In Vivo expression of two novel tumor-associated antigens and their use in immunolocalization of human hepatocellular carcinoma

In Viuo Expression of Two Novel Tumor- Associated Antigens and Their Use in Immunolocalization of

Human Hepatocellular Carcinoma

HIROSHI TAKAHASHI, MEHMEI OLTLRK. R ~ R O N WILSON A T ~ L ~ H I K O MAKI, KAZLIE OZAWA. MITSURLJ KOIZUMI, KEIGO ENDO WII,I,IAM STRAIISb, r)ANIEI S H O r r V A i 1 AND .JACK WANDS

Molecular Hepatology Labornton cind Ma\aclic*wtt\ (knvrril H(i\pitni C uncvr Center Massac husetts General Hospital East, Charlestown, Massachuspth 0.21 5 4 I ) I L /\Lon of h iu lear M ~ d i ~ i n r Mn\mhuwtt\ ( h e m 1 Hospital, Boston Massachusetts 021 14

nnd Facult\ I,/ M p d r c i n e and S L C I ~ C K\oto I 'n/Lf>rsLl\ 606, Japan

We are investigating the antigenic changes on the cell surface of human hepatocytes that distinguish the normal from the transformed phenotype using monoclonal antibodies. In uiuo expression of antigens has been di- rectly assessed by in situ radioimmunohistology. This technique allows one to determine the distribution and density of antigen expression at the individual cellular level on fresh hepatoma and adjacent uninvolved liver tissue. We have found two antigens recognized by monoclonal antibodies XF-8 and AF-20 that are uniformly present on 15/15 hepatocellular carcinomas tested thus far. Most if not all tumor cells highly express these antigens. Such antigens were not evident on adjacent normal liver and the XF-8 epitope was not found on other normal human tissues. AF-20 antigen distribution revealed low-level expression on a subpopulation of cells in the zona glomerulosa of the adrenal gland and on crypt cells of the small intestinal tract. We have studied the capability of radiolabeled XF-8 and AF-20 monoclonal antibodies when administered either alone or in combination to localize a hepatitis B virus-related hepatocellular carcinoma cell line (FOCUS) grown as subcutaneous tumors in nude mice. Biodistribution experiments demonstrated an excellent localization to tumor of 15 to 22% of the injected dose of ""I-labeled antibodies. Indeed, it was possible to enhance the delivery of '''1 to the tumor cell surface by the use of XF-8 and AF- 20 in combination. Nuclear imaging studies showed sharp visualization of tumor and demonstrate that these monoclonal antibodies have sufficient specificity and sensitivity to be strongly considered as immunotargeting agents.

Primary hepatocellular carcinoma t HCC) i h one of the most common tumors in Subsaharan Africa and in Asia

Received August 15, 1988; accepted Oclcdwr 11. 19XX. This work was supported in part bv (;rants AA-02666. ('A-:33711

J . W. is the recipient of the Research S(.ientist Developmeni Award

t Present address: Department o f Mrdiciiie, Hadashah I itliversir? Hospital, Jerusalem, Israel.

Address reprint requests to: Jack Wantlh. 12.1.b.. Molecular Hepatol- ogy Laboratory, Massachusetts General Hospital Cancer Center. 149 13th St.. Charlestown, Massachusetts 01'13

and HD-20469 from the National Institutes of Health.

AA-00048.

( I.-:< 1 . One of the major risk factors for the development of HCC is chronic infection with hepat.itis B virus (HBVI (1 1. Affected individuals generally have a poor prognosis unless this tumor is detect,ed early (<3 cm) by ultrasound o r by wfetoprotein screening programs in high-risk pop- ulations (5). Effective treatment is usually restrict.ed to hepatic reaction, when possible, since conventional ra- diotherapy or chemotherapy results in little benefit (5).

Despite HCC being one of the most prevalent tumors in the world today, very litt,le is known regarding the antigenic properties of human hepatoma cells (6) . In order to study cell surface changes associated with malignant transformation of hepatocytes, we have produced several large libraries of monoclonal antibodies (MAbs) to a HRV-related HCC cell line (7). Using this panel of antihodies, we are searching for and characterizing antigens common to the transformed phenotype and that are not present to any appreciable degree on normal hepatocvtes. If such ant.igens appear and are identified, it may then be possible to devise immunotargeting strat- egies. at least in experimental animal models. In this report, we present our observations on two antigens that are highly expressed in ~1iuu and will explore their use in irnmunolocalization of HCC grown as solid tumors in nude mice.

MATERIALS AND METHODS All cells used for immunization o r monoclonal

antitiodv testing were maintained in Earle's modified Eagle's medium (M.A. Bioproducts, Walkersville. MD) supplemented with 1 0 L fetal bovine serum inactivated at 56"C, 10 F M non- essential amino acids, 1,000 units per ml of penicillin and 100 pg per in1 of streptomycin. Cells were harvested from the rnonoiayer cultures by washing three times with 20 m M phosphate-buffered saline (PBS). p H 7.2, followed by treatment with 0.04% F,DTA/Versene buffer in the absence of trypsin. Single cell suspensions thus obtained were used for immunization of BALB/c mice. The HCC cell line, FOCUS, was developed in our laboratory (7) and all other cell lines were ohtained from the American Type Culture Collection (Rock- ville. MD). For production of HCC in nude mice, FOCUS cells grown as monolayers in culture were treat,ed with 0.05% trypsin and 0.02';; EDT.4 solution i n Hanks' balanced salt solution

Cell Lines.

626 TAKAHASHI E T AL. HEPATOLOGY

(GIBCO Laboratories, Grand Island, NY) for 5 min. The cells were washed once with Earle's modified Eagle's medium, supplemented as described above, and resuspended to appropriate concentration in PBS. All the cell lines used were routinely tested for mycoplasma by nucleic acid hybridization technique using Mycoplasma T.C.11 (Gen-Probe, Inc., San Diego, CA). Also cell lines were tested by DNA staining (generously performed by the Mycoplasma Testing Laboratory, Massachusetts General Hospital), and cultures were found to be free of con- tamination.

The MAbs were produced as follows: A second passage of the FOCUS cell line, which had been derived from the original tumor and stored in liquid nitrogen, was harvested with EDTA/Versene buffer and used for primary immunizations. Six-week-old BALB/c female mice were given i.p. injections of 4.0 x lo6 FOCUS cells emulsified in 50% (v/v) complete Freund's adjuvant. After 6 to 10 weeks, the secondary immunizations were performed by an intrave- nous inoculation of 4.0 X lo6 cells in 200 pl PBS. Splenocytes were fused with the parent myeloma cell lines (NS1 or X63) 3 days later. The resulting hybridomas were selected and maintained in HAT medium as previously described (8). Antibody- producing hybridoma cells were screened by binding activity to a panel of human tumor cell lines and were cloned twice by limiting dilution. Ascites fluids were prepared by injection of hybridoma cells i.p. into BALB/c mice primed with 2,6,10,14- tetramethylpentadecane.

MAbs from double-cloned cell lines of the IgG subclass were purified by using Sepharose 4B staphylococcal protein A-affinity col- umns (Pharmacia Fine Chemicals, Piscataway, NJ). One to two milliliters of ascites fluid were placed on a column equilibrated with 0.1 M sodium phosphate buffer (pH 8.0). The ascites fluid was allowed to remain for 60 min at 4"C, and unbound material was removed by washing extensively with the same buffer. Bound mouse IgG isotypes were eluted with 0.05 M citrate buffer by a pH step gradient, using several void volumes each of various pH values (pH 6.0, pH 5.5 and pH 3.5) (9). Purified antibodies thus obtained were dialyzed overnight at 4°C against twice-normal concentrated saline, and the protein concentration was determined by the method of Lowry et al. (10).

MAbs were labeled with "'I or "'1 using the Iodogen method to a specific activity between 5 and 15 pCi per pg protein. Briefly, 100-p1 aliquots of Iodogen (1,3,4,6-tetrachloro-3,6-di- phenylglycoluril; Pierce Chemical Co., Rockford, IL) at 50 pg per ml in chloroform were evaporated to dryness under nitrogen gas in 10 x 75 mm glass tubes; then 500 mCi of N a T or Na':"I (Amersham Corp., Searle Div., Arlington Heights, IL) and 100 pg of MAb were added to the tubes and incubated for 6 min at 20°C. The radiolabeled MAb was then separated from free iodine on a PD-10 column (Pharmacia) equilibrated with 0.9% NaCl. Iodinated MAbs were always tested to assure that there was no loss of their immunoreactivity by direct radioimmunoassay as described below.

Direct Binding of Antibodies to Human-Derived Tu- mor Cell Lines. Direct binding of MAbs to human cancer cell lines was performed by radioimmunoassay using "'I-labeled MAb XF-8 and AF-20. The assay was carried out in filter- bottom 96-well plates (V & P Scientific, Inc., San Diego, CAI. Plates were first filled with 100 pl of bovine serum for 30 rnin at room temperature, in order to block nonspecific protein binding sites. Radiolabeled MAbs adjusted to lo5 cpm per 100 pl PBS containing 20% calf serum were incubated with 105 target cells for 1 hr at 20°C. The cells were then washed three times with PBS, and the radioactivity of dried filters was determined in a gamma counter. A radiolabeled nonrelevant

Monoclonal Antibodies.

Purification and Radiolabeling of Antibodies.

MAb designated B2TT (anti-tetanus toxoid IgGl and IgG2b) served as a negative control in the assay. Results were expressed as the ratio of cpm bound to human-derived tumor cell lines compared with a negative control monkey kidney cell line (Cos- 1) and designated as the signal/noise ratio.

Monoclonal Antibody Binding to the Membrane Preparations. Crude membrane fractions were prepared from normal human tissues as well as from hepatocellular carcinoma and adjacent normal liver obtained from surgical specimens (Kyoto University Hospital). Human HCC cell lines (FOCUS) grown as solid tumors in nude mice served as a positive control. Tissues were homogenized by Polytron ho- mogenizer in 10 volumes of 20 d PBS, pH 7.2, containing 0.1% NaN,, and were centrifuged a t 20,000 x g for 30 min. Homogenization and centrifugation were repeated three times. Subsequently, the pellet was suspended in PBS containing 20% glycerol and frozen at -80°C until further use. For the binding assay, this crude membrane preparation was diluted to a protein concentration of 1 mg per ml with 20% calf serum PBS. A direct binding assay was performed in 96-well filter plates using 50 p1 of membrane preparation. Specificity of binding was confirmed by competitive inhibition test with relevant cold antibody (MAb XF-8 or AF-20) compared with a nonrelevant MAb (B2TT). The difference between cpm bound per mg of tissue in the presence of cold relevant MAb and cpm bound per mg of tissue in the presence of nonrelevant MAb was designated as specific binding. Results were expressed by the ratio of specific binding in normal tissue to specific binding in FOCUS tumor [(cpm per mg of normal tissue)/(cpm per mg of tumor)] and expressed as a percentage. Positive results are binding >lo%. The use of human tissues in this investigation was approved by the Subcommittee on Human Studies at the Mas- sachusetts General Hospital.

