Decreased hepatic function in patients with hepatoma or liver metastasis monitored by a hepatocyte specific galactosylated radioligand

9'Tc-galactosylated neoglycoalbumin ("mTc-NGA) is a hepatocyte-specific tracer that, after injection into the blood stream, delivers radioactivity selectively to the liver. This is based upon chemical recognition and binding by the hepatic binding protein (HBP), a receptor specific for galactosylated glycoproteins. Liver tissue samples were obtained intraoperatively from patients undergoing surgery for various cancers. The concentration of specific HBP receptors in the liver (normal liver, hepatoma, liver metastasis) was calculated from the in vitro binding of 99mTc-NGA. One week after surgery, the in vivo HBP density was also measured in some of these patients after injection of 3.5 mg (50 nmol per patient) "mTc-NGA (150-200 MBq) for simulation of 9'Tc-NGA kinetics. Comparison of in vitro and in vivo HBP concentration in the liver showed values in the same concentration range. In patients with hepatoma or liver metastasis a significantly (P<0.01) decreased global HBP density was found in vivo compared to controls. The values obtained for in vivo HBP concentration in the liver amounted to 0.38 ± 0.05 gmol 1-' liver for patients with hepatoma, to 0.4 ± 0.1 gAmol 1in patients with liver metastasis and to 94 ± 0.05 timol 1liver in cancer patients without liver malignancy. In vitro investigation of HBP density revealed the malignant liver tissue to have a significantly (P<0.0001) decreased or almost (completely) absent HBP receptor density compared to the normal tissue apart from the cancer area. It is concluded that determination of HBP density in vivo via a specific tracer is a new, simple and reliable approach for the determination of remaining hepatic function in patients with primary or secondary liver cancer.

Methods for the determination of functional liver mass in patients with liver disease are still being improved. In contrast to creatinine clearance as a parameter of kidney function, the measurement of a single liver function parameter does not reflect overall hepatic capacity due to the multitude of metabolic tasks of the liver including synthesis, uptake, degradation and secretion of bile. Although a number of quantitative tests of liver function, i.e. elimination of bromosulphophtalein (Haecki et al., 1976), antipyrine (Andreasen et al., 1974) or aminopyrine , exist, they are usually time consuming, difficult to apply and may have adverse effects and are therefore not extensively used.
Recently, Stadalnik and co-workers (Stadalnik et al., 1985;Vera et al., 1985a) introduced a model for in vivo binding and simulation of a hepatocyte specific tracer, 99mTc-neoglycoalbumin C9mTc-NGA), to human hepatic binding protein (HBP; Stockert & Morell, 1983) in patients with liver disease in order to evaluate hepatic function from global HBP receptor density and hepatic blood flow. In these studies NGA was seen to be hepatocyte-specific and its rate of accumulation was dependent on the amount of ligand injected and its affinity to the receptor (Vera et al., 1985b).
Direct evidence for reduction of HBP concentration as a consequence of hepatocellular pathology was obtained in studies with chemical carcinogens (Stockert & Becker, 1980), and positive correlation of reduced in vitro HBP binding activity and increased circulating inhibitors resulted from a study of galactosamine-induced liver disease (Sawamura et al., 1981). Based on these observations we addressed the question of whether the in vivo HBP density measured by 9'Tc-NGA scintigraphy would be changed in patients with primary or secondary liver cancer. This study investigated the in vivo binding of 99mTc-NGA to HBP in patients with normal livers, hepatomas and liver metastasis. Furthermore, in order to validate the method described, the in vivo HBP concentration was compared to the concentration of HBP measured in vitro at hepatic membranes. This was assessed in liver tissue samples (normal liver, hepatoma, liver metastasis) obtained intraoperatively from the same patient one week before 99mTc-NGA scintigraphy was performed.
