a unique morphology of epstein-barr virus-related …vol. 3, 607- 6 1 1, october/november 1994...

Vol. 3, 607- 6 1 1, October/November 1994 Cancer Epidemiology, Biomarkers & Prevention 607

A Unique Morphology of Epstein-Barr Virus-relatedEarly Gastric Carcinoma1

Yoshiko Uemura, Masayoshi Tokunaga,2Junnkou Arikawa, Noriko Yamamoto,Yasuhiro Hamasaki, Sadao Tanaka, Eiichi Sato,and Charles E. LandDepartment of Pathology, Kagoshima City Hospital, Kajiya-cho 20-1 7,

Kagoshima 892, Japan lY. U., M. T., J. A., N. Y., Y. HI; Department ofPathology, Kagoshima City Medical Association Hospital, Kagoshima 892,

Japan IS. TI; Second Department of Pathology, Kagoshima UniversityFaculty of Medicine, Kagoshima 890, Japan IE. 5.1; and RadiationEpidemiology Branch, National Cancer Institute, EPN 408, Bethesda,

Maryland 20892 IC. E. L.I

Abstract

Epstein-Barr virus (EBV) involvement in gastric cancer isdemonstrated by uniform presence of viral RNA incarcinoma cells as deteded by EBV-encoded small RNAin situ hybridization, monoclonal proliferation of EBV-infeded carcinoma cells, and elevated antibodies. Ourreview of selected early gastric cancers found that 46 of49 EBV-positive lesions (94%) but only four of 97 EBV-negative lesions (4%) conformed to a uniquemorphology, in which carcinoma cells formed lacepatterns of branching and/or anastomosing strudureswith lymphocytic infiltration in and around thecarcinoma nests in the mucosa. We conclude that EBV-related gastric carcinoma has a distind andcharacteristic morphology in the early stage ofdevelopment, and this lace pattern is a biomarker ofEBV involvement in early gastric cancer.

Introduction

The well known, and strongly EBV3-Iinked neoplasia ofBurkitt’s lymphoma and certain nasophamyngeab camcino-mas have characteristic momphologies; i.e., the “starry sky”appearance conferred by scattered macrophages in diffuseproliferation of pme-B lymphocytes in Bumkitt’s lymphomaand the unique LE among nasophamyngeab carcinomas.

Direct evidences of EBV involvement in a substantialminority of gastric carcinoma have been provided in recentstudies by the uniform presence of EBV in all carcinomacells by EBER-1 ISH, monocbonal EBV episomes detected bySouthern blot analysis, and the presence of elevated IgGantibodies against viral capsid antigen in response to anti-gens produced by replication of the virus in the carcinoma(1 , 2). EBV involvement was observed in 6.7% of gastric

Received 4/7/94; revised 7/8/94; accepted 7/1 1/94.i This study was supported in part by a Grant-in-Aid from the foundation for

the development of the community.2 To whom requests for reprints should be addressed.

3 The abbreviations used are: EBV, Epstein-Barr Virus; H & E, hematoxylin

and eosin; EBER-l ISH, EBV-encoded small RNA in situ hybridization; LE,bymphoepithelioma-like (tumors).

carcinomas in a barge Japanese series (3) and in 1 6% of asmaller North American series (4). EBV involvement washigher among males than females. By site, EBV prevalencewas highest for cancers in the gastric stump of previouslygastrectomized patients (5), followed by the cardia and themiddle stomach, and the antrum (3).

Extremely high (90%) levels of EBV involvement wereseen in gastric cancers with extensive lymphoid stroma andless differentiated features of carcinoma cells (LE tumors)(6); although EBV-positive cases were observed for all otherhistological types, there was significant nonhomogeneity bytype, with relatively high levels for moderately differenti-ated tubular adenocarcinoma and solid-type, poorly differ-entiated adenocarcinomas (3) (8 and 10%, respectively).

During the pathology review of the large Japanese

series (3), theme was repeated observation of a possiblyunique growth pattern in the mucosa, in which carcinomacells formed relatively small glands with back-to-back,anastomosing, or branching structures recognizable as a“lace pattern.” This pattern occurred in “early” tumors con-fined to the mucosa and submucosa, and a large number ofthe tumors with this property were found to be EBV rebated.The present study was conducted to investigate the repeat-ability of the pattern as a morphological event and to pro-vide a formal test of the possibility, raised by this serendip-itous observation, that the pattern might be a biomarker forEBV involvement in early gastric cancer.

