intraassay CV (n 9) was 11% at a concentration of 50 g/l, and the interassay CV (n 7) was 11% at a concentration of 20 g/l. serum concentrations and percentages of msafp in hcc patients and lc patients with nondiagnostic afp As shown in Table 1 and Fig. 4, we were able to determine the serum concentrations and percentages of msafp in the HCC patients and LC patients with nondiagnostic AFP by both the IEF-Western blot assay and the GISA. As expected, serum total AFP was not significantly different between the two groups of patients. However, the msafp percentages and concentrations were significantly higher in the HCC patients regardless of the quantification methods. Using ROC curve analysis (Fig. 5), we found that the serum total AFP was not useful in discriminating HCC from LC [area under the ROC curve, 0.54 (95% CI, ); P 0.31], but both the msafp concentration and its percentage were useful in discrimination, regardless of the assay methods. The areas under the ROC curves of the msafp concentration and its percentage, measured by the IEF-Western blot assay, were 0.81 (95% CI, ; P ) and 0.89 (95% CI, ; P ), respectively. The areas under the ROC curves of the msafp concentration and the msafp percentage, measured by the GISA, were 0.73 (95% CI, ; P 0.001) and 0.74 (95% CI, ; P 0.001), respectively. In summary, both the msafp concentration and percentage of msafp, which were measured by either the IEF-Western blot analysis or the GISA, were useful in discriminating between HCC and LC, but the total AFP concentration was not. comparison and correlation between msafp values obtained by the two assays To examine whether the two assays measure the same target molecules in the serum, the data obtained from all

5 Clinical Chemistry 48, No. 7, respectively, as measured by the IEF-Western blot assay, and 61% and 50% as measured by the GISA. sensitivities of msafp in groups of patients with different tumor sizes The effect of tumor size on the sensitivities of msafp concentration and percentage was analyzed by grouping the patients with measurable HCC into three subgroups Fig. 4. msafp concentrations (A) and percentages (B) measured by IEF-Western blot assay plotted against the values measured by the GISA., HCC patients; E, LC patients. 54 cases were subjected to Wilcoxon signed-rank test analysis and Spearman rank-order correlation analysis. When the msafp values obtained by the two assays were compared, both the msafp concentration (P 0.001) and percentage (P 0.001) were significantly higher when measured by the IEF-Western blot assay (Table 1), whereas the msafp concentrations (r 0.70; P ) and msafp percentages (r 0.49; P ) obtained by the two assays were highly correlated. At cutoff values to achieve a specificity of 80% (i.e., only 3 of 18 LC cases were wrongly classified as HCC), sensitivities for HCC (36 cases) were 61% and 81% for msafp and msafp%, Fig. 5. ROC analysis of the ability of the msafp concentration and the msafp percentage, as measured by IEF-Western blot assay (A) and GISA (B), to discriminate between HCC and LC patients with nondiagnostic serum total AFP.

6 1026 Poon et al.: msafp in Diagnosis of Hepatoma according to tumor size (Table 2). The tumor size did not significantly affect the sensitivities of msafp concentration/percentage, as measured by the IEF-Western blot assay ( 2 test with Yates correction). It is worth noting that three of five HCC patients with a tumor size 2.5 cm were positive for either msafp concentration or msafp%, as measured by the IEF-Western blot assay. Discussion In the present study, we have developed two methods for the quantitative analysis of msafp. The results suggest that both the percentage and the absolute concentration of msafp are potential markers for the diagnosis of HCC and are especially useful in discriminating HCC from LC when the total AFP concentrations are within the diagnostic gray region. Differentiation between HCC and LC is particularly important in Southeast Asian countries, including Hong Kong and China, because 80% of HCC cases arise in patients with preexisting LC (11, 12). Studies of the clinical value of AFP-L3 (L. culinaris agglutinin-reactive AFP glycoform) have shown that it is a potential marker for the detection of small tumors (17, 18, 25), tumor recurrence (25), distant metastasis, and poor prognosis for HCC (26, 27). In a case of pancreatoblastoma, it was observed that both the asialylated and monosialylated AFP glycoforms disappeared during chemotherapy but recurred when the chemotherapy was withdrawn (28). Future studies could explore the clinical value of msafp. We are currently investigating the prognostic value of msafp in HCC. The present study was of the case-control type, in which there may be a positive bias; additional studies are