Radioimmunohistology. Hepatocellular carcinoma, adjacent nontumor counterparts and other normal tissues were obtained fresh from surgery (Kyoto University Hospital) or as rapidly as possible from autopsies (kindly provided by Dr. Michael Kew). These tissues were immediately embedded in OCT compound (Miles Scientific, Naperville, IL) and frozen in liquid nitrogen and stored at -80°C. Tissue sections were cut from snap-frozen tissues, dried onto glass slides and fixed in cold acetone for 5 min. Sections were incubated overnight with IY5I-labeled XF-8 or AF-20 in 20 mM PBS, pH 7.2, containing 20% calf serum at 4°C. After washing in 20 d PBS, pH 7.2, for 60 min with three changes, the sections were fixed in a saline-containing 10% formalin solution for 10 min and washed again in PBS for 10 min. Tissues were stained with eosin (1% eosin Y in 100% ethanol) for 5 min, destained with 100% ethanol and dried. Sections were then dipped for 2 sec in autoradiographic emulsion (Kodak autoradiography emulsion type NTB 3; Eastman Kodak Co., Rochester, NY), diluted 1:l with distilled water and maintained at 45°C in a water bath. After the slides had been dried for 60 min, they were placed in light-tight wooden boxes (Fisher Scientific Co., Pittsburgh, PA) and exposed at 4°C for between 6 hr and 10 days. Exposed slides were then developed in Kodak Dektol, D19 for 90 sec, dipped in liquid hardener for 15 sec to stop reaction and fixed in Rapid Fix for 3 min. After washing sections in Hypo clearing agent for 2 min and in distilled water for 5 min, they were air dried and counterstained with Gill's no. 3 hematoxylin (Fisher) and mounted. In some experiments, autoradiography using Kodak X-ray film X-Omat-AR was performed before mounting tissue sections. In all experiments, "'I-labeled B2TT was used as a negative control MAb.

Specificity of Radioimmunohistology. Tissue sections were incubated with relevant MAb (XF-8 or AF-20) or nonrelevant MAb (BBTT) adjusted to 1.0 mg per ml in 20% calf

serum/PBS for 1 hr at room t,eiiiperature in wet (.hamher. After removing the MAb solution from ihr slide glass with a b s o r b ance with filter paper. tissue sections were incubated with "..'l- labeled XF-8 or AF-20 overnight at 4 c'; autoradiographs were made as described above.

Foiir- t o six-week-old BALB/c nude mice were obtained from the .Jackson Laboratory (Bar Harhor. ME). To establish xenografts of human HCC. the mice were injected subcutaneously in the left shoulder with 1 x 10' FOCUS cells in 0.2 ml of PBS. or explanted with a 2-mm'piecr obt.ained from a n exist,ing tumor-hearing mouse. Animals were maintained under specific-pat,hogen-free-conditions. Such t u - mors were used for both MA11 hiodistribution and histologic, studies when t,hey reached a size of ahout 10 x 10 mm. The animal studies done in this investigation were approved by the Subcommittee on Animal Welfare of' the Massachusetts Ckn- eral Hospital.

Ten microcuries of "'I-labeled specific antibody were injected i . \ , . via the tail vein into tumor-bearing nude mice. For dual tracer st,udies 1 1 ) . mice were given simultaneous injections of 10 p C i of "'I-1al)eled specific antibody of interest (XF-8 and AF-20) and 1 pCi of ' ' ' ' 1 nonspecific antibody (B2TT). T h e mice were killed and dis- sect.ed at 24,48 and 7'2 hr after injection. Tumor. hlood, thyroid. heart, lung, kidney, stomach, intestine. liver and spleen were weighed on an analytical balance and assayed for radioact ivit? using a multichannel, well-type ganima counter ( a window from 15 to 50 KeV for 'T and from 50 to :?JO KeV for I i i I ) . Results were expressed as follows: (a) specific activity of antihody in tumor and normal t.issues (cpm per mg); ib) percentage of' injected dose of '"I-labeled specific MAb per gni of tissue ("; of injected dose per gm of tissue). and (c ) localization index, 1.e. the ratio of specific '"I to nonspecific "I activity in tumor and in tissues divided by the same ratio in blood ( 1 1).

For nuclear imaging studies, 250 p('i of ""I-labeled MAbs were injected i.v. via tail vein into each tumor-bearing mouse. Images were obtained at 6, 2!4,48, 74 and 120 hr. The mice were anesthetized. and images were obtained in posterior views using a scintillation camera interfaced to a computer. Thyroid uptake of free '-'-'I was not blocked with KI. No background subtraction nor computer smoothing was employed. After imaging was performed. the ratio of radioactivity (cpm) in tumor t.o nontumor was calculated using a gamma camera interfaced to a computer.

At the terniiiiation of' nuclear imaging, tumors were excised and fixed i n 1 0 % formalin containing 0.9% NaCl and embedded in paraffin. Tissue sections ( 6 pm) were placed on glass slides, deparaffinated with xylene and rehydrated by washing through decreasing alcohol concentra- tions (100, 90 and 75%). Slides were dipped for 2 sec in photographic emulsion (Eastman Kodak ). and autoradiography was performed.

Tumor Xenografts.

In Viuo Biodistribution o f MAbs.

Nuclear Imaging.

Autoradiography.