Materials and Methods

A total of 49 EBV-positive early gastric carcinomas from 47patients (38 males and 9 females) was identified from ourfiles of 2344 cases diagnosed during 1976-i 993 and ex-amined by EBER-1 ISH. Twenty-one lesions were intramu-cosab carcinomas and the other 28, including 8 of 9 LE-typecarcinomas, had submucosab invasion.

For comparison, two EBV-negative early gastric carci-nomas from the files were individually, and exactly,matched to each positive case by sex and tumor site, and as

closely as possible by age. The matching was successful forcases occurring in the cardia, middle stomach, and antrum,but only one EBV-negative female early gastric remnantcase was found to match the two EBV-positive female cases.The only early gastric remnant cancers among males weretwo negative tumors, which were included as approximatematches. Thus, 15 EBV-negative female cancers, plus twonegative male gastric remnant cancers, were matched to 9EBV-positive female cancers, and 80 negative male cancerswere matched to 40 EBV-positive male cancers (Table 1).Two EBV-positive cases but no negative cases were repre-sented by two tumors each. The matching was quite closefor age at diagnosis, with mean 60.80 years for positivetumors and 60.75 for negative tumors. Mean ages weresimilarly close between positive and negative tumors when

on January 27, 2020. © 1994 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

http://cebp.aacrjournals.org/

608 EBV.related Early Gastric Cancer

Table 1 Number of cases and average ages (in parentheses), by EBVinvolvement, sex, and site

SiteEBV-po�-

Male

sitive-�-�

Female

EBV-ne

Male

gative

Female

Cardia 5(57.4) 0 10(56.8) 0

Middle 28 (61 .9) 3 (58.0) 56 (62.0) 6 (58.0)

Antrum 7(59.1) 4(60.0) 14(59.1) 8(60.1)

Stump 0 2 166.0) 2 165.0) 1 168.0)

subdivided by sex and site. EBV-positive and -negative early

tumors were not matched on presence on absence of sub-mucosab invasion. The proportion of EBV-negative tumors

with submucosal invasion was 50 of 97, compared to 28 of

49 positive tumors, a nonsignificant difference (P= 0.52). If

the 9 LE-like tumors (all of which were EBV positive and all

but one of which had submucosal invasion) are excluded,the proportions of positive and negative tumors with sub-mucosab invasion are nearly identical.

The intnamucosal lesions with on without submucosalinvasion were classified according to the classification

scheme of the Japanese Gastric Cancer Society (7) as pap-illamy adenocamcinoma (pap), well differentiated on modem-ately differentiated tubular adenocarcinoma (tubi on tub2,respectively), poorly d iffementiated adenocamci noma withno tubular structure (pom), and signet-ring cell carcinoma

(sig). This is a slightly different classification from that usedin our previous work (1 , 3, 5) reflecting the virtual absenceof certain types (e.g., muc, pon2) in early gastric cancers.

Historically, our pathology review reflects the fact thatthe present study developed from a serendipitous finding.Two of us (M T., Y. U.) first reviewed the main histologysections for EBV-positive early gastric cancers and foundexamples of lace pattern in the mucosal lesions. Theselesions were then jointly reviewed by four pathologist (S. T.,E. S., Y. U., M. T.) who agreed that this was an unusualgrowth pattern in gastric cancer histology. Next, a sample ofrecent EBV-negative early gastric cancer lesions was me-viewed (by Y. U. and M. T.), and the results were submitted

to this journal for publication. Following the suggestion of areferee, this control sample was dropped and a formalcontrol series of EBV-negative cancers was selected, asdescribed above. H & E-stained main sections of individualtumors were reviewed (by J. A. and M. T.) without knowl-

edge of EBV involvement, using a discussion microscope,and the results of that review, which agreed with the pre-vious one for those lesions included in both series, is thebasis for this article.

In the statistical analysis, the matching scheme was

broken and the matching variables (sex, site, and age) wereentered into the model as stratification or regression van-ables, as were level of invasion, presence on absence ofLE morphology, and histological type. The analyticalmethod of choice was logistic regression fitted by maximumlikelihood, using the General Modelling of Binomial Out-

comes program from the EPICURE package of statisticalprograms for analyses of epidemiobogical data (8). The mea-sure of association used was the odds ratio, defined as the

ratio of the odds corresponding to one particular combina-tion of factor levels divided by the odds for some other

combination.

