Gastric neckcell and intestinal goblet cell in gastric ... · Hughes,Bhathal and sigmoid colon (four cases) obtained from routinelyprocessedtissue werealso studied. Two micron thick

7 Clin Pathol 1997;50:741-748

Gastric mucous neck cell and intestinal goblet cellphenotypes in gastric adenocarcinoma

N R Hughes, P S Bhathal

AbstractAim-To investigate the phenotype of cellscomprising diffuse and intestinal-typegastric cancers using monoclonal anti-bodies to two antigens. One antigen(designated D10) is characteristic of gas-tric mucous neck cells, cardiac glands,pyloric glands, and Brunner's glands. Thesecond antigen (designated 17NM) is spe-cific to the mucous vacuole of intestinalgoblet cells.Methods-Thirty two gastrectomy speci-mens with adenocarcinoma were studied.Serial paraffin sections were stained im-munohistochemically for D10 and 17NMand histochemically for acid and neutralmucins. The cancers were classified histo-logically as of either diffuse or intestinaltype according to Lauren.Results-Of 15 diffuse-type gastric carci-nomas, 11 showed the majority of cancercells staining for Dl0 while four were typi-cal signet ring cell cancers staining pre-dominantly for 17NM; five tumoursdisplayed both phenotypes with the twophenotypes segregated in different areasof the tumours. In contrast, of 16intestinal-type cancers, six expressed17NM, three Dl0, five neither antigen, andtwo expressed both antigens. Oneindeterminate-type cancer expressed bothantigens. The staining of individual cellsfor D1O and 17NM was mutually exclusivein both diffuse and intestinal types. Incontrast to the diffuse cancers, intestinal-type cancers typically expressed eitherantigen only in occasional small groups ofcells and individual cells.Conclusions-In disease, the gastric stemcell can assume the capacity ofthe duode-nal stem cell for divergent differentiationinto either intestinal goblet cells (forexample, as in intestinal metaplasia) orBrunner's gland cells (for example, as inpyloric gland/Brunner's gland metapla-sia). With neoplastic transformation, thispotential for divergent differentiation ismaintained and gives rise to diffuse-typecancers that display either the D1O pheno-type, the 17NM phenotype, or the clonalexpression of both phenotypes. In themore cell cohesive (intestinal-type) tu-mours, differentiation for antigen expres-sion is poorly developed and morefrequently directed towards the intestinalgoblet cell phenotype.( Clin Pathol 1997;50:741-748)

Keywords: gastric carcinoma; mucous neck cells; intes-tinal goblet cells

The histological classification of gastric adeno-carcinomas into Lauren's diffuse or intestinaltype' has some biological significance as theintestinal-type cancer shows variation in geo-graphical incidence and has a somewhat differ-ent clinical behaviour to the diffuse type.2 3Although intestinal metaplasia is common instomachs bearing intestinal-type cancers,4 thecell lineages of the two classes of tumour andthe reason for the frequent histological andhistochemical heterogeneity seen in individualspecimens of gastric carcinoma remain un-clear. We report on the use of two phenotypicmarkers that identify two distinct neoplasticcell populations in gastric carcinoma and throwsome light on the histogenesis of thesetumours.

MethodsTissues were collected over a four year periodfrom gastrectomy specimens (as approved bythe Ethics Committee of The Royal Mel-bourne Hospital) from 22 men (28-80 yearsold) and 10 women (45-82 years old) admittedto this hospital for surgery for gastric adenocar-cinoma. The specimens were consecutive andunselected except that the fresh specimenneeded to be available immediately after resec-tion, and the duty pathologist was prepared torelease tissue for the research after diagnosticneeds had been met. Blocks of tumour andadjacent mucosa and, when possible, longstrips of non-involved mucosa were taken fromthe fresh specimens for this study. In addition,sections of the cancers were taken subsequentlyfrom the routinely processed paraffin blocksused in producing the surgical pathologyreport.

For optimal antigen preservation, the freshtissue was either fixed for three to four hours incold fixative containing 2% paraformaldehyde,75% ethanol, and 0.4% cetylpyridinium chlo-ride (EPF), or in Zamboni's fixative5 for onehour followed by a further three hours in EPF.Tissues were dehydrated in isopropanol,cleared in chloroform, and vacuum embeddedin four changes of paraffin wax over a period ofone hour. The long strips ofmucosa were fixed,dissected free from the muscularis propria,fashioned into "Swiss rolls", and processed forsectioning as described above.

Paraffin sections of four signet ring cell can-cers of the caecum and sigmoid colon, adeno-carcinomas of the ascending colon (one case)

Department ofAnatomical Pathology,The Royal MelbourneHospital, Parkville,Victoria 3050,Australia

Correspondence to:Professor Bhathal.

Accepted for publication13 May 1997

741

on March 30, 2020 by guest. P

rotected by copyright.http://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.50.9.741 on 1 Septem

ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Hughes, Bhathal

and sigmoid colon (four cases) obtained fromroutinely processed tissue were also studied.Two micron thick serial paraffin sections

were stained immunohistochemically for anti-gens D 10, 1 7NM, and the antigen detected bythe monoclonal antibody 7HGM-1D7-B; inaddition, the sections were stained histochemi-cally for acid and neutral mucins (AB-dPASstain).

PRODUCTION OF MONOCLONAL ANTIBODIESPreparation of hybridoma 5HL-5D 11 -D10,which recognises antigen DI 0, has beendescribed elsewhere.6 The hybridoma wasgrown intraperitoneally in BALB/c mice andthe same batch of ascites fluid used for the pre-vious study was used in the present study.Hybridoma 1 7NM-20-20 recognises antigen

17NM and was produced by immunising micewith isolated colonic glands from subjects withcolonic carcinoma; the procedures for prepar-ing cell hybrids, screening, and cloning thishybridoma have been described elsewhere.7The same batch of ascites fluid used in the pre-vious study was used here.A third monclonal antibody (7HGM-

1D7-B) used in this study was produced byimmunising BALB/c mice with epithelium iso-lated from surgically resected human gall blad-ders. Cell hybrids were prepared from thespleens of these mice and screened against amosaic of tissues as described for antibody5HL-5D1 1-D 10.6 Hybridoma 7HGM-1D7-Bwas selected, cloned by limiting dilution,grown intraperitoneally in BALB/c mice, andthe batch of the resulting ascites fluid usedhere.

IMMUNOHISTOCHEMISTRYA four stage immunohistochemical stainingprocedure was used as previously described.6Briefly, sections were incubated sequentially inTris buffer (140 mM NaCl, 50 mM Tris,2.7 mM KC1, 0.01% merthiolate, pH 7.2)containing 10% fetal calf serum (FCS), mousemonoclonal antibody, rabbit immunoglobulinsto mouse immunoglobulins (Z259; Dako,Botany, Australia), 1.5% H202 in Tris buffer,swine immunoglobulins to rabbit immu-noglobulins (Z 196; Dako) and rabbit horserad-ish peroxidase antiperoxidase (Zi 13; Dako).The substrate 3-3'-diaminobenzidine (DAB)was used to demonstrate the sites of boundenzyme. All immunological reagents werediluted in Tris buffer containing 10% FCS and0.1% NaN3 (0.04% NaN, was used with theperoxidase antiperoxidase solution) and thesections were washed between changes for 10minutes in Tris buffer. The three monoclonalantibodies were all used at a dilution of 1/400.A 1/400 dilution of ascites fluid produced by anon-reactive mouse hybridoma or the Trisbuffer containing 10% FCS were used as nega-tive controls.To obtain maximum staining of antigens in

paraffin sections of the routinely fixed andprocessed tissue, sections were taken down towater, placed in 0.05 M citrate buffer, pH 6,which had been brought to the boil in a micro-wave oven, again heated to boiling point in the

oven over one minute, and allowed to cool for20-30 minutes before staining.8 Followingimmunohistochemical staining, sections werecounter stained for five minutes with 1%Alcian blue in 3% acetic acid (pH 2.5) for acidmucins and then with haematoxylin.Double staining for Dl 0 and proliferating

cell nuclear antigen (PCNA) was done bysequentially reacting sections with a 1/500dilution of antibody to PCNA (Cat. No. ABT152; American Biotech, Florida, USA), rabbitantibodies to mouse immunoglobulin (Z259;Dako), and gold labelled goat antibodies torabbit immunoglobulins (Cat. No. RPN 470;Amersham International, Amersham, UK),and then enhancing the gold label with theIntenSE M kit (Amersham International).Sections were rinsed in 0.1 M glycine HCIbuffer, pH 2.5, for two minutes, in Tris buffer,pH 7.2, for five minutes, and stained for D10as before.

HISTOLOGICAL CLASSIFICATION OF GASTRICCANCERSThe tumours were classified according toLauren' and the WHO International Histologi-cal Classification of Tumours,9 either as beingof a diffuse or intestinal-type, or of an indeter-minate category, based on the predominanthistological growth pattern ofthe tumours seenin the sections stained with antibody 7HGM-1D7-B. The indeterminate category refers totumours with about equal proportions of thediffuse and intestinal growth patterns.9 Otherhistological features of the tumours wererecorded as solid, tubular, papillary, ormucinous,9 and the general degree of differen-tiation of the tumours was noted.

HISTOCHEMICAL STAININGSections were digested with salivary amylasefor one hour at 37°C, washed in distilled water,stained with Alcian blue, pH 2.5, and then bythe periodic acid Schiff (PAS) reaction'" forcarbohydrate (AB-dPAS stain).

ResultsANTIGEN EXPRESSION IN NORMAL TISSUEAs described elsewhere, the normal tissue dis-tribution of D10 is limited to gastric mucousneck cells, the glands of the cardia and pylorus,Brunner's glands, peribiliary glands, and theperiductal glands of the pancreas; staining forDl 0 was present as cytoplasmic granules in allsamples of the tissues tested.6 The distributionofD1 0 in the 32 specimens of stomach studiedhere was similar to that reported previously.Figure 1A illustrates the spatial relation ofmucous neck cells staining forD 10 with cells ofthe proliferative zone of gastric corpus mucosa,as revealed by staining for PCNA.The normal tissue distribution of antigen

17NM has previously been shown to be limitedto the mucous vacuole of the goblet cells of thesmall and large intestine. It was demonstratedin all 136 samples of normal colon and 32 ofsmall bowel, in the goblet cells of intestinalmetaplasia in six stomachs and in 61 of101 primary colorectal adenocarcinomas.7 "17NM was not present in the normal mucosa

742



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Cell phenotypes in gastric carcinoma

of any of the 32 stomachs examined in the cur-rent study, but was present in the goblet cells ofall examples of intestinal metaplasia seen ineight of the 15 stomachs with a diffuse-typecancer and seven of 13 stomachs with anintestinal-type cancer for which "Swiss rolls" of

macroscopically normal mucosa were availablefor study (table 1).Antibody 7HGM-1D7-B stained the cyto-

plasm of the epithelial cells of stomach,intestine, breast, bile ducts, and the transitionalepithelium of the kidney calyces, as well as

q~~~~~~~~e~~~~~~; ~~~O'

AA

Figure 1 Immunohistochemical staining of sections ofhuman stomach for antigens DlO, 17NM, and PCNA. (A) Corpusmucosa showing cell nuclei stainedfor PCNA and mucous neck cells for antigen D10. PCNA positive nuclei (black) aremainly present in the lower pit and isthmus regions and above the mucous neck cells (staining brown for Dl 0). (B) Atypical diffuse-type gastric carcinoma (specimen 6 in table 1) showing most tumour cells staining strongly for D10. (C-G)Serial sections of a Dl 0 positive, diffuse-type gastric carcinoma (specimen 8). (C) Shows that most of the cancer cells stainfor D10 and are widely disseminated throughout the stomach wall. Two areas (arrows) near the surface of the tumour stainfor acid mucins with the Alcian blue counter stain and contain relatively few cells stainingfor D10. (D) Shows that cells inthe two Alcian blue positive areas marked in C, together with scattered, solitary cells in the adjacent tissue, stain strongly forantigen 17NM (therefore, this Dl 0 positive tumour includes foci of diffuse tumour cells of the intestinal goblet cellphenotype). (E) A higher power view ofC to show part of the deeper edge of one of the Alcian blue stained areas and thepresence of numerous large, round, Alcian blue positive cancer cells which, unlike the surrounding smaller tumour cells, donot stain for antigen D10. (F) A higher power view ofD corresponding to part of the section area seen in E to show thatthe majority of the Alcian blue positive, D10 negative cells seen in E stain strongly for 1 7NM. (G) An area of a sectionstained with the AB-dPAS stain corresponding to part of that seen in E; this shows the large Akian blue positive cells seenin E staining purple for acid mucins, and the smaller, adjacent cancer cells in the right half of the photomicrograph(corresponding to the DJ0 positive ceUs seen in E) staining redfor neutral mucins. (H) An intestinal-type cancer (specimen13) showing the focal cell stainingfor D10 (arrows) typical of intestinal-type cancers that express D10. (Originalmagnifications: A, B, E, lj G,H x250; C, D x12.)

743



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Hughes, Bhathal

Table 1 Antigen expression and mucin synthesis by 15 diffuse, 16 intestinal, and oneindeterminate type gastric carcinomas

Antigen OtherSpecimen histological Intestinalnumbers* D10 17NM Mucin type featurest metaplasiat

Diffuse type6 +++§ M, Al +9 +++ N +10 +++ N Tubular +14 +++ N18 +++ N23 +++ - M7 ++ N

32 ++ - M +12 +++ + M Tubular +8 +++ ++ N, A

17 +++ ++ M,A +2 + +++ A +5 + +++ A Tubular

16 +++ M +24 +++ AIntestinal type25 ++ N, A Tubular +31 ++ - N Tubular, pd13 + A Papillary11 + + N, A Tubular +28 + ++ N, M Tubular, pd -

3 - +++ M Mucinous1 - ++ M Tubular NA

15 - ++ M,A Tubular20 - ++ M, A Tubular, pd NA21 - ++ M Tubular +27 - ++ M,A Tubular +19 - -- Papillary +22 - - N Papillary +26 - - Tubular, solid NA29 - - N Tubular, solid +30 - - M Tubular, solidIndeterminate type4 +++ + M, A Solid, diffuse

*The specimens are not in order of their accession, but for both diffuse and intestinal types havebeen arranged in order of their expression of antigen, first for D10 and then for 17NM.tRecords the presence of tubules in some areas of the otherwise diffuse-type cancers, and the gen-eral histological pattern(s) of tumour growth seen in sections of the intestinal-type cancers. pd,poorly differentiated; solid, growing mainly as solid cords of tumour cells.tIntestinal metaplasia in the non-neoplastic mucosa as demonstrated by staining of goblet cellswith antibody 17NM-20-20. +, metaplasia present; -, metaplasia not found; NA, insufficientmucosa available for evaluation.§A subjective estimate of the proportion of cancer cells staining for antigen made by comparisonwith the numbers ofcancer cells seen in an adjacent section stained with antibody 7HGM-1D7-B.-, no cells stained; +, an occasional cell stained; ++, numerous cells stained; +++, most cellsstained.¶The predominant histochemical staining of intra- and extracellular mucins by the AB-dPASstain, seen generally or in different areas of the cancers. -, no staining; A, acid mucin; N, neutralmucin; M, a mixture of acid and neutral mucins.NA, "Swiss rolls" of mucosa not available.

hepatocytes, and the neuroendocrine cells ingastric mucosa. It did not stain the stratifiedepithelium of skin or oesophagus. Therefore,the tissue distribution of the antigen identifiedresembles that of cytokeratins 8 and 18.12 Theantibody stained all cancer cells in all of thegastric adenocarcinomas tested, and was par-ticularly useful in evaluating sections of thediffuse-type carcinomas.

TUMOUR CLASSIFICATIONFifteen of the cancers were classified as diffuse-type (table 1). These were from 10 men (aver-age age 54 years, median 52) and five women(average age 54 years, median 55), with cancersites in cardia (one), corpus (four), and antrum(10). Of the remaining 17 tumours, 16 were ofthe intestinal type and one was indeterminate;they were from 12 men (average age 61 years,median 60) and five women (average age 73years, median 74), with cancer sites in the car-dia (six), corpus (three), and antrum (seven),and one unspecified site.

All 15 tumours of diffuse type consisted pre-dominantly of small, pleomorphic cancer cells

widely disseminated throughout the tissue assingle cells, or in loose, irregular groups. Threeof the diffuse cancers (specimens 5, 10, and 12in table 1) showed occasional tubules in someareas, which in each case constituted less than- 10% of the sectioned part of the tumour.These tubules lacked the more ordered colum-nar cell arrangement of those seen in theintestinal-type cancers.

All but one of the 16 tumours classified asintestinal-type showed well formed glands/tubules or a papillary pattern of growththroughout the greater part of the tumour. Themucinous cancer (specimen 3) consisted pre-dominantly of lakes of extracellular mucusenclosing groups of cells, but with some moresuperficial parts of the tumour consisting oftubules, the appearance suggesting an origin inan intestinal-type rather than a diffuse-typecancer. I

One tumour (specimen 4) was anindeterminate-type carcinoma consisting pre-dominantly of solid cords of cells with one areaof papillary and tubular carcinoma, as well asone intramucosal and another submucosal areaof a diffuse growth pattern that together repre-sented less than - 10% of the tumour tissue inthe sections examined.

ANTIGEN EXPRESSION IN CANCERS

Antigen DI 0 was present in gastric cancer cellsas cytoplasmic granules. Granular and diffusestaining for D10 was also seen in the stroma ofmany areas containing large numbers of DI0positive cancer cells in diffuse-type cancers, butpools of mucin were not present in thesetumours. Antigen 17NM was present in themucigen granules or mucous vacuoles ofcancer cells, in extracellular mucus surround-ing tumour cells, and in pools of secretedmucus.A subjective estimate of the proportions of

cancer cells staining for D10 or 1 7NM inrepresentative areas of sections of the 32tumours was made using adjacent sectionsstained with antibody 7HGM-1D7-B as ameasure of total numbers of cancer cellspresent. Results are shown in table 1.

Diffuse-type gastric cancersEleven of the 15 diffuse-type cancers showednumerous to most cancer cells staining stronglyfor antigen D10 (fig 1 B). When not completelyobscured by D10 positive cytoplasmic gran-ules, these tumour cells generally showed somedegree of cytoplasmic staining with the Alcianblue counter stain and slight to strong stainingfor predominantly neutral mucins with theAB-dPAS stain. Sections from three of these 1 1diffuse-type tumours (specimens 12, 8, and 17)also showed some areas with occasional tonumerous, large signet ring cells with granularcytoplasm that did not stain for D10 but themajority of which stained strongly for antigen17NM. These large signet ring cells stainedpredominantly for acid mucins with theAB-dPAS stain, in contrast with the neutralmucins of the D10 positive tumour cells seenthroughout the bulk of these three cancers (figs1C-G). Foci of cancer cells expressing the

744



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/


intestinal goblet cell phenotype were, therefore,present in these three otherwise uniformlyD 10positive, advanced diffuse-type cancers.The remaining four diffuse-type gastric can-

cers were signet ring cell carcinomas from onemale and three female patients (specimens 2, 5,16, and 24). Sections of these four tumourswere characterised by discrete areas showingintense staining for antigen 17NM (fig 2A);these areas corresponded to pools of extracel-lular mucin staining predominantly for acidmucins with the AB-dPAS stain. 17NM inthese areas was also mostly extracellular. Thestroma lying outside of the pools ofmucin con-tained few tumour cells in two of thesetumours; one contained numerous cancer cells(including signet ring cells) many of which didnot stain for 17NM (as shown in figs 2A and Bfor specimen 2); and the fourth tumour showedinnumerable, small, irregularly shaped and dif-fusely scattered cells all of which appeared tostain for 17NM in their cytoplasm. Tumourcells in these areas also stained predominantlyfor acid mucins with the AB-dPAS stain (figs2A-F). Two of the four tumours (specimens 2and 5) showed occasional small groups of cells(< 40 cells) staining for D10 in areas outsidethe pools of mucin and in which tumour cellsdid not stain for 17NM. Therefore, foci of can-cer cells expressing the DI0 phenotype werepresent in these two tumours which otherwiseuniformly expressed the intestinal goblet cellphenotype.

Intestinal-type gastric cancersStaining for antigens D10 and 17NM was notas consistent in the tumour cells of theintestinal-type cancers as in the diffuse-typeand, when present, was generally only focal andseen in occasional, scattered cells (figs 1H, 2Gand H). Ofthe 16 intestinal-type cancers, threestained exclusively for D10, two stained forboth D10 and 17NM, and another six stainedexclusively for 17NM. D1 0 positive cellsstained predominantly for neutral mucins, and17NM positive cells for acid mucins. Theremaining five cancers did not stain for eitherDI0 or 17NM, and showed little or no mucinwith the AB-dPAS stain. When present in thesame tumour, the two antigens were generallysegregated in different regions of the tumours.

Indeterminate-type gastric cancersIn specimen 4, the solid cords of cancer cellsand the glandular areas showed the greatmajority of cells staining for D10, but therewere also rare, scattered cells staining for17NM. The small areas of diffuse-type cancerpresent in this specimen (see above) did notstain for either 17NM or D10 antigen, butstained for acid mucins with the AB-dPASstain.

Colonic signet ring cell carcinomas andadenocarcinomasNone of the four signet ring cell cancers of thecolon or the five adenocarcinomas of the colonstained for D1 0, but all of these tumoursexcept for two of the adenocarcinomas stainedfor 17NM; the two adenocarcinomas that did

not stain for 17NM did not stain for mucinswith the AB-dPAS stain.

DiscussionAntigen D10 is characteristic of certain cells oforgans derived from the foregut that typicallysynthesise neutral mucins.6 It is invariablyexpressed in normal gastric mucosa and,therefore, could be expected to occur in somemucus secreting gastric tumours. On the otherhand, antigen 17NM is specific to intestinalgoblet cells and only appears in the stomach inintestinal metaplasia. Therefore, these twoantigens represent universally expressed, mu-tually exclusive and distinctive tissue specificcell markers of potential use in tracing the celllineages of gastric and intestinal adenocarcino-mas. In the following discussion, we will referto the D10 phenotype as that of a mucous neckcell, although the similarity in cytology ofmucous neck cells to the mucous cells of thecardiac, pyloric and Brunner's glands is welldocumented'3 '4 and further supported by thepresence of D10 in each of these tissues.6 Cellsstaining for 17NM will be referred to as of anintestinal goblet cell phenotype.

This study has shown that the majority(84%) of gastric carcinomas display one orboth of the cell phenotypes investigated, andthat there is some correlation between the pre-dominant phenotype displayed and the histo-logical classification of the tumours. As bothD10 and 17NM are normally associated withmucin producing cells, the generally lowerprevalence of antigen positive cells in theintestinal-type cancers was probably related totheir generally poorer mucin content as dem-onstrated by histochemical staining for mucins.The four diffuse cancers that showed extensivepools of extracellular mucin, as well as the oneintestinal-type mucinous cancer, all displayedpredominantly the intestinal goblet cell pheno-type, and this was in keeping with the activesecretion of acid mucins normally exhibited bythis cell phenotype. In contrast, the diffusecancers with cells exclusively or predominantlyexpressing D 10 did not show pools of extracel-lular mucin except in those cancers containingsegregated cells of the intestinal goblet cellphenotype (as illustrated in fig 1D). The twophenotypes studied revealed that Lauren'sdiffuse-type cancers are a heterogeneous groupthat can be subdivided into cancers that exclu-sively express either the mucous neck cell orintestinal goblet cell (signet ring cell) pheno-type, or both phenotypes in segregated areas.As either antigen can be expressed in a diffuseor intestinal-type cancer, the phenotypes donot strictly characterise histological tumourtype. However, if the apparent differences inprevalence ofthe two phenotypes in diffuse andintestinal-type cancers are real, the differencesin behaviour of two such disparate cancer cellphenotypes could help to explain the generallyaccepted biological relevance of Lauren'stumour classification.

All of the diffuse-type gastric carcinomasdisplayed predominantly one or other of thetwo cell phenotypes. Generally, the two antigenmarkers and associated mucin profiles dis-

745



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Hughes, Bhathal

played by the malignant cells matched those ofthe two normal cell types (the mucous neck celland the intestinal goblet cell), suggesting thatmany tumour cells in the diffuse-type cancerswere fully differentiated, at least with respect tothese tissue specific components. Because oftheir continued synthesis of mucins, Laurensuggested that the cells of diffuse-type cancers

are rather more biochemically mature thantheir lack of cell cohesion would indicate,' andthe continued expression ofD10 by the major-ity of these cancers supports such aninterpretation. However, there were many can-cer cells that did not stain for D 10 or 17NM insections of most of the diffuse-type cancers,and it is these unstained cells that may be the

. 1",....~I I.,

*

S.,U.

.ps.

A7A~1.

.9..

D

X...*.-

..v

.-W.2

H

..1

Figure 2 Immunohistochemical staining of sections ofhuman stomach for antigens D10 and 1 7NM. (A-F) Serialsections of a 17NM positive diffuse-type gastric carcinoma (specimen 2 in table 1). (A) Low power view showing areascontaining cancer cells and pools of extracellular mucin (lower left and upper right) staining strongly for antigen 1 7NM. Inan adjacent section (not shown), stained with AB-dPAS, these 17NM positive areas stained strongly for acid mucins.Scattered groups of cancer ceUs showing cytoplasmic stainingfor 17NM (arrow) are also present in the generally unstainedstroma between the pools of mucin. (B) The area corresponding to that outlined in A in a section stained with antibody7HGM-ID7-B to show the presence of numerous cancer cells in the stroma, many of which did not stain for 1 7NM insection A. (C) A higher power view of the area marked by the arrow in section A to show a group of the 1 7NM positivecancer cells. (D) The area corresponding to that in C in a section stainedfor antigen D10 to show large signet ring cellssimilar in size to the 17NM positive cells in C stainingfor acid mucins with the Alcian blue counterstain but notfor Dl 0.(E) The area corresponding to that in C and D in a section stained with the 7HGM-1D7-B antibody showing that thereare many other smaller cancer cells present besides the large, Alcian blue positive signet ring cells. (F) The areacorresponding to that in C, D, and E in a section stained with AB-dPAS showing that the large signet ring cells containpredominantly acid mucins. (G and H) Serial sections ofan intestinal-type gastric cancer (located in the right half of themucosa and submucosa) and adjacent mucosa that contains normal pyloric glands and an area of intestinal metaplasia(arrow). (G) Shows the pyloric glands stainingfor D10 while the area of intestinal metaplasia (arrow) and the carcinomaare unstained. (H) Shows the reverse, wvith the normal pyloric glands unstainedfor 1 7NM, but goblet cells in the intestinalmetaplasia stainingfor 17NM andfocal staining of the carcinoma. (Original magnifications:A x25;B x60; C, D, E, Fx250; G, H xl5.)

746

B,I

...e..11,

'11.

K.! 'o.,

.% 40'i.

V4,1

i

-:..w

t.."4.

.:t.

g,

,

Az:A

14S

'f"I S1, -

it, f.%-' .

.- i.I Ito1,

'S. .k.. N 1-- 1 "4



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/


most active in proliferation. In the case of thefour diffuse-type cancers with tumour cellspredominantly or exclusively of a phenotypeforeign to the stomach, one could argue thatthe cells have arisen from the same cell lineageas that giving rise to intestinal metaplasia in thestomach. Intestinal metaplasia has been re-ported to be more frequent and widespread instomachs with intestinal-type cancers than inthose with the diffuse-type,' 4 although therelation remains uncertain.'5 We found intesti-nal metaplasia in almost 50% of the stomachsbearing each type of cancer as demonstrated bystaining for the 17NM antigen in "Swiss rolls"of gastric mucosa from this small series of 32stomachs. These proportions need to beconfirmed in a larger series of cancers.The consistent expression of D10 by the

majority of the diffuse-type cancers suggeststhat this group of tumours could have arisenfrom the cell lineage that normally differenti-ates into mucous neck cells and the glands ofcardia and pylorus. Previous investigators haveidentified dysplasia in the proliferative zone ofnon-metaplastic gastric glands and proposedthis as the site of origin of diffuse-typecancers.'6 Also, histological and histochemicalevidence that the mucous neck cell may giverise to diffuse-type carcinoma has been pre-sented by others,'7 and similarity in lectinstaining and in histochemical and ultrastruc-tural features between normal mucous neckcells and the neoplastic cells demonstrated.'8Our previous observations on cell proliferationin normal human gastric mucosa suggest thatmucous neck cells are mature cells that rarelydivide, although pre-mucous neck cells in theisthmus that show some cytoplasmic granulesstaining for D10 can do so.'9