RESlTLTS

Monoclonal Antibody Binding. Table 1 shows results of direct binding of MAbs XF-8 and AF-20 to human tumor cell lines. Both MAbs bound well to all human-derived hepatoma cell lines tested and demonstrate that these antigens are expressed on the surface of living hepatoma cells. MAbs XF-8 and AF-20 also bind tumor cell lines derived from colon, lung, kidney, ovary and uterus; however, the binding activities were generally lower than to hepatoma cells, with a notable exception of high binding to an adrenal adenocarcinoina cell line, SW-13 (Table 1)

' l (AHLE 1. Direct binding of XF-8 and AF-20 monoclonal antibodies to human adenocarcinoma cell lines

Origin Binding of XF-8 Binding of AF-20 (S/N) (WN)

Cell line6

19.53 t o..v; 17.59 * 1.21 11.:11 * 0.13

7 . 7 0 f i 1 .X N"" NT

10.72 f 0.711 4.99 f U.21 :{.?I * :{.22 2.44 * 0.2s :3.16 +- 0.64 7.35 f 0.6s 8.47 ? 0.44

:11.06 zk 1.39 2.46 f I).:3x 4.79 * 0.56

3.00 zk U.2S :;.05 f 0.4.1 5.57 f 0.14

1 1 . 1 0 t 2.42 N T

0.61 f 0.08

N T

:30.68 * 2.1*5 '2.88 f 1.83

13.22 * 1.87 "2.09 & 0.36 23.17 f .5.93 17.40 f 1.93 8.72 f 0.27 3.11 * 2.70

N T N T N T N T

4,554 f 2.03 7.51 f 2.24

N T 6.86 f 0.96 8.16 f 1.44 9.58 +- 1.28

13.08 f 3.11 NT

6.35 f 1.80 1.10 f 0.08

11.31 -c n.51

~ ~~ ~~~

Yl/lonkey kidney cell line (Cos-1) was used as a negative control. liesillis are expressed as signal t o noise ratio (S /N) . i.u. the ratio of cpin Iiound to adenocarcinoma cell line divided by cpm bound to Cos- 1 Positive results were S/N > 2.5, All cell lines are of human origin.

' N?' = not rested.

Monoclonal Antibody Binding to Membrane Preparations. We examined the capability of '"I- labeled XF-8 and AF-20 to bind to membranes prepared lrom FOCUS tumors, human HCC as well as other normal human tissues. Resu1t.s are expressed as 96 specific binding relative to FOCUS membranes and are displayed in Tahle 2. XF-8 demonst,rat.ed a high specific binding t,o human HCC (78.57 k 6.13%). We were unable t o show significant binding t,o other human normal tissues ( < l O % ) . Similarly, AF-20 demonstrated a high specific binding to human HCC but not to membranes derived from normal human tissues, with the notable exception o f the adrenal gland (Table 2 ) , where enhanced binding was observed.

Radioimmunohistology. Figure 1 shows an example of this technique by autoradiography and illustrates the tumor-specific binding of the MAbs. Hepat.0- cellular carcinoma tissue as well as nontumorous liver from the same patient was obtained at surgery, incubated with radiolabeled XF-8 or AF-20 MAb and exposed on X-ray film. Histologic examination of the nontumorous liver in this patient showed hepatic cirrhosis of the postnecrotic type. The expression of' XF-8 or AF-20 antigens is demonstrated as a dense radioactive spot on the film. As shown in Figure lI b0t.h XF-8 and AF-20 antigens are highly expressed on hepatoma tissue in uiuo hut not on the adjacent nontumorous part, of the liver. The binding of each '"I-labeled MAb was blocked by

628 TAKAHASHI ET AL. HEPATOLOGY

TABLE 2. Direct binding of XF-8 and AF-20 monoclonal antibodies to membranes prepared from human tissues and

FOCUS

’% specific binding-

XF-8 AF-20 Tissue

~~

FOCUS tumor HCC‘ Thyroid Lung Small intestine Colon Liver Pancreas Spleen Kidney Adrenal Myocardium Skeletal muscle

100.00 f (2,350 f 504)* 79.00 f 6.13

(10 < lo (10 < lo <lo 4 0 (10 <lo <lo 4 0 (10

100.00 f (10,395 f 742) 42.00 f 5.11

(10 4 0 <lo 4 0 <I0 <lo 4 0 <lo

<10 <10

18.0 f 3

“The ratio of specific binding to membrane derived from normal human tissue relative to FOCUS [(cpm per mg normal tissue)]/[(cpm per mg FOCUS tumor)] is expressed as % specific binding.

The number in parentheses represents the cpm bound to FOCUS tumor membranes by ‘2sII-labeled antibody. Positive binding to membrane preparations is >lo%.

HCC membranes prepared from a tumor.

’251-X F - 8

‘251-XF- 8 + Cold XF-8

‘251- X F- 8 + Cold 8 2 T T

4 2 5 1 - ~ ~ - 2 0

’251-AF - 20 +Cold AF- 20

Normal Hepatoma L iver

prior incubation of tissues with unlabeled “cold XF-8 or AF-20 but not by a nonrelevent MAb, thus demonstrating the specific binding to HCC tissue (Fig. 1).

Figure 2 demonstrates in uiuo expression of XF-8 antigen on human hepatoma tissues by radioimmunohistology compared with nontumorous counterparts of liver and other human normal tissues. This antigen was highly expressed on HCC tissues (Hepatoma 1 to 4 in Fig. 2) as well as FOCUS tumors grown in nude mice. In contrast, this antigen was not expressed on nontumorous adjacent counterparts (Normal Liver 1 to 4 in Fig. 2) nor on other normal human tissues. Tissue expression of the antigenic determinants recognized by ‘251-XF-8 was further confirmed histologically by in situ autoradiography. Figure 3 shows typical examples demonstrating dense antigen distribution on HCC tissue and not on an adjacent normal liver counterpart. Figure 3, B and A, are same tissue sections as “Hepatoma 3” and “Normal Liver 3,” respectively, shown in Figure 2. Figure 3C is a tissue section of the original HCC from which FOCUS was derived and illustrates the authenticity of antigen rep-

Skin

Muscle

8 a

Colon

@ Focus

Hepatoma 4

Lung K“ Normal L i ve r 1

Stomach Hepatoma 2

I .a ’251-AF - 20+ Cold B2TT

Liver

Pancreas

Hepatoma 3

Normal L iver 3

FIG. 1. In uiuo expression of XF-8 and AF-20 antigens in hepatoma tissue compared with adjacent uninvolved liver by radioimmunohistology. The binding is specific due to inhibition by preincubation with “cold” unlabeled specific antibody and not by a nonrelevant MAb (BZTT). Both antigens are highly expressed, as shown by the dense radioactive spots.

r, Spleen Normal L i ve r 4

FIG. 2. In uiuo expression of XF-8 antigen on different hepatomas compared to adjacent uninvolved liver counterparts as well as distribution on normal tissue. This autoradiograph was overexposed to show the presence of normal tissues. There is little if any expression of XF- 8 in normal tissue by this technique.