therefore needed to formally elucidate the diagnostic accuracy. It appears that both the IEF-Western blot assay and GISA are capable of quantifying msafp in terms of either concentration or percentage, and from one value we can calculate the other if we know the total AFP concentration. However, these two assays appear to behave differently. Although the msafp values obtained by the two assays were highly correlated, the correlation coefficients were far less than 1.0. Furthermore, the msafp concentrations and percentages obtained by the GISA were significantly lower than those obtained by the IEF-Western blot assay. This suggests that the GISA may measure only a particular fraction of the msafp molecules that can be measured by the IEF-Western blot assay. Such a suggestion is consistent with our previous observation that the tumor-specific AFP isoforms focused as Band II by the IEF analysis were composed of a mixture of msafp isoforms that were microheterogeneous at their carbohydrate side chains (24). Successful detection of an msafp molecule by GISA relies on the presence of a terminal galactose residue to which the labeled sialic acid molecule can bind. Although those msafp isoforms missing the terminal galactose residue cannot be detected by the GISA, they can still be detected by the IEF-Western blot assay. This may also explain why the IEF-Western blot assay is more sensitive than the GISA in differentiating HCC from LC. However, despite being less sensitive, our results suggest that the GISA may be useful in identifying HCC cases with nondiagnostic AFP concentrations. Furthermore, because it has a format similar to ELISA, the GISA is a highthroughput assay and has the potential to be fully automated. The assay steps could be reduced by direct detection of the label attached to msafp after the labeling step. Because the antibody used in the current GISA also carries carbohydrate side chains containing terminal galactose residues, direct detection would lead to an unacceptably high background. However, pretreating the antibodies with unlabeled CMP-5-neuraminic acid and sialyltransferase may solve this problem. An alternative strategy would be the use of carbohydrate-free Fab regions rather than intact antibody. There is also scope for us to improve the clinical value of the current GISA. One possibility is to develop a similar assay that can quantify nongalactosylated AFP isoforms, by labeling the terminal glucose residue, or other tumorspecific isoforms by similar strategies. This combination of multiple labeling systems may enable the GISA to detect the majority of tumor-specific AFP isoforms and have a sensitivity comparable to that of the IEF-Western blot assay. Theoretically AFP-L3, which carries an extra fucose residue, may be detected by use of a specific glycosyltransferase to add a labeled monosaccharide to the fucose residue. Furthermore, the GISA technology could be also applied to quantify other serologic glycoproteins with disease-specific glycosylation profiles, e.g., 1 -acid glycoprotein (29) and human chorionic gonadotropin (30). The GISA technology avoids the use of lectins, which are less specific and more difficult to integrate into a simple assay system (31). Table 2. Sensitivities of the msafp concentration and percentage, measured by the IEF-Western blot assay and by the GISA, in relation to tumor size. Tumor size, a cm IEF-Western blot analysis msafp >26.0 g/l msafp% >15% msafp >9.4 g/l msafp% >8.9% % (3/5) 60% (3/5) 40% (2/5) 20% (1/5) % (5/7) 71% (5/7) 71% (5/7) 43% (5/7) % (11/18) 89% (16/18) 67% (11/18) 61% (11/18) a Tumor size did not significantly affect the sensitivities of the assays ( 2 test with Yates correction). GISA

7 Clinical Chemistry 48, No. 7, In conclusion, we have successfully developed two strategies for the quantitative analysis of msafp. One is based on a novel technology, named GISA. Although future studies are needed for less biased elucidation of the diagnostic accuracy, the present study indicates that both the serum concentration and percentage of msafp are potential diagnostic markers for HCC with nondiagnostic AFP concentrations. Quantitative analysis of msafp could provide an objective and practical way to improve the diagnostic value of total AFP, the conventional tumor marker for HCC. We are grateful to the Providence Foundation Ltd. (Hong Kong), the Hong Kong Research Grants Council, and the Hong Kong Cancer Fund for continuing support of liver cancer research. References 1. Johnson PJ, Portmann B, Williams R. -Fetoprotein concentration measured by radioimmunoassay in the diagnosing and excluding of hepatocellular carcinoma. Br Med J 1978;2: Sheu JC, Sung JL, Chen DS, Yang PM, Lai MY, Lee CS, et al. Growth rate of asymptomatic hepatocellular carcinoma and its