Table 2 Distribution of early gastric carcinoma with lace pattern, byhistological type and EBV invoIvement�’

Lace pattern

Histological . .

type EBV-positive EBV-negative

Yes No Yes No

pap 1(1) 1 0 12

tubl 111) 1 0 26

tub2 (non-LE”) 26 (2) 1 3 13) 33

(LE) 6 0 0 0

por Inon-LE) 9 (31 0 1 11) 12

ILE) 3 0 0 0

sig 0 0 0 10

Total 46 17) 3 4 14) 93

a pap, papillary adenocarcinoma; tub 1 , well differentiated tubular adeno-

carcinoma; tub2, moderately differentiated tubular adenocarcinoma; por,poorly differentiated adenocarcinoma; sig, signet ring cell carcinoma. Num-bers in parentheses, numbers of carcinoma with focal lace pattern includedin the larger number to the left.b Lymphoepithelioma-like morphology with extensive lymphoid stroma and

less differentiated features of carcinoma cells.

Results

All but four of the EBV-positive lesions were classified inone of two histological types (tub2 and pon) whereas theEBV-negative tumors were more evenly distributed in allhistological types (Table 2). The positive versus negativedifference in distribution was highly significant statistically(P< 0.0001). Nine EBV-positive lesions, 6 classified as tub2and 3 as pom, and no EBV-negative lesions, conformed tothe LE morphology.

Detailed histological observation of representative sec-tions stained by H & E revealed that 50 of the 1 46 lesions(34%) conformed to the lace-pattern described in the “In-troduction” (Fig. 1 ). Although the determination was basedon H & E-stained main sections, the lace pattern was par-ticubarly clean in specimens stained by EBER-i ISH (Fig. 2).Theme was minimal disagreement among our reviewersabout presence on absence of lace-pattern growth in thernucosal layer of reviewed tumors. Individual carcinomacells were cuboidal and had oval hypemchmomatic nucleifocally with small but distinct nucleoli. These features wereseen even in two minimal carcinomas measuring less than5 mm in catheter (Fig. 3). Similar features were seen in themucosal lesions of all 9 LE-like carcinomas, although thesubmucosal invasion areas (Fig. 4) displayed less differen-tiated features similar to lymphoepithelioma arising in thenasophamynx. For all lace-pattern lesions, regardless ofwhether there was infiltration with LE-like pattern, exces-sive lymphocytic infiltrates were closely intermingled withthe carcinoma glands giving obscure outline. In 1 1 lesions,the lace pattern was observed in only a focal area of theordinal histological type associated with dense lymphocyticinfiltration.

The proportion of lesions with lace pattern did notdepend upon age (P= 0.71), sex (P= 0.71), site (P= 0.75),or presence of submucosal invasion (P = 0.25). These find-ings did not change after adjustment for other factors, in-cluding age, sex, site, submucosal invasion, histologicaltype, LE morphology, and/or EBV involvement (Table 3).There was significant variation by histological type

(P = .001 5), a finding that persisted after adjustment forother variables, including EBV involvement.



Cancer Epidemiology, Biomarkers & Prevention 609

Fig. 1 . Unique lace-pattern of EBV-positive intramucosal adenocarci-noma, showing anastomosing pattern

IH & E).

Fig. 2 . EBV-positive carcinoma with

lace-pattern demonstrated by EBER-l

ISH.

Forty-six of the 49 EBV-positive lesions, and only 4 ofthe 97 EBV-negative tumors, exhibited lace growth patternin the mucosal layer (Table 2), a highly significant differenceeven after adjustment for the other variables (P < 0.0001). Inthose EBV-negative lesions exhibiting the lace pattern, it oc-curred in only a focal area ofthe tumor. The same was seen in7 of 46 EBV-positive lesions with lace pattern growth. Of thetumors with lace pattern growth, the bymphocytic reaction in

the four EBV-negative tumors was less intense than that seen inthe EBV-positive lesions, including those with focal growth.

All LE tumors exhibited the lace pattern. Given thepreponderance of histological types tub2 and por amongthe EBV-positive lesions, and the close association be-tween EBV involvement and the lace pattern, it was to beexpected that the prevalence of lace-pattern growthshould be nonhomogeneous with respect to histological



Fig. 3 . Minimal carcinoma with

lace-pattern demonstrated by EBER-1ISH.

Fig. 4 . Submucosal )ymphoepithebi-

oma-like carcinoma demonstrated byEBER-1 ISH showing EBV-positiveundifferentiated carcinoma withmarked lymphoid stroma which is

EBV negative.

i#{149}.�

type (P < 0.002).The association between lace-patternappearance and histological type was not, however, me-moved by adjustment for EBV involvement, nor by ad-justment for both EBV and LE morphology (P < 0.02).

Moreover, EBV and histological type conformed to amultiplicative interaction model, in which the odds ratiofor EBV-positive versus EBV-negative status was fairlyhomogeneous across histological types, and those for the

various types did not vary appreciably by EBV involve-ment (P = 0.68 for departure from this model).

DiscussionA general observation in pathology is that advanced carci-nomas often change their histological picture, formingmixed patterns during invasive growth. Early carcinomas,

6 10 EBV-related Early Gastric Cancer

“b ,. � .�b’!’ ,

8. � ‘ . �1’. .. 4:v�� . - ..lkaAt.

. �:. �,, � - � � � �

p., � � ‘‘4�: (� � .�; � � -�

.. . ,�, S. , � � � i: . � � � � . �.�.. � . .,. .,. ...�

� �� #{149}:� �: � � � : ‘,�a�W , -. .

?‘ � . . � � � : . . -

�, . .: � .

. � � � �‘ � �



Cancer Epidemiology, Biomarkers & Prevention 6 1 1

Table 3 Distribution of early gastric carcinoma with lace pattern, bysex, site, age, depth of invasion, and EBV involvement

Lace patternlrotal Odds � valueFactor Level

EBV� EBV Total ratio Unadjusted Adjusted’

Sex Male 38/40 3/82 41/122 1.00

Female 8/9 1/15 9/24 1.19 0.71 0.53

Site Cardia 5/5 0/10 5/15 0.95

Middle 30/31 2/62 32/93 1.00

Antrum 10/11 2/22 12/33 1.09

Stump 1/2 0/3 1/5 0.48 0.75 0.30

Age (yr) 27-50 1 1/1 1 2/22 1 3/33 1 .23

51-60 9/10 0/20 9/30 0.81

61 -70 1 6/1 7 2/35 18/52 1 .00

71-87 10/11 0/20 10/31 0.90 0.71 0.30

Depth Mucosa 19/21 1/47 20/68 0.67

Submucosa 27/28 3/50 30/78 1 .00 0.25 0.67

Total 46/49 4/97 50/146

Odds Ratios 356.4 1 .00 <0.0001 <0.0001

a Adjusted for EBV, histological type, and LE morphology.

on the other hand, usually present a simple histologicalpicture and are more likely to reflect the influence of aunique carcinogenic event with consequently unique

morphology.The observed uniformity or (possibly) near-uniformity

of EBV expression in tumor cells within an EBV-positivecancer (1 , 2) suggests that EBV infection is a cause of gastriccancer or, at the very least, that if infection of tumorsoccurs, it is at a very early stage of tumor development andthat infected tumor cells have a strong reproductive advan-

tage over noninfected cells. The results of the present studysuggest that the lace pattern of mucosab tumor growth de-

scribed here is characteristic of EBV-mebated gastric cancersin their early stages, but not of EBV-negative tumors. Thispattern is therefore a biomaker of EBV involvement in earlygastric cancers.

The lace-pattern was observed in both intestinal (pap,tubi , tub2) and diffuse (pom) types in which EBV-positive

early carcinomas were represented, suggesting that thisgrowth pattern is not based on the cell origin oftumor cells.It would appear from our present and earlier results (1 ) thatEBV-infected tumor cells proliferate with identical morpho-logical expression in the early, mucosal stage and maychange momphobogically, to be classified as mostly tub2 andponl, during invasive growth. LE-like carcinomas, most ofwhich are EBV-melated (6), conform to the lace pattern in themucosa and dediffementiate and are associated with exten-sive bymphocytic infiltrates during invasive spread. SevenEBV-positive carcinomas and 4 EBV-negative ones exhib-ited the lace pattern in only a small part of the carcinoma,although in the positive tumors, EBER-1 ISH demonstratedEBV infection in all tumor cells. Thus, it cannot be said thatlace-pattern growth is unique to EBV-positive tumors, ab-

though it remains a possibility if the focal form is excludedfrom consideration.

In EBV-positive tumors, it is conceivable that marked

inflammatory reaction due to EBV infection may affect tu-mom histology resulting in an abnormal, lace-patterngrowth. This does not explain the much rarer instances offocal lace-pattern growth in EBV-negative tumors, butagents other than EBV might also produce inflammatoryreactions beading to that type of growth.

In our barge series (1, 3), no EBV-positive cells weredemonstrated by EBER-l ISH in precursor lesions such asatrophic gastnitis, intestinal metaplasia, and various types of

dysplastic changes observed in adjacent mucosa of gastriccancers, nor did we find EBV-positive dysplastic lesions notadjacent to cancers. Shibata et a!. (4), however, reportedfinding of EBER-l ISH-positive dysplastic lesions in their

smaller U.S. series. Possibly, EBV may interact with back-ground precursor changes in the mucosal epithelium at afairly late stage in carcinogenic development. Another pos-sibility is that different criteria have been used by differentinvestigators for what constitutes dysplasia and what iscancer. Further study, using serial cut sections of the stom-ach from cases with EBV-positive early gastric carcinomas,might resolve this issue.

We conclude that EBV-rebated gastric carcinoma is apeculiar cancer not only etiobogically but also morphobog-cabby, distinguished from other gastric cancers in much the

same way that Bumkitt’s lymphoma is distinguished fromother malignant lymphoma and lymphoepithelioma is dis-tinguished from other nasopharyngeab carcinoma. In the

early stages of tumor development, the lace pattern of

growth presented here is a biomarker of EBV involvement.

AcknowledgmentsWe thank the following laboratory staff in the Department of Pathology,Kagoshima City Hospital for their technical assistance: K. Yamashita,H. Nagasato, Y. Shinmura, M. Tsuruda, K. Sakanoue, T. Yamashita, N. Emori,K. Kubo, H. Fujisaki, and S. Oyamada.

References1 . Tokunaga, M., Land, C. E., Uemura, Y., et al. Epstein-Barr virus in gastric

carcinoma. Am. I. Pathol., 143: 1250-1254, 1993.

2. lmai, 5., Koizumi, S., Sugiura, M., et al. Gastric carcinoma: monocbonal

epithebial malignant cells expressing Epstein-Barr virus latent infection pro-

tein. Proc. NatI. Acad. Sci. in press, 1994.

3. Tokunaga, M., Uemura, Y., Tokudome, T., et al. Epstein-Barr virus related

gastric cancer in Japan: A molecular patho-epidemiobogical study. Acta

Pathol.lpn., 43:574-581, 1993.

4. Shibata, D., Weiss, L. M. Epstein-Barr virus-associated gastric adenocar-cinoma Am. J. Pathol., 140: 769-774, 1992.

5. Yamamoto, N., Tokunaga, M., Uemura, Y., et al. Epstein-Barr virus andgastric remnant cancer. Cancer (Phila.), 74: in press, 1994.

6. Shibata, D., Tokunaga, M., Uemura, Y., et al. : Association of Epstein-Barrvirus with undifferentiated gastric carcinoma with intense lymphoid infiltra-tion: lymphoepithebioma-like carcinoma. Am. I. Pathol., 139: 469-474,

1991.

7. Japanese Research Society for Gastric Cancer. The General Rules for

Gastric Cancer Studies in Surgery and Pathology, Ed. 1 2, Tokyo: Kanehara

Press (in Japanese), pp. 64-67, 1993.

8. Preston, D. L., Lubin, J. H., and Pierce, D. A. : Epicure User’s Guide.

Seattle: HiroSoft International, 1 991.



1994;3:607-611. Cancer Epidemiol Biomarkers Prev Y Uemura, M Tokunaga, J Arikawa, et al. carcinoma.A unique morphology of Epstein-Barr virus-related early gastric

Updated version

http://cebp.aacrjournals.org/content/3/7/607

Access the most recent version of this article at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/3/7/607To request permission to re-use all or part of this article, use this link



http://cebp.aacrjournals.org/cgi/alerts

mailto:[email protected]



a unique morphology of epstein-barr virus-related …vol. 3, 607- 6 1 1, october/november 1994...

Documents