It should be noted that DI 0 is alsocharacteristic of pyloric gland metaplasiafound in inflammatory conditions of the stom-ach, small bowel, biliary tree, and gall bladder.6This form of metaplasia is said to be derivedfrom an ulcer associated cell lineage (UACL)distributed throughout the gastrointestinaltract, which, in response to mucosal ulceration,buds from the base of glands and then prolifer-ates to form the metaplastic glands.20 AntigenD10 has been demonstrated in the UACL inthe small intestine.6 The UACL has beendescribed as having the differentiation pro-gramme of Brunner's glands but acquiring theproliferative organisation of the gastric gland.2'If the UACL is distinct from the gastric stemcell lineage normally giving rise to mucousneck cells, it is possible that DIO positive,diffuse-type cancers may also arise indirectly,that is by transformation of cells of the UACLas they proliferate in response to the chronicgastritis which frequently precedes gastric car-cinoma. The UACL does not stain for 17NMand is not known to be directly associated withintestinal metaplasia or intestinal goblet cells,and is unlikely therefore to give rise to gastriccancers other than those displaying antigenD1O.The small group of four diffuse-type cancers

that were distinguished by predominantly orexclusively displaying the 17NM intestinal

goblet cell phenotype and secreting acidmucins histologically were classic signet ringcell carcinomas.9 The cell phenotype displayedby these four tumours suggests a relation with,and perhaps origin in, intestinal metaplasia,this being the only non-neoplastic tissue in thestomach to express the 17NM antigen. Intesti-nal metaplasia is reported to arise from withinthe isthmus of the glands of intact gastricmucosa,22 or during the reparative process of

23erosions. In the early stages of the develop-ment of intestinal metaplasia in intact mucosa,the proliferative zone may remain in theisthmus with the surface epithelium and pitsshowing metaplastic cells in continuity with thedeeper normal gastric glands.22 Transformedcells developing at such a site could give rise to17NM positive diffuse-type carcinomas in ananalogous fashion to the D10 positive diffusecancers that may develop from pre-mucousneck cells, as suggested above. Since there wasevidence of tubule formation in three of the 15diffuse-type carcinomas, it is also possible thatsome of the tumours studied were originallyintestinal-type cancers that were then over-grown by a diffuse-type tumour, although therewas no histological evidence to support this.Both the D10 and 17NM cell phenotypes

were co-expressed in individual tumours of aquarter of all the gastric cancers examined.Since no more than 10% of subjects with gas-tric carcinoma are reported to show macro-scopically discrete synchronous cancers,24 colli-sion tumours are an unlikely explanation of themixed cell phenotypes found in some of ourspecimens. A stratification of tumour cellswhich parodies normal gastric cell types hasbeen observed in the early stages of develop-ment of gastric carcinoma by otherinvestigators,25 and segregation of the two cellphenotypes was noted in the mucosa ofsome ofthe diffuse-type cancers studied here. There-fore, it appears that some transformed gastriccells are sufficiently plastic and primitive to becapable of divergent differentiation into eitherthe DI0 or 17NM phenotype. Clonal selectioncould account for the predominance with timeof one or other phenotype or, when there is noselective advantage, the growth of both cellphenotypes would result in the regional differ-ences in phenotype expression that we ob-served within some individual tumours.

Since normal Brunner's glands form bybudding off from the crypts of the duodenalmucosa,26 the stem cells in the crypts of theduodenum must be capable of divergent differ-entiation into either mucosa containing intesti-nal goblet cells (displaying antigen 17NM) orBrunner's glands (displaying antigen Dl 0). Itappears that in the course of development ofdiffuse-type cancers the transformed gastricstem cells consistently regain such plasticityand go on to express one or both of the twophenotypes studied. Why the mucous neck cellphenotype rather than that of the intestinalgoblet cell is then more frequently expressed bysuch cancers may be explained on the basis thatthis would be the more normal direction in dif-ferentiation for gastric cells (differentiationinto mucous neck cells or mucous cells of the

747



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Hughes, Bhathal

cardiac and pyloric glands). The co-existenceof both phenotypes found in some individualcancers is also in keeping with a capacity fordivergent differentiation by the neoplastic cells,and helps to explain the confusing histochemi-cal heterogeneity of these tumours.This study has shown that gastric adenocar-

cinomas are characterised by two cell pheno-types, one of which is foreign to the normalstomach mucosa and both of which arenormally present in the duodenum. Reversionof the transformed gastric stem cell to a moreprimitive gut stem cell lineage possessing thesame capacity for divergent differentiation asthe normal duodenal stem cell is the most eco-nomical explanation of the consistent associ-ation of the two phenotypes with the non-cohesive diffuse-type carcinoma. Thesecancers typically show one of the two antigenspredominating, usually DI0, and beingstrongly expressed by the majority ofthe cancercells. In contrast, intestinal-type cancers withtheir more ordered tubular structure more fre-quently display the intestinal goblet cellphenotype and typically express the antigens inrelatively few tumour cells.

The work was supported by a grant from the National Healthand Medical Research Council of Australia. This work was pre-sented in part at the Australian Gastroenterology Week 1996,the Annual Scientific Meeting of the Gastroenterological Soci-ety of Australia, held in Adelaide, September 1996, andpublished in abstract form in J Gastroenterol Hepatol 1996;11(Suppl):A94.

1 Lauren P. The two histological main types of gastriccarcinoma: diffuse and so-called intestinal-type carcinoma.Acta Path Microbiol Scand 1965;64:31-49.

2 Correa P The epidemiology of gastric cancer. World J Surg1991;15:228-34.

3 Janssen CW Jr, Lie RT, Maartmann-Moe H, Matre R. Theinfluence of age on the growth and spread of gastric carci-noma. BrJ Cancer 1991;63:623-5.