clinical implications. Gastroenterology 1985;89: Nomura F, Ohnishi K, Tanabe Y. Clinical features and prognosis of hepatocellular carcinoma with reference to serum -fetoprotein levels. Cancer 1989;64: Lange PH, Fralay EE. Serum AFP and HCG in the treatment of patients with testicular tumors. Urol Clin North Am 1977;4: Javadpour N. The role of biologic tumor markers in testicular cancer. Cancer 1980;45: Kay PH, Wells FC, Goldstraw P. A multi-disciplinary approach to primary nonseminomatous germ cell tumours of the mediastinum. Ann Thorac Surg 1987;44: Nichols CR, Saxman S, Williams SD, Loehrer PJ, Miller ME, Wright C, et al. Primary mediastinal nonseminomatous germ cell tumors. A modern single institution experience. Cancer 1990;65: Lok ASF, Lai CL. -Fetoprotein monitoring in Chinese patients with chronic hepatitis B virus infection: role in the early detection of hepatocellular carcinoma. Hepatology 1989;9: Okuda K. Early recognition of hepatocellular carcinoma. Hepatology 1986;6: Trevisani F, D Intino PE, Morselli-Labate AM, Mazzella G, Accogli E, Caraceni P, et al. Serum -fetoprotein for diagnosis of hepatocellular carcinoma in patients with chronic liver disease: influence of HBsAg and anti-hcv status. J Hepatol 2001;34: Kew MC, Popper H. Relationship between hepatocellular carcinoma and cirrhosis. Semin Liver Dis 1984;4: Johnson PJ, Williams R. Cirrhosis and the aetiology of hepatocellular carcinoma. J Hepatol 1987;4: Mizejewski GJ. -Fetoprotein structure and function: relevant to isoforms, epitopes, and conformational variants. Exp Biol Med 2001;226: Smith CJP, Kelleher PC. -Fetoprotein molecular heterogeneity: physiologic correlations with normal growth, carcinogenesis and tumour growth. Biochim Biophys Acta 1980;605: Buamah PK, Harris R, James OFW, Skillen AW. Lentil-lectinreactive -fetoprotein in the differential diagnosis of benign and malignant liver disease. Clin Chem 1986;32: Du MQ, Hutchinson WL, Johnson PJ, Williams R. Differential -fetoprotein lectin binding in hepatocellular carcinoma. Cancer 1991;67: Li D, Mallory T, Satomura S. AFP-L3: a new generation of tumor marker for hepatocellular carcinoma. Clin Chim Acta 2001;313: Kumada T, Nakano S, Takeda I, Kiriyama S, Sone Y, Hayashi K, et al. Clinical utility of Lens culinaris agglutinin-reactive -fetoprotein in small hepatocellular carcinoma: special reference to imaging diagnosis. J Hepatol 1999;30: Burditt LJ, Johnson MM, Johnson PJ, Williams R. Detection of hepatocellular carcinoma-specific -fetoprotein by isoelectric focusing. Cancer 1994;74: Ho S, Cheng P, Chan A, Leung N, Yeo W, Leung T, et al. Isoelectric focusing of -fetoprotein in patients with hepatocellular carcinoma frequency of specific banding patterns at non-diagnostic serum levels. Br J Cancer 1996;73: Johnson PJ, Leung N, Cheng P, Welby C, Leung T, Lau WY, et al. Hepatoma-specific alphafetoprotein may permit preclinical diagnosis of malignant change in patients with chronic liver disease. Br J Cancer 1997;75: Johnson PJ, Ho S, Cheng P, Chan A, Leung T, Yuen J. Germ cell tumors express a specific -fetoprotein variant detectable by isoelectric focusing. Cancer 1995;75: Johnson PJ, Poon TCW, Hjelm NM, Ho CS, Ho SKW, Welby C, et al. Glycan composition of serum -fetoprotein in patients with hepatocellular carcinoma and non-seminomatous germ cell tumour. Br J Cancer 1999;81: Johnson PJ, Poon TCW, Hjelm NM, Ho CS, Blake C, Ho SKW. Structures of disease-specific serum -fetoprotein. Br J Cancer 2000;83: Taketa K, Endo Y, Sekiya C, Tanikawa K, Koji T, Satomura S, et al. A collaborative study for the elevation of lectin-reactive -fetoproteins in early detection of hepatocellular carcinoma. Cancer Res 1993;53: Hayashi K, Kumada T, Nakano S, Takeda I, Sugiyama K, Kiriyama S, et al. Usefulness of measurement of Lens culinaris agglutininreactive fraction of -fetoprotein as a marker of prognosis and recurrence of small hepatocellular carcinoma. Am J Gastroenterol 1999;94: Yamashiki N, Seki T, Wakabayashi M, Nakagawa T, Imamura M, Tamai T, et al. Usefulness of Lens culinaris agglutinin-reactive fraction of -fetoprotein (AFP-L3) as a marker of distant metastasis from hepatocellular carcinoma. Oncol Rep 1999;6: Chan MHM, Shing MMK, Poon TCW, Johnson PJ, Lam CWK. -Fetoprotein variants in a case of pancreatoblastoma. Ann Clin Biochem 2000;37: Ryden I, Lundblad A, Pahlsson P. Lectin ELISA for analysis of 1 -acid glycoprotein fucosylation in the acute phase response. Clin Chem 1999;45: Cole LA, Shahabi S, Oz UA, Bahado-Singh RO, Mahoney MJ. Hyperglycosylated human chorionic gonadotropin (invasive trophoblast antigen) immunoassay: a new basis for gestational down syndrome screening. Clin Chem 1999;45: Turner GA. N-Glycosylation of serum proteins in disease and its investigation using lectins. Clin Chim Acta 1992;208: