[CANCER RESEARCH 53. 1421-1430, March 15. 1993]

Expression of Thrombospondin (TSP1) and Its Receptors (CD36 and CD51) inNormal, Hyperplastic, and Neoplastic Human Breast1

Philippe Clezardin,2 Lucien Frappart, Magali Clerget, Christine Pechoux, and Pierre D. Delmas

INSERM V234. Pavillon F ¡P.C, M. C. P. D. D.¡and Department of Pathology IL F.. C. P.I. Edouard Herrin! Hospital. Place d'Arsonval. 6V4.Ì7Lyon, France

ABSTRACT

We have previously shown that thrombospondin (TSP) is present innormal breast secretions, and high levels of TSP are observed in malignantbreast secretions and cytosols. Three genes encoding for three distinctTSPs (TSP1, TSP2, TSP3) have recently been described. In this study,using both immunohistochemistry and in situ hybridization, we report onthe distribution of TSP1 in normal, hyperplastic, and neoplastic humanbreast. Its immunolocalization was also compared with that of two knowncell surface receptors for TSP1: CD36 and CD51. In nonlactating ducts ofnormal and hyperplastic breast, TSP1 and CD51 are expressed in thebasement membrane and in the basal surface of myoepithelial cells, respectively. In lactating adenomas, both TSP1 and CD51 disappear fromthe myoepithelial-stromal junction of ducts. However, TSP1 becomes selectively expressed at the apices of secretory epithelial cells of lactatingducts together with CD36, suggesting that the distribution of TSP1and theappearance of its receptors are dependent on the secretory activity ofhuman mammary ducts. In neoplastic human breast, a strong immuno-staining for TSP1 is observed in the basement membrane surrounding insitu carcinomas (preinvasive cancer), and excessive TSP1 deposits are alsoobserved in desmoplasia of invasive ductal carcinomas. TSP1 mRNA islocalized in myoepithelial cells surrounding in situ carcinomas and infibroblasts present in desmoplastic areas. On the other hand, few invasiveductal carcinoma cells (10%) express TSP1, while CD51 is moderatelyexpressed by some neoplastic clusters, and no immunoreactivity is observed for CD36. By contrast, TSP1 is codistributed with CD51 in most ofthe invasive lobular carcinoma cells (40 to 80%) and with CD36 in asubpopulation (30 to 40%) of these invasive tumor cells. As previouslyobserved with lactating adenomas, it is likely that the coexpression ofTSP1 and CD36 is related to the secretory activity of invasive lobularcarcinoma cells. The different distribution of TSP1 in invasive ductalversus lobular carcinomas may well reflect biological differences betweenthese two main types of breast carcinoma. In this regard, the coexpressionof TSP1 and CD36 may, in part at least, account for the variably invasivebehavior of lobular carcinoma cells.

INTRODUCTION

TSP1 is a large, trimeric glycoprotein synthesized and secreted by

a wide range of normal and transformed cells, and it is incorporatedinto the extracellular matrix (1-3). TSP binds specifically to a variety

of molecules including heparin, fibronectin, vitronectin, collagen (forreview, see Ref. 2), and osteonectin (4). Because of its extremely widedistribution, TSP has been involved in a variety of physiopathologicalcontexts such as development (5), wound healing (6), atherosclerosis(7), angiogenesis (8), tumorigenesis (9), and tumor cell metastasis(10). In this regard, TSP functions in modulating cellular adhesion (3,10-15), migration (16), proliferation, and differentiation of tumor

cells (9).

Received 8/19/92; accepted 1/6/93.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

1This study was supported by grants from La FédérationNationale des Centres de

Lutte contre le Cancer (P. C.), La FédérationNationale des Groupements des EntreprisesFrançaisesdans la Lutte contre le Cancer (P. C.). and ARC (L. F).

2 To whom requests for reprints should be addressed, at INSERM U. 234. Pavillon F.Hôpital Edouard Herriot. 3 Place d'Arsonval. 69437 Lyon Cedex 03, France.

'The abbreviations used are: TSP. thrombospondin; IgG. immunoglobulin G; ADH,

atypical ductal hyperplasia; NOS, not otherwise specified; PBS. phosphate-buffered saline; cDNA. complementary DNA; SSC, standard saline citrate.

It is now known that three related but not identical genes for TSPare expressed in mice (Thbs 1, Thbs 2, and Thbs 3) and humans(THBS\, THBS2, and THBS3) (17-20). TSP1 and TSP2 are made upof several protease-resistant domains, including NH2-terminal,COOH-terminal, and procollagen homology domains, and type I, typeII, and type III repeats (2, 17-20). The three type I repeats share

homology with the human von Willebrand factor, circumsporozoiteproteins from Plasmodium falciparum, and the complement proteinsproperdin and factors C6 through C9 (2, 20). The three type II repeatsshow homology with epidermal growth factor (2, 20). The type IIrepeats are followed by seven type III repeats which constitute theprimary calcium-binding domain of TSPs (2, 20). As opposed toTSPl

and TSP2, TSP3 lacks the procollagen homology domain and type Irepeats and contains four rather than three type II repeats (19). Basedon comparison of cDNA sequences, the NH2 terminus containing theheparin-binding domain is poorly conserved between TSP1 versus

TSP2 and TSP1 versus TSP3 (38 and 10% identity, respectively) (17,19, 20). These findings suggest that TSP2 and TSP3 may not beheparin-binding proteins or may bind less well to heparin comparedwith TSP1 (17, 20). In addition to the heparin-binding domain which

binds heparin sulfate proteoglycans (11) and sulfatides (12), TSP I andTSP2 contain binding sites for three other cellular receptors: theArg-Gly-Asp sequence which binds to CD51/CD61 (the avß3inte-grin) (13), the Val-Thr-Cys-Gly sequence which binds to CD36 (alsonamed platelet GPIIIb) (21, 22), and the COOH-terminal domain

which binds a M, 105,000/80,000 receptor (14). Not all of theseinteractions occur in all cells, and it is possible that separate cellsurface receptors are used in a cooperative manner to bind TSP1 andTSP2 (15). Moreover, TSP3 lacks the Val-Thr-Cys-Gly sequencewhich is present in type I repeats and the Arg-Gly-Asp sequence

which is present in the last type III repeats of TSP1 and TSP2 (19).The existence of at least three structurally and genetically distinctTSPs suggests therefore that any given TSP must have a specificfunction on some domains of TSP that are lacking in others. It is alsopossible that TSP1, TSP2, and TSP3 are regulated in a cell type-

specific fashion (20).We previously reported that TSP is present in normal breast secre

tions (23), and high levels of TSP are observed in malignant breastsecretions and cytosols (23, 24). A recent immunohistochemical studyhas demonstrated that excessive TSP deposits are present in the desmoplastic stroma of human breast carcinomas, whereas normal breasttissue exhibits no TSP staining (25). However, the source(s) and thetype(s) of TSP expressed in normal and neoplastic breast tissues arenot known nor is there any indication of the function of TSP in thiscontext. CD36 has been related to a mammary epithelial cell surfaceprotein present in breast secretions, namely PAS-IV (26, 27). BothCD36 and PAS-IV interact with TSP (21, 22, 28). However, thefunction of PAS-IV in normal and malignant breast tissues is notknown. The distribution of some adhesive proteins (fibronectin, lami-

nin, tenascin) and integrins (including CD51) has also been describedin normal breast tissue and reported to be altered in human breastcarcinomas (29-31). In the present study, using immunohistochemis

try and in situ hybridization, we have investigated the expression ofTSP1 and the distribution of two of its cell surface receptors (CD36,CD51) in normal, hyperplastic and neoplastic human breast.

1421

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BREAST CARCINOMAS

Fig. 1. Tissue sections of normal and neoplastia breast immunostained with a negative control mouse monoclonal antibody directed against rat K light chains. No staining wasobserved. A and B, normal ducts observed at original magnifications (X 160 and X 500), respectively; C and D, in situ carcinoma observed at original magnifications (X 160 and x500), respectively; E and F, invasive carcinoma observed at original magnifications (X 160 and X 500). respectively.

MATERIALS AND METHODS

Antibodies. Mouse monoclonal antibodies PIO and PI2 are directedagainst human blood platelet TSPI. The characterization and specificity ofthese antibodies have been described earlier (32). Mouse anti-TSP monoclonalantibodies MA-I and MA-II (33) were kindly provided by Dr. Jack Lawler ofBrigham and Women's Hospital, Boston, MA. The epitope for MA-I is in the

sequence DFDHDSUP (residues 915-922) of the type III repeats and that forMA-II is IEDANLIP (residues 43-50) in the heparin-binding domain of TSPI.