A solution of 0.1 mol 1' A and 0.01 mol I -' CH3ONa in absolute methanol was kept at room temperature for 48 hours and then stored as stock solution at -15C (up to 3 months). It contained on average 0.055 mol I' 2-imino-2methoxyethyl-l-thio-p-D-galactopyranoside (B, coupling reagent). A measured aliquot of this stock solution (125 pl; 0.055 mol 1-) was evaporated to dryness, redissolved in fresh 0.2 mol 1`borate buffer, pH 8.6, a precise amount of human serum albumin (HSA; 17 pLI, 20% HSA = 3.4 mg = 50 nmol; Immuno AG, Vienna, Austria) was added and incubated over night at room temperature to produce the NGA-ligand. This was routinely isolated by repetitive ultrafiltration through a membrane with 20 kDa exclusion limit separating unbound coupling agent into the filtrate. The number of galactose residues per HSA-molecule was synthetically controlled by the molar ratio of coupling agent/HSA. According to a relation set up by Vera et al. (1984), a molar ratio of coupling agent/HSA = 138 was employed, resulting in about 21 galactose residues per HSA molecule. For each patient, 3.5 mg NGA (50 nmol 3 ml-' per patient) was labelled with 9'Tc in 0.15 mol I' NaCl at pH 2.5 by adding the desired activity of 99"TcO4 (patient dose 150-200 MBq) and reducing it with 32 jig Sn2+ generated in situ from a tin anode and platinum cathode, by applying a d.c. current of 5 mA for 11.4 s in 1 ml labelling volume. After stirring for 30 min, the product was neutralised and finally filtered through a sterile 0.2 jim membrane. Radiochemical purity was routinely monitored by cellulose-acetate electrophoresis in 0.1 mol 1' barbitone buffer, pH 8.6, run at 300 V for 20min. This system offered the advantage of determining both free TcO4-and reduced hydrolysed Tc (TcO2 x H20) in single analysis. Radiochemical purity was typically >97%, i.e. the 99mTc-NGA peak contained >97% of total "mTc on the electrophoresis strip. The labelling yield after filtration through low protein adsorption membranes amounted to about 95%, in vitro stability at room temperature extended through more than 10 hours. Liver membrane preparation Normal hepatic tissue samples were obtained intraoperatively from a total of 25 patients aged 44-81 years undergoing surgery for various cancers of the abdominal tract. In some patients with liver metastasis (n = 8) or hepatoma (n = 5) samples from the malignant area were obtained as well. Histological diagnosis was assessed by haematoxylin and eosin staining. The tissue removed and designated for the receptor study was transported immediately to the laboratory at 4°C.
After calculation of the liver volume (approximately 1 ml), the tissue was cut into pieces which were suspended in 15-20 ml 50 mmol I`Tris-HCl buffer, pH 7.5, and homogenised by means of ultraturrax (Typ 18/10, IKA-Labortechnik, Staufen, FRG) and ultrasound (Heat Systems Ultrasonic, sonicator W 220F, New York, USA). In order to study in vitro binding and to calculate the HBP density for the whole liver this homogenate was directly used for the binding assays. The homogenate was taken up in assay buffer containing 50 mmol -' Tris-HCI, 5 mmol [' MgCI2, 5 mmol 1' CaCI2, 1 mol 1' NaCI, pH 7.5, 4°C, and measured for its protein content by the assay kit provided by BIO-RAD (Commassie Blue G-250, Richmond, CA, USA).
The protein content in g I`liver was then calculated and amounted to approximately 100 g -'.
In separate experiments hepatic plasma membranes were isolated by the method described previously (Neville, 1968;Virgolini et al., 1989a). The membranes floating on the top of the 42.1% sucrose were removed and taken up in the same assay buffer (pH 7.5, 4°C) at a protein concentration of about 200jig 100 ILl-.

