Vol. 2, 1867-1871, November 1996 Clinical Cancer Research 1867

3 The abbreviation used is: EPO, eosinophil peroxidase.

New Marker for Blood Vessels in Human Ovarian and Endometrial

Cancers’

Michael Samoszuk,2 Fritz Lin, Paul Rim, and

Gary Strathearn

Pathology Department, University of California-Irvine, Irvine,

California 92697-4800 lM. S., F. L., P. Ri, and Isotope Products

Laboratories, Burbank, California 91504 [G. S.]

ABSTRACT

Angiogenesis plays a critical role in tumor biology and

may someday be a target for novel therapeutic interventions.

To date, however, relatively few markers have been identi-

fled that can specifically distinguish between microvessels in

benign versus malignant lesions. Here we report that thecationic heme-protein eosinophil peroxidase (EPO) was lo-

calized by in situ immunohistochemistry on the vascular

endothelial cells and/or connective tissue stroma in 16 of 16cases of human endometrial carcinoma and in 12 of 15 cases

of ovarian carcinoma. Similar deposits of EPO were notdetected in normal endometrial tissues or ovaries from fivehealthy subjects, in adjacent uninvolved tissues from fourtumor-bearing subjects, or in any normal organs from five

other subjects. These findings imply that eosinophil degran-ulation is a significant and previously unappreciated corn-ponent of the interaction between ovarian and endornetrialcancers and the host. Moreover, the abundant and highly

specific nature of the EPO deposition near and within the

microvessels of these cancers suggests that eosinophil de-

granulation is a new marker for tumor blood vessels that

potentially could be exploited to treat these important types

of cancers that currently lack highly effective therapies.

INTRODUCTION

A novel approach to the treatment of solid tumors would be

to target toxic drugs to the microvasculature of the tumor rather

than to the tumor cells themselves ( 1). As summarized by

Thorpe (2), such an approach could have several advantages.

For example, tumor vascular endothelial cells are directly ac-

cessible to drugs in the blood, whereas tumor cells are mostly

inaccessible, especially to large drugs such as monoclonal an-

tibodies. In addition, such an approach might be applicable to a

wide variety of solid tumors because all of them require a blood

supply for their survival and growth. Furthermore, the endothe-

hal cells within tumors are nonmalignant, thereby making it

unlikely that drug-resistant mutants would develop. Finally, this

approach has inherent amplification properties because many

Received 4/1 5/96; revised 6/20/96; accepted 8/16/96.

I Supported by Grant R01CA69079 from the National Cancer Institute.2 To whom requests for reprints should be addressed, at Pathology

Department. Medical Sciences D-440, University of California-Irvine,Irvine, CA 92697-4800. Phone: (714) 824-2323; Fax: (714) 824-2160.

tumor cells rely on each microvessel for nutrients and oxygen-

ation. Therefore, even partial damage to the blood vessels within

tumors might result in relatively massive destruction of tumor

cells.

The ideal target molecule for such a therapeutic approach

would be expressed at high density on the surface of vascular

endothelial cells in solid cancers and be absent on endothelial

cells in normal tissues. To date, several promising candidate

molecules with these properties have been identified in humans,

including tissue factor in breast cancer (3, 4), endoglin (5),

endosialin (6), an endoglin-like molecule (7), a fibronectin

isoform (8), receptors (9) for vascular endothelial cell growth

factor, and vascular endothelial growth factor itself (10).

Using a murine monoclonal antibody called EOS, we have

been studying the deposition ofEPO,3 a markedly cationic heme

protein, in human lymphomas (1 1, 12) and in breast cancers

(13). Normally, EPO is exclusively confined to the intracellular

granules within intact eosinophils. In a variety of pathological

conditions, however, eosinophils release their granule contents,

which then bind avidly to the anionic surfaces of adjacent cells.

Here we describe immunohistochemical evidence of a similar

degranulation phenomenon occurring in the microvasculature of

human ovarian and endometrial cancers.