4 Ming SC, Goldman H, Freiman DG. Intestinal metaplasiaand histogenesis of carcinoma in human stomach. Cancer1967;20:1418-29.

5 Stefanini M, De Martino C, Zamboni L. Fixation of ejacu-lated spermatozoa for electron microscopy. Nature 1967;216:173-4.

6 Hughes NR, Bhathal PS, Francis DMA. Phenotypic identityof gastric mucous neck cells and mucous cells of cardiac,pyloric, and Brunner's glands. J Clin Pathol 1994;47:53-7.

7 Hughes NR, Walls RS, Newland RC, Payne JE. Antigenexpression in normal and neoplastic colonic mucosa: threetissue-specific antigens using monoclonal antibodies to iso-lated colonic glands. Cancer Res 1986;46:2164-71.

8 Cattoretti G, Pileri S, Parravicini C, Becker MHG, Poggi S,Bifulco C, et al. Antigen unmasking on formalin-fixed,paraffin-embedded tissue sections. J Pathol 1993;171:83-98.

9 "Watanabe H, Jass JR, Sobin LH, in collaboration withpathologists in eight countries. Histological typing ofoesophageal and gastric tumours. In: WHO internationalhistological classification of tumours. Berlin: Springer-Verlag,1990:1-43.

10 Cook HC. Carbohydrates. In: Bancroft JD, Stevens A, eds.The theory and practice of histological techniques. Edinburgh:Churchill Livingstone, 1982:180-216.

11 Hughes NR. Gland heterogeneity and colorectal neoplasia.University of Sydney: PhD thesis, 1988.

12 Moll R, Franke WW, Schiller DL, Geiger B, Krepler R. Thecatalog of human cytokeratins: patterns of expression innormal epithelia, tumors and cultured cells. Cell 1982;31:11-24.

13 Bloom W, Fawcett DW. A textbook of histology. Philadelphia:WB Saunders Company, 1975:639-57.

14 Katsuyama T, Spicer SS. Histochemical differentiation ofcomplex carbohydrates with variants of the concanavalinA-horseradish peroxidase method. Jf Histochem Cytochem1978;26:233-50.

15 Dixon MF. Progress in gastric cancer. In: Kirkham N, HallPA, eds. Progress in pathology. Vol 1. Edinburgh: ChurchillLivingstone, 1995:13-29.

16 Ghandur-Mnaymneh L, Paz J, Roldan E, Cassady J.Dysplasia of nonmetaplastic gastric mucosa. A proposal forits classification and its possible relationship to diffuse-typegastric carcinoma. Am J Surg Pathol 1988; 12:96-114.

17 Yamashina M. A variant of early gastric carcinoma.Histologic and histochemical studies of early signet ringcell carcinomas discovered beneath preserved surfaceepithelium. Cancer 1986;58: 1333-9.

18 Tatematsu M, Furihata C, Katsuyama T, Miki K, Honda H,Konishi Y, et al. Gastric and intestinal phenotypicexpressions of human signet ring cell carcinomas revealedby their biochemistry, mucin histochemistry, and ul-trastructure. Cancer Res 1986;46: 4866-72.

19 Hughes NR, Bhathal PS. Do gastric mucous neck cellsdivide? [abstract]. IntJ' Surg Pathol 1996;3:217.

20 Wright NA, Pike C, Elia G. Induction of a novel epidermalgrowth factor-secreting cell lineage by mucosal ulcerationin human gastrointestinal stem cells. Nature 1990;343:82-5.

21 Ahnen D, Poulsom R, Stamp GWH, Elia G, Pike C, JefferyR, et al. The ulceration-associated cell lineage (UACL)reiterates the Brunner's gland differentiation programmebut acquires the proliferative organization of the gastricgland. J Pathol 1994;173:317-26.

22 Hattori T, Fujita S. Tritiated thymidine autoradiographicstudy on histogenesis and spreading of intestinal metaplasiain human stomach. Path Res Pract 1979;164:224-37.

23 Mukawa K, Nakamura T, Nakano G, Nagamachi YHistopathogenesis of intestinal metaplasia: minute lesionsof intestinal metaplasia in ulcerated stomachs. J Clin Pathol1987;40:13-8.

24 Kosaka T, Miwa K, Yonemura Y, Urade M, Ishida T, Take-gawa S, et al. A clinicopathologic study on multiple gastriccancers with special reference to distal gastrectomy. Cancer1990;65:2602-5.

25 Fujimori Y, Akamatsu T, Ota H, Katsuyama T. Proliferativemarkers in gastric carcinoma and organoid differentiation.Hum Pathol 1995;26:725-34.

26 Krause WJ, Leeson CR. The origin, development anddifferentiation of Brunner's glands in the rat. J Anat 1967;101:309-20.

748



j.com/

J Clin P


ber 1997. Dow

nloaded from

http://jcp.bmj.com/

Documents

Gastric neckcell and intestinal goblet cell in gastric ... · Hughes,Bhathal and sigmoid colon (four cases) obtained from routinelyprocessedtissue werealso studied. Two micron thick