Binding assays
In preliminary studies we used the assay conditions recently applied also by Vera et al. (1 985b) in order to evaluate 99mTc-NGA binding to rabbit HBP. Due to the small and varying size of liver tissue samples available we finally reduced the total assay volume from 500 to 200 Il (in triplicate). Receptor-ligand interaction was studied previously (Virgolini et al., 1989a). For calculation of the number of HBP-receptors in normal liver tissue, hepatoma and liver metastasis, saturation experiments were carried out. Therefore increasing concentrations (0.01-200 nmol 1') of 99mTc-NGA were incubated with the protein suspension (200-500 jig protein 100 ;1I-') in the absence (determination of the total binding) and presence (determination of the nonspecific binding) of 100 jimol 1' NGA. The difference of total and non-specific binding is referred to as the specific binding. In competition experiments increasing concentrations of unlabelled NGA (0.01-1000 jLmol 1-') were tested to displace S nmol 1`of 99mTc-NGA.
To ensure equilibrium (Virgolini et al., 1989a,b), the incubation time was fixed for exactly 60 min for each sample. The incubation was performed at room temperature (22°C). Since the non-specific binding amounted to 5% only (blank limit), in later experiments only the total binding was assayed.

Gamma-camera imaging
Kinetic study In all patients the in vivo binding of 'Tc-NGA to HBP was estimated. The patients were placed in supine position under a gamma camera (Searle Radiographics Inc., Netherlands) connected to a data processor (PDP 11/34, Digital Equipment Int. Ltd, Galway, Ireland). The gamma-camera was equipped with a low energy collimator (140 Kev; Searle). Computer acquisition of gamma camera data was performed at a rate of two frames per minute and a matrix of 64 x 64. Time-activity curves were recorded over precordium and the liver. The total acquisition time was 30 minutes.
The exact dose of 9'Tc-NGA given to a patient was calculated from the dose in the syringe before injection and immediately thereafter and amounted to 4-5 mCi per 3.5 mg NGA (50 nmol). The exact volume was calculated from the syringe weight before injection and thereafter. Two minutes after injection of 99Tc-NGA blood was drawn and transferred into a preweighed plastic tube. The blood concentration of 99'Tc-NGA was calculated using the activity/gram of this blood sample and a diluted standard of the labelled product (1:5,000). Morphological study Liver morphology was studied by SPECT scintigraphy performed right after dynamic acquisition with a double-headed gamma-camera equipped with a low energy collimator (ROTA-camera, Siemens, FRG). Using a matrix of 128 x 128, 60 pictures were obtained within a total exposure time of 5 minutes (angle 300, one turn lOs).

Analysis
In vitro experiments All data were corrected for the half-life of "mTc. Calculation of the binding data in terms of Scatchard analysis was performed by a computer program (kindly provided by K. Neumann, Ing., Bender & Co., Vienna, Austria) which searched systematically for the highest level of correlation unto the model of two straight lines in the given interval testing against the alternative of a single straight line approximation (Neumann, 1988).
This program is based on classical least squares methodology for the lines fit. The program uses a straightforward partitioning of regression sums of squares followed by a standard regression F test.
The corresponding test has been shown within a Monte Carlo simulation to be rather reliable and on the conservative side. The purpose of the Monte Carlo simulation was to show that the implicit multiple decision problem does not seriously affect the significance levels.
Values are presented as the mean ± standard deviation. In vivo experiments The in vivo HBP concentration and hepatic blood flow (Q) were calculated from the timeactivity curves. The kinetic model was developed (and later published) by D.R. Vera and co-workers (Dept of Nuclear Medicine, University of California, Sacramento, USA; Vera et al., 1986). It consists of the haemodynamic subsystem which delivers the ligand to the target organ, and of the receptor-binding subsystem in which the formation of the receptor-ligand complex within the target organ take place. Following this model, system state equations can be obtained of the kinetic system which are mathematically represented as a system of first order non-linear differential equations.
The program runs on a MicrovaxII computer and produces both the graphical representation of the experimental and the fitted curves and additional numerical output of the system parameters, the most important of which are the concentration of HBP in the liver and the forward binding rate constant Kb for the reaction of the ligand with the receptor in the liver. Furthermore, the program gives estimates on the goodness of fit and of the errors for the various parameters.
At present we are using two observations: (a) the time course of radioactivity in the extrahepatic blood which can be obtained by a region of interest over the precordial area; (b) the time course of radioactivity in the area of the liver, which is the sum of two components, the radioactivity of the free ligand and the radioactivity of the ligand-receptor complex. In vitro binding studies In vitro binding experiments with normal liver parenchyma revealed high specific binding of 99mTc-NGA to HBP amounting to 91 ± 7% in the presence of 5 nmol 1-' of 99'Tc-NGA (Figure 1). The corresponding IC50 (i.e. concentration causing half maximal inhibition) value was 10-7 mol 1-'. However, in tissue samples obtained from a malignant area no relevant in vitro binding activity was observed (<40%) in the high affinity ligand range. In normal liver tissue the NGA binding capacity (Bmax) amounted to 6.8 ± 0.9 pmol mg' total liver protein, being equivalent to 1.13 ± 0.05 tLmol 1' liver (Table I). In tissue samples derived from hepatomas (Table II) or liver metastasis (Table III) the NGA binding capacity was significantly (P<0.0001) lower and amounted to 0.3 ± 0.05 pmol mg-' total liver protein in hepatoma and to 0.07 ± 0.05 pmol mg-' total liver protein in liver meta-0.1 1 10 NGA (umol 1-') Figure 1 Displacement of "9mTc-NGA binding to normal human liver (n = 10), hepatoma (n = 5) and liver metastasis (n = 7). 5 nmol 1 of b9mTc-NGA (total binding) were incubated with increasing concentrations of unlabelled NGA (non-specific binding) and the liver homogenate (200-500 jg 100 il-') in the presence of 50 mmol 1' Tris-HCI, pH 7.8, 5 mmol 1-' MgCl2, 5mmoll-' CaC12 and I moll' NaCI for 60min at room temperature (22°C). In the presence of 100-1,000timoll-' specific binding (the difference of total and non-specific binding) amounted to 91±7% in normal liver parenchyma and to <40% in malignant tissue.
stasis. The affinity constant (Kb) amounted to 1.21 ± 0.34 nmol I' in normal liver tissue, to 60.7 ± 11.8 nmol I' in hepatoma (P<0.001) and to 102 ± 35.4 nmol 1-' in liver metastasis (P<0.0001). In vivo binding (kinetic) studies In vivo simulation of QQmTc-NGA-kinetics allowed quantification of 9"Tc-NGA binding to HBP. In patients without liver malignancy a normal HBP-concentration of 0.94 ± 0.05 .mol 1I liver was found (Table I). However, in patients with hepatoma or liver metastasis a significantly decreased (P<0.01) NGA binding capacity was simulated. The in vivo measured HBP concentration amounted to 0.38 ± 0.06 ftmol I`liver in patients with hepatoma (Table II) and to 0.4 ± 0.1 ttmol I`in patients with liver metastasis (Table   III).