FIG. 3. In situ autoradiography i , f HC(’ dernonstratinp the distribution and density o f ?IF-8 antigen expression. ( A ) Normal liver counterpart. lox. (B) HCC. 1 0 ~ . (C) Original HCC 1rr-m which the FOClrS cell line was derived. 2 5 ~ . Note that most i f not all t umor cells highly express this antigen.

resentation on the cell line. Black grains in the figure correspond to “I-XF-8 bound t o tumor cells (Fig. 3, €3 and C). Note that most if not all cells illustrabe diffuse expression of XF-8 ant,igen. Indeed, this antigen was highly expressed in all 17 patients (100%) wit,h HCC examined. In these 17 tumor specimens, almost all cells expressed XF-8 similar to that observed in Figure 3. €3 and C. In contrast, this antigen was absent in 14/14 adjacent nontumorous liver. XF-8 antigen was also not detectable in normal human tissues by this technique, including esophagus, stomach, small int.estine, colon. pancreas, adrenal gland, kidney. spleen. t hvroid. skin, myocardium and skeltal muscle.

Figure 4 shows an autoradiograph demonstrating in uico expression of AF-20 antigen on human HCC tissues compared with adjacent liver counterparts and other human normal tissues. The antigen is highly expressed on FOCUS cells and HCC tissues obtained from patients

at the time of surgery (Hepatoma 2 to 4 in Fig. 4). It is of int,erest that the tissue section indicated as “Hepat,oma 1” in Figure 4 was t,he HCC from which the FOCUS cell line was derived. Thus, further evidence is presented that the antigens expressed in the original HCC are faithfully represented in FOCUS and that this antigen is highly conserved in other HCC tumors as well. Figure 5 is the histologic autoradiograph of AF-20 binding. Figure 5, A and H, is from Patient 4 (“Hepatoma 4” and “Normal Liver 4,” respectively) presented in Figure 4. Figure 5C is a high-power view of Figure 5A. Figure 5D is HCC t,issue from which the FOCUS cell line was derived. N0t.e that all tumors demonstrate diffuse cellular distribution of AF-20 antigen (Fig. 5, A, C and D). Indeed, all 14 cases (100% I of HCC tested were highly positive for AF- 20 antigen in t.he primary tumor. In these specimens, most if not all tumor cells expressed AF-20 as shown in Figure 5. This antigen was usually absent in the adjacent

TAKAHASHI ET AL. HEPATOLOGY 630

Muscle

Lung

Liver

Pancreas

Colon

Kidney

Adrenal

Spleen

Focus

Hepatoma 4

Hepatoma 2

Normal Liver 2

Hepatoma 3

Normal Liver 3

Hepatoma 4

Normal Liver 4

FIG. 4. In vivo expression of AF-20 antigen on hepatoma tissue, adjacent uninvolved liver and other normal human tissue. Hepatoma 1 is the original tumor from which the FOCUS cell line was derived. The autoradiograph is overexposed to show the presence of normal tissue. Note that AF-20 is expressed in adrenal (see text).

nontumorous liver as shown in Figure 5B. Indeed, 11 of 13 adjacent counterparts did not express this antigen. It was noteworthy, however, that two nontumorous liver samples were positive for AF-20 antigen; both were found to have active cirrhosis with hyperplastic nodules. How- ever, the expression of antigen was much greater in the primary tumor than in the adjacent cirrhotic nodules. There was low-level expression in the zona glomerulosa of the adrenal cortex and in a subpopulation of crypt cells of the intestinal epithelium; it was not present on other normal human tissues examined, including esophagus, stomach, pancreas, spleen, kidney, lung, myocardium, skeletal muscle and skin.

Nude mice bearing FOCUS tumors were given injections of l2'II-XF- 8, and the biodistribution in uiuo was determined. The results are summarized in Table 3, which displays the specific activity of '"I-XF-8 in tissues, percentage uptake of injected dose per gm of tissue and localization index at 72 hr. 1251-XF-8 demonstrated specific localization to FOCUS as shown by high specific activity in tumors

Biodistribution of XF-8 and AF-20.

(1,598.86 f 976.86 cpm per mg) as well as high uptake in tumor (19.62 f 11.98% of injected dose per gm of tissue) compared to normal tissues (1.37 * 0.51 to 7.07 f 2.25% of injected dose per gm of tissue). Further evidence for specific localization of lZ5I-XF-8 in human HCC was established by the localization index defined as the ratio of specific ('*'I-XF-S) to nonspecific (1311- B2TT) activity in tumor and normal tissues relative to blood. In tumor, the localization index was found to be 2.73 k 1.25, whereas in normal tissues, values varied between 0.93 f 0.13 and 1.35 f 0.66. These experiments demonstrate, therefore, a selective and specific binding of 12'I-XF-8 in uiuo to human HCC.

