Proc. Natl. Acad. Sci. USAVol. 92, pp 8999-9003, September 1995Medical Sciences

Targeting cancer micrometastases with monoclonal antibodies:A binding-site barrier

(pharmacology/metastasis/immunotherapy)

TSUNEO SAGA*t, RONALD D. NEUMANN*, TOSHIRO HEYAt§, JUN SATOt§, SEIGO KINUYA*, NHAT LE*,CHANG H. PAIK*, AND JOHN N. WEINSTEINtf*Nuclear Medicine Department, Clinical Center, and kLaboratory of Molecular Pharmacology, Developmental Therapeutics Program, Division of CancerTreatment, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892

Communicated by Judah Folkman, Harvard Medical School, Boston, MA, January 3, 1995

ABSTRACT Monoclonal antibodies penetrate bulky tu-mors poorly after intravenous administration, in part becauseof specific binding to the target antigen. Experiments pre-sented here demonstrate an analogous phenomenon in mi-crometastases; poor antibody penetration, attributable to a"binding-site barrier" phenomenon, can be seen in guinea pigmicrometastases as small as 300 ,um in diameter. Increasingthe dose of antibody can partially overcome this limitation,but at a cost in specificity.

When monoclonal antibodies (mAbs) are administered i.v. todetect or treat a solid tumor, they generally distribute non-uniformly in the mass (1-8). A number of possible reasons,reflecting the heterogeneity seen in almost all characteristicsof cancer (9-11), have been suggested: heterogeneous antigen(Ag) expression within the tumor; heterogeneous blood sup-ply; elevated interstitial pressure; and mechanical barriers-e.g., tight junctions between cells (12-23).

Mathematical analysis led to another suggestion. Accordingto the binding-site barrier hypothesis (19-23), the very fact ofsuccessful antibody (Ab) binding to target Ag (or bindingfollowed by metabolism) can retard deep penetration into atumor. That is, Ab molecules leaving the capillaries can besequestered by Ag and prevented from diffusing or beingcarried by convection to more distant Ag sites in the tumor.Lower Ab dose, higher binding affinity, and higher Ag densitywere all predicted to exacerbate the problem (19-23). Inprevious studies, we obtained direct experimental evidence ofa binding site barrier phenomenon in intradermally implantedLine 10 (L10) carcinoma of guinea pigs (5), and similar resultshave been obtained in other systems (6, 7). In the presentexperiments, we ask whether micrometastases are subject tothe same limitation of access and, if so, whether anything canbe done about it.

MATERIALS AND METHODSExperimental Lung Metastases. L10 is a chemically induced

guinea pig carcinoma of bile duct origin (24, 25). Ascites cellswere collected 2 weeks after i.p. injection of 4 x 106 cells. Toproduce lung metastases, fresh ascites cells (107; >95% viableby trypan blue exclusion) were injected via the penile vein inguinea pigs weighing 400-450 g.mAbs. D3 is a murine IgGI mAb raised against a 290-kDa

heterodimeric glycoprotein on the surface of L10 cells (26).The affinity constant and maximal binding of D3 were esti-mated as 109 M-1 and 3.55 x 105 molecules per cell, respec-tively (8). BL3, a murine IgG1 directed against a human B-celllymphoma idiotope, was used as an isotype-matched control(27).

Radioiodination and Immunoreactivity. mAbs were radio-iodinated with chloramine T (28). Labeled Ab was separatedfrom free iodine on a PD-10 column (Pharmacia LKB) and,when necessary, affinity purified on glutaraldehyde-fixed L10cells (8). No aggregates were found by size-exclusion HPLC.Specific activities of D3 and BL3 mAbs ranged from 61.8 to142.5 MBq/mg and from 104 to 138 MBq/mg, respectively.The immunoreactivity of labeled D3, determined by cellbinding assay, varied between 82% and 89% for differentpreparations. BL3 did not bind detectably to L10 cells.

Biodistribution of Labeled D3 and BL3 mAbs in GuineaPigs. For low-dose experiments, 30 ,ug of 131I-labeled D3, withor without 30 ,ug of 125I-labeled BL3, was injected i.v. intoguinea pigs with lung metastases. Six and 72 hr after injection,animals were sacrificed by CO2 inhalation. Organs were rap-idly removed and weighed, and tissue radioactivity was deter-mined. Lung tissues with metastases were frozen immediatelyin OCT compound (Tissue-Tek, Miles) and then processed forautoradiography and immunostaining. For high-dose experi-ments, 1000 ,ug of unlabeled D3 was added to 30 ,ug of labeledD3.Autoradiography and Immunostaining. Serial 20-,um sec-

tions were cut from the frozen lung tissues. For autoradiog-raphy, sections were placed on SB-5 film (Eastman Kodak) andexposed for 1-4 days. When 125I-labeled BL3 was coinjectedwith I311-labeled D3, images of the latter were obtained shortlyafter sacrifice, images of the former after eight 131I half-lives.The avidin-biotin peroxidase complex (ABC) method (29)

was used to compare Ab and Ag distribution. Briefly, tissuesections were acetone-fixed for 10 min and incubated withblocking serum for 20 min. To assess Ag expression, sectionswere incubated with D3 Ab (20 ,ug/ml) for 1 hr at roomtemperature, biotinylated horse anti-mouse IgG for 45 min,ABC for three 30-min periods, and diaminobenzidine (VectorLaboratories). When injected D3 was assessed, incubationwith D3 was omitted. Some sections were also stained forblood vessels by use of polyclonal rabbit anti-factor VIII andan alkaline phosphatase substrate, Vector Red (30).

RESULTS AND DISCUSSIONIn this study, we have assessed the distribution of i.v. admin-istered mAbs in experimental guinea pig lung micrometasta-ses. Two weeks after i.v. injection of L10 cells, multiplemetastases 100-1000 ,um in diameter could be seen in thelungs of all animals. This model (metastasis of guinea pigcancer in guinea pigs) is in many ways more relevant to the

Abbreviations: Ab, antibody; mAb, monoclonal Ab; Ag, antigen.tPresent address: Department of Nuclear Medicine, Kyoto UniversityHospital, Kyoto 606-01, Japan.§Present address: Takeda Inc., Yodogawa-ku, Osaka 532, Japan.ITo whom reprint requests should be addressed.

8999

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

9000 Medical Sciences: Saga et al.

Table 1. Biodistribution of 131I-labeled D3 Ab in guinea pigs with lung metastases

Low dose High dose

Tissue 6 hr 72 hr 6 hr 72 hr

% injected dose/gtBlood 1.849 ± 0.145 0.151 ± 0.054 1.722 ± 0.280 0.238 ± 0.010Lung (metastases) 0.991 ± 0.175 0.355 ± 0.074 0.914 ± 0.029 0.209 ± 0.046*Liver 0.522 ± 0.095 0.054 ± 0.022 0.477 ± 0.115 0.072 ± 0.008Kidney 0.465 ± 0.019 0.055 ± 0.019 0.480 ± 0.177 0.074 ± 0.003Muscle 0.057 ± 0.007 0.015 ± 0.004 0.056 ± 0.022 0.020 ± 0.001

Tissue/blood ratio*Lung/blood 0.533 ± 0.056 2.657 ± 1.305 0.540 ± 0.076 0.883 ± 0.236Liver/blood 0.284 ± 0.063 0.353 ± 0.019 0.274 ± 0.022 0.303 ± 0.043Kidney/blood 0.253 ± 0.019 0.368 ± 0.029 0.277 ± 0.024 0.310 ± 0.010**Muscle/blood 0.031 ± 0.004 0.100 ± 0.016 0.032 ± 0.008 0.083 ± 0.001

Localization index§1.319 ± 0.113 3.632 ± 1.490 ND ND

*, P < 0.001, compared with low dose; **, P < 0.05, compared with low dose [however, one of the 18pairwise comparisons for organs other than lung is likely to have P < 0.05 by chance alone (the "multiplecomparisons" problem)]; ND, not done.tMean ± SD for 3 or 4 guinea pigs.*Mean ± SD of ratios calculated for individual guinea pigs.§Lung/blood ratio for D3 Ab divided by that for BL3 Ab.

clinical setting (metastasis of human cancer in humans) thanare human cancer xenografts in nude mice.

Pharmacokinetics. Table 1 shows basic pharmacokineticdata for 1311-labeled D3. As expected, the distribution (ex-pressed as % injected dose/g) to organs other than lungappeared essentially independent of dose. Because microme-tastases were distributed diffusely throughout the lungs, wecould not measure radioactivity in them without includingnormal lung tissue. At 6 hr, D3 uptake in lung (includingmetastases) appeared to be independent of dose. At 72 hr,however, when radioactivity had largely cleared from normallung, the uptake was greater at the low dose (P < 0.001,unpaired two-tail t test), as would be expected if Ag bindingsites were becoming saturated. As also expected with Agsaturation, lung/blood ratios and localization indices weregreater at the low dose than at the high dose.

Microscopic Distribution of Bindable Ab (D3). To test thebinding-site barrier hypothesis, one must determine the his-tological distribution of four different entities: the injectedmAb, a nonbindable control immunoglobulin, the Ag, andblood vessels. Fig. la shows D3 Ag (dark brown) expressedquite uniformly in metastatic nodules. In contrast, Fig. lbshows injected D3 Ab (dark brown) at a dose of 30 ,ug localizedafter 6 hr in the peripheral one to two cell layers of the noduleand near blood vessels (stained red for factor VIII) at itscenter.

Fig. 1 shows poor penetration of Ab into the substance ofmicrometastases, as reported for bulky tumors (1-8, 13), andalso for spheroids (31, 32). However, these findings did notsuffice to identify the mechanism. Possibilities included (i) amechanical barrier-e.g., tight junctions between cells; (ii)rapid, nonspecific metabolism of Ab molecules as they enterthe metastases; and (iii) a binding-site barrier phenomenon,

for which Weinstein and coworkers (19-23) have predictedseverely limited penetration even over distances of a fewhundred micrometers.The arrows in Fig. 2 a and b indicate poor penetration of

low-dose D3 Ab into a neovascularized (33, 34) metastasisabout 700 ,Am in diameter. Also seen are smaller, apparentlynonvascular nodules with Ab only at the periphery. Ab distri-bution 72 hr after injection (Fig. 2 c and d) was too faint forconvincing photographic reproduction, but, when viewed inthe microscope, it was clearly heterogeneous within the nod-ules.

High-dose D3 appeared to have penetrated more deeply(two to six cell layers from the nodule margin or intranodularvessels) after 6 hr than had low-dose D3 (Fig. 2 e and f).However, appearances may be deceptive in this case. The34-fold higher dose would be expected to achieve a givenintensity of staining at greater depth in the nodule, even if thepenetration process were simply proportional to concentrationgradient (i.e., even in the absence of Ag saturation). By 72 hr(Fig. 2 g and h), Ab staining was essentially identical to that ofAg in most nodules.

Autoradiography reinforced the immunohistochemical find-ings of inhomogeneity. It showed poor penetration of low-dose131I-D3 at 6 hr after injection (Fig. 3 a and b) and more diffusebut still somewhat heterogeneous localization at 72 hr (Fig. 3c and d). The findings for high-dose D3 at 6 hr were similar(Fig. 3 e andf), consistent with the hypothesis that Ab had notsaturated very much of the Ag at either dose. In contrast, Fig.3 g and h indicate greater homogeneity at 72 hr.The qualitative relationships in Figs. 1-3 were substantiated

by semiquantitative microscopic analysis (Table 2). Both im-munohistochemistry and autoradiography showed predomi-nantly heterogeneous distribution in metastases after 6 hr. The

FIG. 1. Double-chromagen im-munostaining of serial sectionsfrom a lung metastasis in a guineapig injected i.v. with 30 jLg of D3Ab and sacrificed 6 hr later. (a)Staining for Ag (brown) and factorVIII (red). (b) Staining for Ab(brown) and factor VIII (red).(Bars = 300 gm.)

Proc. Natl. Acad. Sci. USA 92 (1995)

Proc. Natl. Acad. Sci. USA 92 (1995) 9001

__,0*. .} .. ..

lbs *

tc v .owz- Iv-$-,^gAgfis 8 at s +,>S w t.t ksXn~~~'d

.hs *>s st X %

. .ti t a* o t4

Ab

,............X;: *

low dose, 72 hr

v*;; S' A4 .r? Abi

-A!tb9> tJ~rX 4 ; * 2 high dose, 6 hr

~~~~high dose, 6hr

AgW.

4

* .S'.,

i.\

.I,Wk..

i.+ W e X w wss% *;t;h P

>. + r .' Xo

>^- - a h': ',: v i

,^ 'saX *

h S o.*.1 w. ^. %i.. .Ab

high dose, 72 hr

FIG. 2. Immunohistochemical staining of serial sections showing micrometastases in the lungs of guinea pigs injected i.v. with 30 ,g (a-d) or

1030 ,g (e-h) of D3 Ab. Animals were sacrificed 6 hr (a, b, e, andf) or 72 hr (c, d, g, and h) after injection. (a, c, e, and g) Staining for Ag. (b,d, f, and h) Staining for D3 Ab. Left-hand (Ag-stained) and right-hand (Ab-stained) panels were cut serially. (Bars = 500 ,um.)

somewhat greater percentage (25%) of nodules appearinghomogeneously stained at the high dose may have related tonodules cut off-center, given the thicker rims of Ab stainingseen. The difference between low-dose and high-dose autora-diograms at 72 hr (Fig. 3 d and h) indicates a degree of Agsaturation at the high Ab dose, consistent with the biodistri-bution data in Table 1. These findings provide circumstantialevidence against a mechanical barrier, which would be ex-pected to have the same proportional effect regardless of Abconcentration. However, they do not discriminate betweenbinding and nonspecific metabolism.Comparison of Bindable (D3) and Nonbindable (BL3) Abs.

For proof that poor penetration is caused by binding, one mustcompare the results for bindable and control immunoglobu-lins. Since unbound immunoglobulin tends to be lost fromhistological sections during immunostaining, we focused on

autoradiography for this comparison. Fig. 4a shows an auto-radiogram of lung tissue from a guinea pig injected with

low-dose 131I-BL3 and sacrificed 6 hr later. Comparison withhematoxylin staining of the same section (Fig. 4b) showsradioactivity principally in blood vessels and diffusely distrib-uted in each nodule, indicating efficient penetration by thenonbindable Ab.The marked difference in distribution between D3 and BL3

Abs was substantiated by coinjection of 1311-D3 and 125I-BL3.Consistent with the binding-site barrier hypothesis, 131I re-mained near the edges of nodules 6 hr later, whereas 125I wasdistributed diffusely throughout them (Fig. 4 c and d). mAbNP-4 has been reported to distribute uniformly in humanGW-39 colon cancer in nude mice when the nodules were <2mm in diameter (35). In those studies, nonuniform distributionin larger nodules and bulky tumor was attributed to centralnecrosis or increased interstitial pressure.The significance of poor penetration at the microscopic level

depends on context. Therapies based on Abs conjugated todrugs, toxins, or short-range radioisotopes are likely to be

Ab

low dose, 6 hr

Medical Sciences: Saga et al.

'k

9002 Medical Sciences: Saga et al.

a low dose, 6 hr

-- ;~' low dose, 72 hr

^~~~ ,is#.

high dose, 72 hrg

rendered less effective (36), insofar as the aim is to reach alltumor cells. However, if the aim is to deliver radiation tonearby tumor blood vessels, the opposite may apply.To circumvent poor penetration due to specific binding (or

specific binding followed by metabolism), the dose of Ab canbe increased. The low and high doses used here were equiv-alent on a weight basis to 5 and 180 mg, respectively, in a 70-kghuman. In these experiments, an even higher dose would havebeen required to saturate Ag fully and speed penetration.However, this dose adjustment is a balancing act because anincrease in dose is expected to decrease the specificity vis-a-visnonbindable Ab and distant sites of toxicity. In other words,the concept of a binding-site barrier implies that not all Ag isavailable for binding unless one is willing to accept lowerspecificity ratios and, hence, a narrower therapeutic window.

FIG. 3. Autoradiograms of mi-,t. crometastases in the lungs of

_I guinea pigs injected with 30 ,ug(a-d) or 1030 jug (e-h) of 1311..

labeled D3 Ab. The animals weresacrificed 6 hr (a, b, e, andf) or 72hr (c, d, g, and h) after injection. (a,c, e, and g) Low magnification. (b,

,,, d, f, and h) High magnification of7 areas shown by rectangles in the

. . 1 _< left-hand panels (Bars = 1000 ,um.)

Table 2. Semiquantitative analysis of the microscopic distributionof D3 Ab in micrometastases

% metastases with

Time, homogeneous D3hr Method Low dose High dose6 Immunohistochemical 0 (0/123) 25 (21/83)*

Autoradiographic 15 (12/79) 23 (17/75)**72 Immunohistochemical 2 (2/90) 80 (78/97)*

Autoradiographic 20 (17/96) 55 (42/77)*Cross-sections counted were 400-700 ,um in diameter, representing

metastases at least that large. Ab was detected immunohistochemicallyor autoradiographically. Significance of differences between high andlow dose (two-tail, Fisher's exact test): *, P < 0.001; **, P = 0.24 (notsignificant).

Proc. Natl. Acad. Sci. USA 92 (1995)

Proc. Natl. Acad. Sci. USA 92 (1995) 9003

1311D3c.y~~~~~~I

The present studies were done with radiolabeled mAbs, butthe same principles of microscopic pharmacology apply toother antibody conjugates (37-39) and to a variety of biologicalligands. They are pertinent, for example, to therapy withcytokines, whether administered exogenously or secreted invivo by genetically modified cells.

We thank L. A. Liotta, B. A. Chabner, E. A. Sausville, S. Sieber,R. L. Dedrick, C. Sung, J. A. Carrasquillo, J. Reynolds, K W. Kohn,B. Bunow, W. van Osdol, J. Casciari, H. Battifora, and S. Rosen forcritiquing the manuscript. We thank E. Owens for technical assistanceand M. Juweid for establishing the autoradiographic methods.

1. Jones, P. L., Gallagher, B. M. & Sands, H. (1986) Cancer Immu-nol. Immunother. 22, 139-143.

2. Pervez, S., Epenetos, A. A., Mooi, W. J., Evans, D. J., Rowlinson,G., Dhokia, B. & Krausz, T. (1988) Int. J. Cancer, Suppl. 3,23-29.

3. Ong, G. L. & Mattes, M. J. (1989) Cancer Res. 49, 4264-4273.4. Del Vecchio, S., Reynolds, J. C., Carrasquillo, J. A., Blasberg,

R. G., Neumann, R. D., Lotze, M. T., Bryant, G. J., Farkas, R. J.& Larson, S. M. (1989) Cancer Res. 49, 2783-2789.

5. Juweid, M., Neumann, R. D., Palk, C., Perez-Bacete, M. J., Sato,J., van Osdol, W. & Weinstein, J. N. (1992) Cancer Res. 52,5144-5153.

6. Blumenthal, R. D., Fand, I., Sharkey, R. M., Boerman, 0. C.,Kashi, R. & Goldenberg, D. M. (1991) Cancer Immunol. Immu-nother. 33, 351-358.

7. Shockley, T. R., Lin, K., Nagy, J. A., Tompkins, R. G., Yarmush,M. L. & Dvorak, H. F. (1992) Cancer Res. 52, 367-376.

8. Yokota, T., Milenic, D. E., Whitlow, M. & Schlom, J. (1992)Cancer Res. 52, 3402-3408.

9. Fidler, I. J., Gersten, D. M. & Hart, I. R. (1978)Adv. CancerRes.28, 149-250.

10. Liotta, L. A. (1992) Sci. Am. 266, 54-63.11. Poste, G. (1986) Cancer Treat. Rep. 70, 183-199.12. Schlom, J., Colcher, D., Hand, P. H., Wunderlich, D., Nuti, M. &

Teramoto, Y. A. (1983) in Understanding Breast Cancer: Clinicaland Laboratory Concepts, eds. Rich, M., Hager, J. C. & Furman-ski, P. (Dekker, New York), pp. 315-358.

13. Sands, H., Jones, P. L., Shah, S. A., Palme, D., Vessella, R. L. &Gallagher, B. M. (1988) Cancer Res. 48, 188-193.

14. Dvorak, H. F., Nagy, J. A., Dvorak, J. T. & Dvorak, A. M. (1988)Am. J. Pathol. 133, 95-109.

15. Jain, R. K. & Baxter, L. T. (1988) Cancer Res. 48, 7022-7032.16. Baxter, L. T. & Jain, R. K (1989) Microvasc. Res. 37, 77-104.17. Baxter, L. T. & Jain, R. K (1991) Microvasc. Res. 41, 5-23.

FIG. 4. Autoradiograms of micrometastasesin the lungs of guinea pigs injected with low-dose(30 ,ug) Ab and sacrificed 6 hr later. (a and b)Autoradiogram and hematoxylin staining, re-spectively, of the same section from an animalinjected with 131I-BL3. (c) Immediate autora-diography of a section from an animal coinjectedwith low-dose 131I-D3 Ab and low-dose 125I-BL3Ab. Radioisotope doses were selected so thatimmediate autoradiography represented almostexclusively the 131I-D3. (d) Autoradiography ofthe same section after 2 months, representingalmost exclusively 125I-BL3. BV, blood vessels;M, metastatic nodule. (Bars = 1000 ;,um.)

18. Sung, C., Shockley, T. R., Morrison, P. F., Dvorak, H. L.,Yarmush, M. L. & Dedrick, R. L. (1992) CancerRes. 52,377-384.

19. Weinstein, J. N., Eger, R. R., Covell, D. G., Black, C. D. V.,Mulshine, J., Carrasquillo, J. A., Larson, S. M. & Keenan, A. M.(1987) Ann. N.Y Acad. Sci. 507, 199-210.

20. Fujimori, K, Covell, D. G., Fletcher, J. E. & Weinstein, J. N.(1989) Cancer Res. 49, 5656-5663.

21. Fujimori, K, Covell, D. G., Fletcher, J. E. & Weinstein, J. N.(1990) J. Nucl. Med. 31, 1191-1198.

22. van Osdol, W., Fujimori, K & Weinstein, J. N. (1991) CancerRes.51, 4776-4784.

23. Weinstein, J. N. & van Osdol, W. (1992) Cancer Res. 52, Suppl.,2747-2751.

24. Rapp, H. J., Churchill, W. H., Jr., Kronman, B. S., Rolley, R. T.,Hammond, W. G. & Borsos, T. (1968) J. Natl. Cancer Inst. 41,1-11.

25. Dvorak, H. F., Form, D. M., Manseau, E. J. & Smith, B. D.(1984) J. Natl. Cancer Inst. 73, 1195-1205.

26. Bernhard, M. I., Foon, K. A., Oeltmann, T. N., Key, M. E.,Hwang, K. M., Clarke, G. C., Christensen, W. L., Hoyer, L. C.,Hanna, M. G., Jr., & Oldham, R. K. (1983) Cancer Res. 43,4420-4428.

27. Colcher, D., Esteban, J. M., Carrasquillo, J. A., Sugarbaker, P.,Reynolds, J. C., Bryant, G., Larson, S. M. & Schlom, J. (1987)Cancer Res. 47, 1185-1189.

28. Hunter, W. M. & Greenwood, F. C. (1962) Nature (London) 194,495-496.

29. Hsu, S. M., Raine, L. & Fanger, H. (1981) J. Histochem. Cyto-chem. 29, 577-580.

30. Sehested, M. & Hou-Jensen, K (1981) VirchowsArch. A. Pathol.Anat. Histol. 391, 217-225.

31. Kwok, C. S., Cole, S. E. & Liao, S.-K (1988) Cancer Res. 48,1856-1863.

32. Langmuir, V. K, McGann, J. K., Buchegger, F. & Sutherland,R. M. (1989) Cancer Res. 49, 3401-3406.

33. Folkman, J. (1985) Adv. Cancer Res. 43, 175-203.34. Folkman, J. & Shing, Y. (1992) J. Biol. Chem. 267, 10931-10934.35. Fand, I., Sharkey, R. M., Grundy, J. P. & Goldenberg, D. M.

(1992) Nucl. Med. Biol. 19, 87-99.36. Fujimori, K., Fisher, D. R. & Weinstein, J. N. (1991) CancerRes.

51, 4821-4827.37. Sung, C., Dedrick, R. L., Hall, W. A., Johnson, P. A. & Youle, J.

(1993) Cancer Res. 53, 2092-2099.38. Saga, T., Weinstein, J. N., Jeong, J. M., Heya, T., Lee, J. T., Le,

N., Paik, C. H., Sung, C. & Neumann, R. D. (1994) Cancer Res.54, 2160-2165.

39. Sung, C., van Osdol, W. W., Saga, T., Neumann, R. D., Dedrick,R. L. & Weinstein, J. N. (1994) Cancer Res. 54, 2166-2175.

Medical Sciences: Saga et aL