MATERIALS AND METHODS

Reagents. Immunohistochemical staining of cryostat

sections of human ovarian and endometrial cancers was care-

fully performed, controlled, and evaluated using the same pro-

cedures and monoclonal antibodies that we recently used to

study human breast cancers ( I 3). In brief, an IgG 2a murine

monocbonal antibody called SF25.5 that binds to human EPO

was derived as described elsewhere ( I I ). This antibody is highly

specific for EPO and does not bind to any other cell types, to

myeloperoxidase, or to normal endometrium or ovaries when

radiolabeled and administered to human subjects ( I 2). Also

included in this study was a mouse monoclonal IgG-negative

control antibody (Dako Corp., Carpinteria, CA) diluted to the

same working concentration as the anti-EPO antibody (5 �ig/ml

PBS with 1% FCS). To visualize the architecture ofthe vascular

endothelium, separate serial sections of the tumors were also

stained with anti-CD34 (a vascular endothelial marker) mono-

clonal antibody (Becton Dickinson, San Jose, CA), using the

same conditions as were used with the other antibodies.

Tissue Specimens. In this study, we examined all of the

cryopreserved, primary ovarian and endometrial carcinomas

(listed in Table 1) that were available in the University of

California, Irvine, Cancer Center Human Tumor Bank (Orange,

CA). All of these cases were well-characterized by routine light

1868 Eosinophil Peroxidase in Tumor Blood Vessels

Table 1 Immunohi stochemical s tudies of human endometrial and ovarian can cers with EPO antibody

Diagnosis

No. of

cases

Degree of EPO deposition Location of EPO deposits

Extensive Present Absent Stroma Vasculature Tumor

Endometrial cancer

Endometrial adenocarcinoma

Adenosquamous carcinoma

Ovarian cancer

Clear cell carcinoma

Endometroid carcinoma

Mucinous carcinoma

Serous carcinoma

Serous papillary adenocarcinoma

15

I

5

4

1

4

I

14 1 0

1 0 0

4 0 1

2 2 0

0 0 1

3 0 1

1 0 0

15 13

1 1

4 2

4 3

0 0

3 21 1

4

0

2

0

0

0

1

microscopy of H&E-strained sections and by other special stud-

ies as appropriate. Normal endometrium, ovary, and other tis-

sues were obtained from autopsies of five nontumor-bearing

subjects. This study was approved by the Institutional Human

Subjects Review Board at the University of California, Irvine.

Immunohistology and Cytochemistry. The tumor and

control tissues were frozen in Histo-Prep embedding medium

(Fisher Chemical, Fair Lawn, NJ), sectioned at 2-6 p.m thick-

ness on a microtome, air dried on a glass slide, and fixed in

acetone for 1 mm. The cryostat sections were then submitted for

two separate assays. The first assay was a cytochemical proce-

dure for directly and specifically detecting EPO (I 1) using a

chromogenic substrate for EPO (aminoethyl carbazole) and 0.01

M potassium cyanide (which inhibits myeloperoxidase but notEPO). For optimal sensitivity, this procedure needed to be

performed within 24 h after tissue sectioning. These slides were

counterstained with hematoxylin prior to mounting.

The immunohistochemical procedure for detecting EPO

and CD34 in tissue sections used an avidin-biotin glucose oxi-

dase detection procedure (Vector Laboratories, Burlingame,

CA) to detect bound primary antibody and was followed by a

nuclear fast red counterstain. This detection method was se-

lected to avoid interference by endogenous peroxidase activity

that was present within the tissues.

The negative controls for each case consisted of tissue

sections that were incubated without the primary antibody or

with the irrelevant mouse monoclonal antibody at the same

concentration as SF25.5. A positive control was also included

and consisted of a cytopreparation of purified human eosino-

phils incubated with the SF25.5 antibody.

Each stained slide was reviewed by two observers. For

purposes of tabulation, EPO activity in each slide was classified

as extensive (present in every X250 microscopic field), present

(detectable in the tissue but not present in every field), or absent

(none visible). In addition, the microanatomic location (stromal,

peripheral, within tumors, within blood vessels, and others) of

the EPO was noted in each case.

RESULTS

The results of our immunohistochemical staining with the

EOS monoclonal antibody are summarized in Table 1. A strik-

ing and consistent finding was the highly specific and obvious

localization of finely granular EPO deposits in a linear pattern,

apparently along and within the microvasculature within the

tumors (Fig. 1, a-d). The distribution of staining was luminal

and abluminal and was most intense at the edges of the tumors.

There was no apparent localization to regions of necrosis. A

similar linear distribution of staining was seen in parallel sec-

tions of the same tissues that were stained with the CD34

monoclonal antibody (Fig. le), but the CD34 staining was

homogeneous and not granular or abluminal. Adjacent unin-

volved normal tissue was available from four of our tumor-

bearing patients, and there was no apparent staining for EPO in

these specimens. There was also no tissue staining when a

negative control monoclonal antibody was substituted for the

EOS monoclonal antibody (Fig. lJ).

At the stromal interface between tumor and peripheral

connective tissue, there were extensive fibrillar deposits of EPO

(Fig. 2a) that were remarkably similar to those that we have

described previously in Hodgkin’s disease and in some human

breast cancers (I I, 13), and faint granular staining frequently

extended into the adjacent tumor cells. An identical wavy pat-

tern of staining in stroma at the tumor edges was also seen when

the tissues were stained with CD34 antibody (Fig. 2b), but the

adjacent tumor cells had no granular staining. Normal endome-

tria and ovarian tissues obtained from autopsies of five nontu-

mor-bearing subjects were devoid of similar EPO deposits when

stained identically with the EOS antibody, as were five speci-

mens of normal heart, liver, spleen, lung, kidney, skin, muscle,

and stomach tissues.

In routinely stained tissue sections, intact eosinophils could

be identified only in about one-half of our study cases and were

easily overlooked unless diligently sought. Eosinophils in our

study cases were most evident at the edges of the tumor, in the

connective tissue stroma adjacent to blood vessels (Fig. 3a), and

extending into the fibrovascular cores between nests of tumor

cells (Fig. 3b). This distribution closely paralleled the distribu-

tion seen with immunochemical staining for EPO. Cytochemical

staining of tumor sections for EPO also confirmed the specific

presence of EPO and eosinophils within the same locations of

the tumors (Fig. 3c).

DISCUSSION

Our in situ immunohistochemical and cytochemical study

has demonstrated that eosinophil degranulation and deposition

of EPO occur within and near the blood vessels and connective

tissues of a significant proportion of human ovarian and endo-

metrial cancers. This deposition of EPO occurred in a distribu-

T

;--: - --�

I , #��‘ � .-�

�-, t� ;r’�� �Irq�’�

1’ --�‘�

,I

�.�-

- -

,4�-;

. - - - - ‘4 - ... .4

-S �4�_� . --- � -.. �

,��. - � . � -

b

q� ‘� � : -

-t� .f-�

�., c..

‘.7,,,, � .�‘-� .: - � #{149}. ,

� � �

,�_ ‘i,- ,,.

- - 4-_:

f

Fig. 1 Eosinophil peroxidase deposits in endometrial and ovarian cancers. In a-d. small blood vessels (marked by arrows) from four different tumorscontained luminal and abluminal blue granules, indicating the immunohistochemical detection of EPO. Similar deposition of EPO was generally

absent or decreased in the adjacent adenocarcinoma tumor cells (T) but usually extended into the fibrovascular stroma (S) between tumor cells (d).

A similar linear distribution ofpurple staining was observed when parallel sections were stained with a monoclonal antibody to CD34 (an endothelialcell marker), but the staining appeared to be exclusively luminal and homogeneous rather than granular (e). By contrast, incubation of the tumor

sections with a negative control monoclonal antibody produced no staining of tumor, blood vessels, or stroma (f). There was also no staining in normaltissues or in adjacent uninvolved tissues from four tumor-bearing subjects. All immunohistochemical photomicrographs are at an original

magnification of X250. Nuclear fast red counterstain and nitroblue tetrazolium substrate chromogen.

tion that was similar to the distribution of intact eosinophils seen blood vessels or tumor cells. This observation is significant

in routinely stained sections of one-half of the tumors that were because it implies that occult eosinophil degranulation is an

diligently examined. Thus, in view of the cytochemical and important and previously unsuspected part of the biological

histological confirmation, we conclude that eosinophils were the interaction between ovarian and endometrial cancers and the

source of the EPO rather than ectopic production of EPO by host.

Clinical Cancer Research 1869

I #{149}

�

k 4�

.- I

‘ -� eal’, .\ . � .:..‘#{149}::�::‘��� �_jc�c�i-

� � , .

. ‘Ii:. �,Jc.4-��-: - ‘�; � .- 4

-.�. , ., .�

::.�. � �.:‘ .

:�, . .‘� . ‘.- -‘ �“ #{149},, -‘ :. - � - .

‘!�i � . ., - ,- -- . ‘ �:

� .‘,tI.�,’ � . .c:�:�,�&:‘� : �

� .., . ...

� � .* �. � �.. � �

I

��.

�‘4 �

t.,

�. $;:-.

�- -

S

�( �� “ci’ � 4’�:k , b�

��,:4I�4 �. . I, � #{149} � . .

#{149}‘Jt1�’ .� ‘ - ., - - ‘:‘�� “-,� A. #{149}2� #{149}

.

4�.,._. 4b�

0“4

-.

S‘. C

. � b

Fig. 3 Histological and cytochemical confirmation of eosinophils in en-

dometrial and ovarian cancers. When diligently sought in H&E-stained

sections, eosinophils (some of which are denoted by arrows) could be

identified in the fibrovascular stroma adjacent to and within blood vessels

(V), as seen in this case of endometrial carcinoma (a). The eosinophils were

readily recognizable by their intensely red, granular cytoplasm and charac-

teristic bi-lobed nuclei. Eosinophils (some of which are denoted by arrows)

could also be detected in a linear distribution in the fibrovascular regions

between some nests of tumor cells (b), duplicating the immunohistochem-

ical distribution of EPO. In a case of ovarian carcinoma, specific cytochem-

ical staining for EPO confirmed the presence of EPO and eosinophils (c),

which stained orange-brown when incubated with the chromogenic sub-

strate aminoethylcarbazole and potassium cyanide (an inhibitor of my-

eloperoxidase). a and b, H&E stain, X450. c, aminoethylcarbazole chro-mogenic substrate and hematoxylin counterstain, X 200.

practical implications for the management of women with ovar-

ian or endometrial cancer.

For example, we have previously shown (12) that iv.

1870 Eosinophil Peroxidase in Tumor Blood Vessels

4. A�

T #{149}-‘ �:: � -...�f- - -

:� ,�:.4j�-’. .

Fig. 2 EPO deposition at the edge of an ovarian adenocarcinoma. EPOwas detectable in wavy fibrillar deposits (a) at the interface between

tumor cells (1) and adjacent connective tissue stroma (S). Fainter

granules of EPO staining also extended into the adjacent tumor cells. A

similar wavy pattern of staining was observed when parallel sections

were stained with anti-CD34, but there was no granular staining within

the tumor itself (b).

. %

In specific, the strikingly localized deposition of EPO that

we observed in the connective tissues and microvasculature of

human gynecological cancers suggests the intriguing possibility

that the degranulating eosinophils somehow participated in the

neovascularization and extracellular matrix formation that were

present within and around these tumors. This possibility is

strengthened by the numerous previous reports that eosinophils

accumulate, activate, and degranulate in the connective tissues

and blood vessels during normal wound healing (14, 15), in

utero during the implantation period (16, 17), in hypervascular

nasal polyposis ( 1 8), and in association with the connective

tissue matrix of a variety of other cancers (19 -2 1 ). The neo-

vascularization and extracellular matrix formation that are ob-

served in all of these conditions have been attributed to trans-

forming growth factors a and �3, both of which are abundantly

produced by eosinophils (14, 15, 18-24).

The findings presented in this report also suggest that EPO

deposition from degranulating eosinophils is a new marker that

can specifically distinguish between microvessels in benign

tissues versus ovarian and endometrial cancers, a property that

is not shared by other tumor markers that have previously been

described in these cancers (25-27). Consequently, EPO depos-

ited within tumor blood vessels could eventually have important

Clinical Cancer Research 1871

administered, radiolabeled EOS antibody localized in a highly

specific manner to tumor sites in 18 patients with a variety of

lymphomas that contained EPO deposits that were similar to

those that we have now also described in ovarian and endome-

trial cancers. Therefore, intravascular EPO may be an ideal

target for a radioimmunoconjugate to be used in women with

ovarian or endometrial cancers. In addition, we have recently

demonstrated that EPO deposition in murine plasmacytomas

sensitized the tumors to killing in vivo by hydrogen peroxide-

generating, anionic Stealth liposomes (28). Based on the data in

this report, a similar experimental approach might also have

value for targeting and chemically killing the EPO-coated blood

vessels in human gynecological cancers that are resistant to

more conventional treatments.

Finally, it is of interest that eosinophils have sometimes

been thought to exert a beneficial antitumor effect in vivo (29).

In view of our findings, additional studies are now warranted to

determine if eosinophils can modulate the clinical course of

ovarian and endometrial cancer in humans.

REFERENCES

I . Burrows, F. J., and Thorpe, P. E. Vascular targeting: a new approachto the therapy of solid tumors. Pharmacol. Ther., 64: 155-174, 1994.

2. Thorpe, P. Antibody-directed targeting of tumor vasculature. Proc.

Am. Assoc. Cancer Res., 37: 668, 1996.

3. Contrino, J., Hair, G., Kreutzer, D. I., and Rides, F. R. in situ detectionof tissue factor in vascular endothelial cells: correlation with the malignant

phenotype of human breast tissue. Nat. Med., 2: 209-213, 1996.

4. Folkman, J. Tumor angiogenesis and tissue factor. Nat. Med., 2:

167-168, 1996.

5. Burrows, F. J., Derbyshire, E. J., Tazzari, P. L., Amlot, P., Gazdar.A. F., King, S. W., Letarte, M., Vitetta, E. S., and Thorpe, P. E.Up-regulation of endoglin on vascular endothelial cells in human solid

tumors: implications for diagnosis and therapy. Clin. Cancer Res., I:

1623-1634, 1995.

6. Rettig, W. J., Garinchesa, P., Healy, J. H., Su, S. L., Jaffe, E. A., and

Old, L. J. Identification of endosialin, a cell surface glycoprotein ofvascular endothelial cells in human cancer. Proc. Natl. Acad. Sci. USA,

89: 10832-10836, 1992.

7. Wang, J. M., Kumar, S., Pye, D., Vanagthoven, A. J., Krupinski, J.,

Hunter, R. D. A monoclonal antibody detects heterogeneity in vascular

endothelium of tumours and normal tissues. Int. J. Cancer, 54: 363-370,

1993.

8. Carnemolla, B., Balza, E., Siri, A., Zardi, L., Nicrotra, M. R., Bigotti,

A., and Natali, P. G. A tumor-associated fibronectin isoform generatedby alternative splicing of messenger RNA precursors. J. Cell Biol., 108:

1138-1148, 1989.

9. Plate, K. H., Breier, G., Welch, H. A., and Risau, W. Vascularendothelial growth factor is a potential tumour angiogenesis factor in

human gliomas in vivo. Nature (Lond.), 359: 845-848, 1992.

10. Dvorak. H. J., Sioussat, T. M., Brown, L. F., Berse, B., Nagy, J. A.,

Sotrel, A., and Manseau, E. J. Distribution of vascular permeabilityfactor (vascular endothelial growth factor) in tumors: concentration in

tumor blood vessels. J. Exp. Med., 174: 1275-1278, 1991.

1 1 . Samoszuk, M. K., Lukes, R. J., and Nathwani, B. Extensive depo-sition of eosinophil peroxidase in Hodgkin’s and non-Hodgkin’s lym-

phomas. Am. J. Pathol., 125: 426-429, 1986.

12. Samoszuk, M., Anderson, A. L., Ramzi, E., Wang, F., Braunstein,

P., Majmundar, H., and Slater, L. Radioimmunodetection of Hodgkin’s

disease and non-Hodgkin’s lymphomas with monoclonal antibody to

eosinophil peroxidase. J. Nucl. Med., 34: 1246-1253, 1993.

13. Samoszuk, M., Nguyen, V., Gluzman, I., and Pham, J. H. Occult

deposition of eosinophil peroxidase in a subset of human breast carci-

nomas. Am. J. Pathol., 148: 701-706, 1996.

14. Wong, D. T. Sequential expression of transforming growth factors

a and 13 1 by eosinophils during cutaneous wound healing in the

hamster. Am. J. Pathol., 143: 130-142, 1993.

15. Todd, R. The eosinophil as a cellular source of transforming growth

factor a in healing cutaneous wounds. Am. J. Pathol., 138: 1307-13 13,

1991.

16. Lea. R. G., Stewart, F., Allen, W. R., Ohno, I., and Clark, D. A.

Accumulation of chromotrope 2R positive cells in equine endometrium

during early pregnancy and expression of transforming growth factor-

�32. J. Reprod. Fertil., 103: 339-347, 1995.

17. McMaster, M. T., Newton, R. C., Dey, S. K., and Andrews, G. K.

Activation and distribution of inflammatory cells in mouse uterus during

the preimplantation period. J. Immunol., 148: 1699-1705, 1992.

18. Elovic, A., Wong, D. T., Weller, P. F., Matossian, K., and Galli,

S. J. Expression of transforming growth factors-a and �3 I messengerRNA and product by eosinophils in nasal polyps. J. Allergy Clin.Immunol., 93: 864-869, 1994.

19. Kadin, M., Butmarc, J., Elovic, A.. and Wong, D. Eosinophils are

the major source of transforming growth factor43 1 in nodular sclerosing

Hodgkin’s disease. Am. J. Pathol., 142: 1 1-16, 1993.

20. Ghiabi, M., Gallagher, G. T., and Wong, D. T. Eosinophils, tissue

eosinophilia, and eosinophil-derived transforming growth factor a in

hamster oral carcinogenesis. Cancer Res., 52: 389-393, 1992.

21. Wong, D. T. TGF-a and oral carcinogenesis. Eur. J. Cancer Part B,

Oral Oncol., 29B: 3-7, 1993.

22. Wong, D. 1. Eosinophils from patients with blood eosinophilia

express transforming growth factor �3l. Blood, 78: 2702-2707, 1991.

23. Ohno, I. Eosinophils in chronically inflamed human upper airway

tissue express transforming growth factor �3l gene. J. Clin. Invest., 89:

1662-1668, 1992.

24. Wong, D. T. Human eosinophils express transforming growth factor

a. J. Exp. Med., 172: 673-681, 1990.

25. Sonoda, K., Nakashima, M., Kaku, T., Kamura T., Nakano, H.,and Watanabe, T. A novel tumor-associated antigen expressed in

human uterine and ovarian carcinomas. Cancer (Phila.), 77: 1501-

1509, 1996.

26. Saffari, B, Jones, L. A., El-Naggar, A., Felix, J. C., George, J., and

Press, M. F. Amplification and overexpression of HER-2/neu (c-erbB2)

in endometrial cancers: correlation with overall survival. Cancer Res.,

55. 5693-5698, 1995.

27. Quadri, S. M., Malik, A. B., Tang, X., Patenia, R., Freedman, R. S.,

and Vriesendorp, H. M. Preclinical analysis of intraperitoneal adminis-

tration of � � ‘In-labeled human tumor reactive monoclonal 1gM AC6C3-

2B12. Cancer Res., 55 (Suppl.): 5736s-5742s, 1995.

28. Samoszuk, M.. Wimley, W. C., and Nguyen, V. Eradication of

interleukin-5 transfected J558L plasmacytoma in mice by hydrogen

peroxide-generating Stealth liposomes. Cancer Res. , 56: 87-90.

1996.

29. Tepper, R. I., Coffman, R. L., and Leder, P. An eosinophil-depen-

dent mechanism for the antitumor effect of interleukin-4. Science

(Washington DC), 257: 548-551, 1992.