Specific activity, percentage of injected dose accumu- lated per gm of tissue and localization index of AF-20 were also investigated (Table 4). High specific uptake in tumor was demonstrated (741.90 f 284.24 cpm per mg) compared to the range observed in normal tissues (57.72 f 13.76 cpm per mg to 366.39 f 22.18 cpm per mg). We also found a high percentage of the injected dose local- izing to the tumor (21.81 f 8.39%). The specific localization index measured was 2.82 f 1.16 in tumor, whereas all normal tissues tested varied between 1.09 f 0.02 and 1.79 f 0.36.

Since both XF-8 and AF-20 demonstrated specific uptake in FOCUS tumors in uiuo, we were interested in investigating the effect of a combination of the two. We asked whether it was possible to deliver more lZ5I to the tumor cell surface by two MAbs directed against different antigens. When we injected a mixture of 10 pCi of '"1- labeled XF-8 and AF-20 and compared our results with 10 pCi of each MAb alone, we found that the localization index in tumor increased from 2.73 f 1.25 (XF-8) or 2.82 f 1.16 (AF-20) to 3.79 f 1.91 (mixture) without an increase of the localization index in normal tissues (Ta- ble 5, Experiment 1). To confirm and extend this observation, we performed another set of experiments using one-half the amount of radioactivity in the mixture (total radioactivity: 10 pCi) and compared results with 10 pCi of each MAb alone. Data are shown in Table 5 (Experi- ment 2). We found a localization index in the mixture (3.87 f 0.69) greater than with either antibody alone: XF-8 (2.33 f 0.33) and AF-20 (3.16 f 0.79).

Radioimaging was performed in nude mice bearing FOCUS tumors. In these experiments, approximately 250 pCi of lZ5I-labeled specific MAb or the same amount of '"I-labeled irrelevant MAb (B2TT) were injected i.v. into nude mice, when tumors reached an approximate size of 1.0 cm in diameter. Imaging was performed at 6, 24, 48, 72 and 120 hr using whole, intact '251-labeled MAbs without background subtraction. Figure 6 depicts the localization in tumors at 48 hr. XF-8 and AF-20 alone and in combination clearly visualized HCC. Ratio of cpm in tumor to an adjacent nontumor area was also calculated by the computer interfaced to the gamma camera. This ratio was 3.4 at 48 hr and 3.7 at 72 hr for XF-8 and 3.8 at 48 hr and 2.9 at 72 hr for AF-20. The ratio was higher in the mixture of the two, showing a value 3.8 at 48 hr and 4.3 at 72 hr.

After completion of the imaging studies, tumors were excised and autoradiographic investigations were performed. Figure 7 shows a representative example. Figure

FIG. 5. In situ autoradiography 01 ' surgically obtained HCC and adjacent uninvolved liver demonstrating the distribution of AF-20 antigen expression. A and B are from Patient 4 in Fig. 4. lox. C is a high-power view of A. 2 5 ~ . D is the original hepatoma from which the FOCUS cell line was derived. 25X. Note the high density of antigen expression and that most if not all tumor cells express AF-20.

7A is a hemat.oxylin-eosin staining of the subcutaneous grown FOCUS t,umor. In Figure 7B, Mack grains are observed over the t.umor, reflecting the in vico binding of '"1-XF-8. Figure 7C is a higher magnification of the FOCUS tumor, illustrating that most if not all h ino r cells show high-density binding of "'I-XF-X to the cell surface.

DISCUSSION

We have extended our library ok MAbs that recognize cell surface changes associated with transformation of human hepatocytes to the malignarit phenotype (6). XF- 8 as well as AF-20 are two such new MAlis. In this report, we have demonstrated that the antigens recognized by both antibodies are highly expressed 112 L'lLv) on HC(' tissue. Indeed, both antigens were expressed 111 all H(Y' tumors studied thus far that were obtained either at surgery or autopsy.

'r.\f$I,ic 3. Biodistribution of '"1-XF-8 in nude mice bea r ing h u m a n hepatocellular carc inomas

'% of injected LocalizaLion Specific binding index"

, , I Isslle (cpm,mg tissue) doseigm tissue

Iilorrtl Inrest ine St oniach [.lver Spleen K 1dnev Hrarr l.?lIlg Neck '1 umor

1,827 f 648 111 f 12 190 f 89 :I40 f 11 1 28x6 f 118 518 f 176 475 -t 164 576 t 1s:; 47.2 f 100

1,599 f 977

1 2 * 7.95 1.37 * 0.51 ?..:I4 f 1.10 '4.17 f 1 -111

:3.r11 ? 1.45 ii.:15 -t 2.11; 5.83 f 2.01 7.07 f 2.2.; ,458 ? 1,2:<

19.62 f 11.98

1.00 f 0.00 1.07 * 0.07 1.21 * 0.07 0.93 rt (1.13 1.40 r 0.20 1.07 f 0.05 1.46 f 0.18 1.17 f 0.05 1.35 k 0.6G 2.72 I l . 2 5

h'lice received simultaneous injectiun of 10 F('i of ""I-labeled XF- ti (spec~fic MAbJ and 1 p('i o f '."I-labeled R T T (nonspecific MAh) and were killed at 7 2 hr after injection. A11 data are expressed as mean i S.1) oi'three animals.