Liver morphology -SPECT scintigraphy
In vivo injection of 99mTc-NGA  to patients at a rate of 3.5 mg (50 nmol) demonstrated the liver to be the only site of tracer uptake. No tracer uptake was found by SPECT scintigraphy in a malignant liver area (Figure 2). Figure 2 9'Tc-NGA binding to the liver of a patient (L.G., 73 years, breast cancer) with liver metastasis. SPECT imaging revealed a lack of tracer uptake in the malignant area (arrows).

Discussion
The objective of the present study was the calculation of the in vitro and in vivo HBP density in the liver of patients with primary or secondary liver cancer through the binding characteristics of a new tracer, 9'9Tc-NGA. A direct comparison of HBP concentration estimated in vivo by 'mTc-NGA functional imaging and HBP concentration measured in vitro on a surgically removed liver biopsy specimen from the same patient with a normal liver showed good matching of these two values, arguing for a good estimation of HBP concentration in vivo. However, the in vivo estimate of HBP concentration was always about 75% of that measured in vitro. This finding in different normal livers indicates a slight but constant underestimation of HBP concentration in vivo. So far, exact quantification of liver function has not been possible in a reliable and clinically applicable way. Nevertheless, several clinical situations such as the evaluation of patients for liver transplantation would make quantitative liver function tests highly desirable. The development of a 99'Tc-labelled ligand of a hepatic receptor protein specific for galactose-terminated asialoglycoproteins could provide the basis for a new approach to the old problem of functional liver cell reserve.
Until now no real definite physiological role has yet been ascribed to HBP, although its study has provided many insights into the biology and pathobiology of the liver. HBP resides at the cell surface of hepatocytes where it recognises galactose-terminated glycoproteins (Schwartz et al., 1981;Stockert & Morell, 1983). Detailed examination of the cellular distribution revealed the parenchymal cells to be the exclusive sites of hepatic uptake (Stockert et al., 1984).
In this study direct binding experiments were performed to assess the feasibility of measuring 'Tc-NGA (specific) binding to human hepatic homogenates or plasma membranes. The high specificity (91 ± 7%) of the chemically synthesised NGA-ligand binding to normal liver tissue has provided the basis for studying changes in receptor density in cancer patients with or without liver metastasis. It is known that, once bound at the surface by HBP, glycoproteins are internalised and transported to prelysosomal vesicles where the majority of the ligand-receptor complex dissociates by a change to an acid pH (Wall et al., 1980). Thereafter, the receptor recycles to the plasma mem-brane (cell surface) while the ligand is degraded in the lysosomal compartment (Haimes et al., 1981). These findings make any comparison between the measured in vitro HBP density and in vivo HBP density somewhat difficult. Although a different in vitro binding behaviour was found between 99mTc-NGA binding to hepatic plasma membranes and to homogenates (Virgolini et al., 1989a), the total binding capacity was similarily pronounced between both preparations. On the other hand the in vivo binding capacities calculated from the time activity curves generated for the liver and precordium were comparable to the in vitro data obtained for normal liver tissue. Thus, in vivo measurement of HBP density using the naturally occurring 'mTc-NGA ligand was found to be a valid method for determination of hepatic function in patients with cancer. The extent of decrease in HBP concentration in patients with liver metastasis or hepatoma does express the non-functioning hepatic mass since metastasis or hepatoma do show no uptake in vivo visualised under the gamma-camera (SPECT scintigraphy) and in vitro no relevant binding activity. In other experiment-al studies on chemically induced carcinogenesis the HBP concentration was reduced too (Stockert & Becker, 1980). The 99'Tc-NGA kinetic analysis and determination of HBP density is therefore a simple and valid approach for quantification of liver function in patients with liver metastasis or hepatoma in vivo. The results suggest that in vivo estimation of HBP concentration in the liver by 99'Tc-NGA functional imaging might be an applicable method to determine functional liver cell mass.
In conclusion, 99'Tc-NGA functional liver imaging may provide a noninvasive means for the selection of medical or surgical management in patients with cancer.