TABLE 4. Biodistribution of Iz5I-AF-2O in nude mice bearing human hepatocellular carcinomas

Tissue

Blood Intestine Stomach Liver Spleen Kidney Heart Lung Neck Tumor

70 of injected dose/gm tissue

Specific binding

(epm/mg tissue)

959 f 165 58 f 14 17 f 30

199 f 54 175 f 64 247 f 43 257 2 76 366 f 22 245 f 61 742 f 285

27.97 f 4.84 1.70 -C 0.40 2.26 f 0.90 5.85 f 1.59 5.13 f 1.90 7.25 f 1.26 7.55 f 2.22

10.77 f 0.65 7.21 f 1.79

21.81 f 8.39

Localization index"

1.00 f 0.00 1.09 f 0.02 1.26 f 0.05 1.16 f 0.11 1.48 f 0.11 1.09 f 0.04 1.28 f 0.13 1.11 f 0.03 1.79 f 0.36 2.82 f 1.16

a Mice received simultaneous injection of 10 pCi of 12511-AF-20 (specific MAb) and 1 pCi of I3'I-B2TT (nonspecific MAb) and were killed at 72 hr after injection. All data are expressed as mean f S.D. of three animals.

TABLE 5. Biodistribution of '261-XF-8, AF-20 and mixture in nude mice bearing human hepatocellular carcinoma

Localization index Tissue

XF-8 AF-20 Mixture

Experiment 1 Blood 1.00 f 0.00 1.00 f 0.00 1.00 f 0.00 Intestine 1.07 f 0.07 1.09 f 0.02 1.09 f 0.04 Stomach 1.21 -+ 0.07 1.26 f 0.05 1.30 f 0.13 Liver 0.93 f 0.13 1.16 f 0.11 1.11 f 0.04 Spleen 1.40 f 0.20 1.48 f 0.11 1.25 f 0.04 Kidney 1.07 f 0.05 1.09 f 0.04 1.12 f 0.02 Heart 1.46 f 0.18 1.28 f 0.13 1.17 f 0.04 Lung 1.17 f 0.05 1.11 f 0.03 1.09 f 0.02 Neck 1.35 f 0.66 1.79 f 0.36 1.46 f 0.13 Tumor 2.72 f 1.25 2.82 f 1.16 3.79 f 1.91

Blood 1.00 f 0.00 1.00 f 0.00 1.00 f 0.00 Intestine 1.21 f 0.08 1.12 2 0.01 1.26 f 0.07 Stomach 1.41 C 0.18 1.20 C 0.08 1.64 f 0.49 Liver 1.08 f 0.12 1.22 f 0.13 1.36 f 0.14 Spleen 1.03 f 0.06 1.09 f 0.06 1.36 f 0.21 Kidney 1.12 f 0.01 1.05 f 0.01 1.16 C 0.15 Heart 0.86 f 0.19 1.01 f 0.00 1.32 f 0.23 Lung 1.05 f 0.01 1.09 f 0.03 1.24 f 0.09 Neck 1.20 f 0.08 1.17 f 0.02 1.39 * 0.18 Tumor 2.33 f 0.33 3.16 f 0.79 3.87 2 0.69

Experiment 2

XF-8 antigen was not expressed on the uninvolved liver nor on other normal human tissues, including pe- ripheral blood lymphocytes, spleen, skin, myocardium, skeletal muscle, esophagus, stomach, small intestine, colon, pancreas, adrenal gland and kidney, as shown by both radioimmunohistology and membrane binding studies. XF-8 antigen was stable after treatment of tissue with formaldehyde or paraformaldehyde and thus may be demonstrated on paraffin-embedded tissues (data not shown). However, there was some loss in reactivity compared to fresh-frozen specimens. To date, this antigen has not been identifiable by Western blotting, by im- munoprecipitation of cell surface '"I-labeled proteins nor after metabolic labeling with 35S-methionine or 3H- glucosamine (data not shown), even though we could

FIG. 6. Nuclear imaging of FOCUS HCC tumor at 48 hr after injection of '251-XF-8, AF-20 or a mixture of the two. Note the sharp demarcation of the tumors.

identify other antigens by the same experimental proce- dures (6). I t is not clear whether the negative data are due to the character of the antigen or the methods employed or both. Thus, the molecular characteristics of XF-8 antigen have yet to be defined. We are led to believe from the density of antigen expression on HCC, however, that XF-8 represents an important cell surface change associated with hepatocyte transformation to the malignant phenotype.

AF-20 was also found to be highly expressed on HCC tissue and most if not all tumor cells demonstrated high- density cell surface labeling with lZ5I-AF-2O. In situ radioimmunohistology proved to be very useful in demonstrating the density and normal tissue distribution, since immunoperoxidase staining even on unfixed tissue gave very low levels of expression, particularly on HCC. In- deed, we would have underestimated the level of antigen expression by both immunoperoxidase and membrane binding studies. Based on the experience we have accu- mulated thus far, the radioimmunohistologic method is the preferred technique for assessment of tissue distribution of some tumor antigens by MAbs. The molecular characteristics of AF-20 have been determined (6).

L o 1 9. No. 4. 1989

FIG. 7. Autoradiograph of FOCI'S t r imor grown in nude mice 7 2 11 after the injection ( ~ 1 '-''T-XF-X. ( A ) Hematoxylin and eosin staining of' rumor. 1 0 ~ . (B) Accumulation of' '' I -NF 5 in the tumor 1 0 ~ . (('1 Cellular densltv 01'"I-SF-X binding to HCC. ?5X.

AF-20 is a 116-kD glycoprotein highly expressed on the cell surface on HCC. We know that AF-20 and XF-8 are different antigens based on their tissue and tumor distribution as well as hy competitive inhibit.ion experiment where cold XF-8 antibodv does not inhibit t,he binding of '""I-AF-20 to its antigen on FOC'IJS cells and L ~ ~ C C

uersa (6). Finally, AF-20 was found to be expressed a t low levels (compared to HCC by radioimmunohistology )

on crypt cells of the small intest,inal t.ract, on a subpopulation of cells in the zona glomerulosa of the adrenal gland and in two uninvolved livers with hvperplastic regenerating nodules. It is possible that AF-20 represents an antigen that is more proliferation associated than XF- 8; further studies are required. Finallv, as shown by the results in Table 1, both antigens are expressed by other human adenocarcinoma cell lines of endodermal origin. and they are not, therefore, hepatoma specific.

The density of antigen expression on individual tumor cells was impressive and unexpected. particularl\, when one considers such a uniform distribution between various hepatomas. U'e found that 100°F of' hepatomas

studied thus far express XF-8 and AF-20 antigen with little if any expression on adjacent uninvolved liver. These antigens appear strongly associated with hepatocyte transformation to the malignant phenotype. Studies regarding the function of these two antigens will be of interest.

In order for MAbs to he potentially useful as immunotargeting agents, several essential properties may be required. In the first place, it would be desirable to have most i f not all t,umor cells express antigen(s) with a relative uniform distribution between various tumors. Second. and more important, i t appears that XF-8 and AF-20 antigens are not shed from the cell surface at least in amounts detectable by immunoblot.ting on nitrocellu- lose membranes or by immunoradiometric assays (6). Such experiment.s have been performed on HCC cell culture supernatants as well as human serum from a hepatoma patient (data not shown). Thus, i.v. injected radiolabeled antibodies would not he hound in immune complexes, and would presumably reach the tumor cell surface in i:iiw without interference. Third, the antibod-


ies must have high affinity to localize to and remain on tumor cells 48 to 72 hr after i.v. injection. Finally, enough antibody must reach the tumor to allow visualization by the gamma camera. It appears that these two antibodies have this capability, since 15 to 22% of the initial injected dose localizes to the tumor 72 hr later. Since we used an inefficient isotope for imaging (12, 13), whole antibody rather than fragments (14) and no computer-smoothing techniques, our imaging results should only improve with these further refinements.

One of the interesting experimental questions that may be approached based on our results is whether HCC may be “attacked from two different sites on the cell surface. Our biodistribution studies suggest that greater amounts of lz5I may localize to the hepatoma cell by the use of these two MAbs in combination. This observation would allow one to evaluate tumor growth following experimental manipulation with one antibody, since it is now possible to determine tumor size by imaging with the other. The capability to target to more than one site on the tumor cell surface needs further investigation, particularly in evaluating the effect of these MAbs in immunotherapy. For example, one antibody could be conjugated to a radioisotope such as I 3 l I (15) and the other to a drug such as adriamycin (16). The fact that both AF-20 and XF-8 localize well to HCC in a nude mouse model and have been demonstrated to be highly expressed in vivo presents unique opportunities to explore their use in therapy of a rapidly fatal disease.

REFERENCES

1. Di Bisceglie AM, Rustgi VK, Hoofnagle JH , et al. Hepatocellular

2. Fu-Sun Y, Kong-Nien S. Epidemiology and early diagnosis of carcinoma. Ann Intern Med 1988; 108:390-401.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

primary liver cancer in China. Adv Cancer Res 1986; 47:297-329. Robinson WD, Miller RH, Klote L, et al. Viral hepatitis and liver disease. New York: Gmne & Stratton, 1984 245-263. Beasley RP. Hepatitis B virus as the etiologic agent in hepatocellular carcinoma-epidemiologic considerations. Hepatology 1982;

Tang Z-Y. Subclinical hepatocellular carcinoma. Berlin, Heidel- berg: Springer-Verlag, 1985. Wilson B, Ozturk M, Takahashi H, et al. Cell-surface changes associated with transformation of human hepatocytes to the malignant phenotype. Proc Natl Acad Sci USA 1988; 85:3140-3144. Lun H, Isselbacher KJ, Wands JR, et al. Establishment and characterization of a new human hepatocellular carcinoma cell line. In Vitro 1984; 20493-504. Wands JR, Zurawski VR Jr. High affinity monoclonal antibodies to hepatitis B surface antigen (HBsAg) produced by somatic cell hybrids. Gastroenterology 1981; 80225-232. Ey PL, Prowse SJ, Jenkin CR. Isolation of pure IgG1, IgG2, and IgGPb immunoglobulins from mouse serum using protein A-sepharose. Immunochemistry 1978; 15:429-436. Lowry OH, Rosebrough NJ, Farr AL, et al. Protein measurement with the Folin phenol reagent. J Biol Chem 1951; 193:265-275. Moshakis V, McIlhinney RA, Raghavan D, et al. Localization of human xenografts after i.v. administration of radiolabeled monoclonal antibodies. Br J Cancer 1981; 4491-99. Esteban JM, Schlom J , Gansow OA, et al. New method for the chelation of Indium-1 11 to monoclonal antibodies. Biodistribution and imaging of athymic mice bearing human colon carcinoma xenografts. J Nucl Med 1987; 28861-870. Kanellos J , Pietersz GA, McKenzie IF, et al. Coupling of the 99m- technetium-nitrido group to monoclonal antibody and use of the complexes for the detection of tumors in mice. J Natl Inst Cancer 1986; 77:431-439. Larson MS. Radiolabeled monoclonal anti-tumor antibodies in diagnosis and therapy. J Nucl Med 1985; 26538-545. Takahashi H, Carlson RI, Shouval D, et al. Radioimmunotherapy of established human hepatocellular carcinoma (HCC) with a monoclonal antibody directed against a p50 cell surface antigen (Abstract). Hepatology 1986; 6:1186. Shouval D, Adler R, Wands JR, et al. Adriamycin conjugates of monoclonal antibodies to hepatoma associated proteins. Proc Natl Acad Sci USA 1988; 85:8276-8280.

2:21S-26S.

Documents

In Vivo expression of two novel tumor-associated antigens and their use in immunolocalization of human hepatocellular carcinoma