Comparison of amino acid sequences of TSPI. TSP2. and TSP3 indicates thatMA-II is specific for TSPI. Mouse monoclonal antibodies FA6-152 and

OKM5. which are directed against CD36 (34. 35), were purchased fromImmunotech (France) and Ortho Diagnostics Systems, respectively. Mousemonoclonal antibody LM142 specific for CD5I (36), the integrin subunit o^,was a gift of Dr. David Cheresh, Scripps Clinic and Research Foundation, LaJolla. CA. Mouse monoclonal antibody MARK I directed against rat K lightchains was purchased from Immunotech (France). Rabbit anti-mouse IgG wasfrom Dakopatt (Copenhagen. Denmark). Goat anti-rabbit IgG conjugated with

horseradish peroxidase was from Clonatech (Paris, France).Immunohistochemistry. Breast tissue was obtained from tumorectomies

and systematized mammary exérèses.Benign lesions included adenomas (n =2), fibroadenoma (n = I), and cystic disease (n = 1). Borderline lesionsincluded ADH (n = 1) and cystosarcoma phyllode (n = 1). Malignant tumors

Table 1 Normal breast and benign and borderline lesions

NormalbreastCystic

disease6Fibroadenoma

''Lactating

adenoma'1ADH'Cystosarcoma

phyllode'"

A moderate toTSPI+

in the basementmembrane ofducts++

in the basementmembrane ofducts++

in the basementmembrane ofducts++

at the apicesofluminalepithelial

cells+++ in the

basementmembrane ofducts++

in the basementmembrane ofductsstrong

staining (+/++) w<CD36"---++

at the apicesofluminalepithelial

cells++ in fibroblasts

surroundingducts—CDS!+

in myoepithelialcells, +/- inluminalcells+

¡nmyoepithelialcells, +/- inluminalcells+

in myoepithelialcells, +/- inluminalcells-+

in myoepithelialcells, +/- in luminal

cells—is

always observed in endothelial cells and inmembranes of adipocytes.

h Benign lesions.' Borderline lesions.

1422

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CD5I IN BREAST CARCINOMAS

Fig. 2. A. normal duct immunostained forTSP; staining is found as a broad pale line in the basement membrane of ducts (arrow), while the surrounding stroma is negative. Originalmagnification. X 4(M).R. normal duct immunostained for CD51 (the o\ integrin suhunit); staining is mostly localized in the basal surfaces of myoepithelial cells (arrow), whereas thereactivity is weaker in luminal epithelial cells of ducts (arrowhead). Original magnification. X 160. C. normal duct immunostained for CD36; no reactivity is observed in the duct.However, CD36 is expressed in endothelia! cells (arrow) and adipocytes (arrowhead). Original magnification, X 4(X). I), fibroadenoma immunostained forTSP; staining is presentas a broad line in the basement membrane immediately adjacent to the basal surfaces of myoepithelial cells {arrow}. Original magnification, X 500. E. fibroadenoma immunostainedfor CD5I; staining was moderately expressed in myoepithelial cells U//rmr). Original magnification. X 5(X). f-\ fibroadenoma immunostained tor CD36; reactivity is only observed

in endothelial cells (arrow) and adipocytes (arrowhead). Original magnification, X 500. G, lactating adenoma immunostained for TSF; TSP becomes strongly expressed at the apicalsurface of luminal epithelial cells (arrow), while the basement membrane is completely negative (arrowhead). Original magnification, X 500. //, laclating adenoma immunostainedfor CD51 ; no reactivity is observed. Original magnification, x 500. /, lactating adenoma immunostained for CD36; strong staining is observed at the apices of luminal epithelial cells(arrow). Original magnification, X 500.

comprised invasive ductal carcinomas NOS in = 12), invasive comedocarci-

nomas (invasive ductal carcinoma with dominant in xitii comedocarcinomas)(n = 2), mucinous carcinomas (n = 3), medullary carcinomas (n = 2),invasive ductal carcinoma with squamous metaplasia (;i = I), and invasivelobular carcinomas (n = 5). Invasive lobular carcinomas, invasive ductal

carcinomas NOS. and mucinous carcinomas were obtained from systematizedmammary excision specimens or tumorectomies in postmenopausal women(aged 49 to 82 yr). Tissues were snap fro/en in liquid nitrogen-cooled isopen-

tane and subsequently stored in liquid nitrogen until used.Fro/en sections (5 to 8 urn thick) were cut in a cryostat at -20°C, air dried

2 h at room temperature, and fixed in cold acetone for 10 min. Immunostainingwas carried out as follows. Tissue sections were incubated with the primaryantibody previously diluted at a concentration of O.I ug/ml in PBS. pH 7.2.containing 1% (mass/vol.) bovine serum albumin. At this stage, a negativecontrol was included using monoclonal antibody MARK 1 (0.1 ug/ml) whichis directed against rat K light chains. After 30-min incubation, tissue sections

were washed 3 times with PBS (2 min/wash) and incubated for 30 min with arabbit anti-mouse IgG [diluted to 1:80 in PBS containing 25% (v/v) human

serum AB]. After washing 3 times with PBS, slides were incubated for 30 minwith a goat anti-rabbit IgG conjugated with horseradish peroxidase [diluted to

1:80 in PBS containing 25% (v/v) human serum AB]. The washing procedurewas repeated, and the reaction was revealed with a solution of 10 mg of3,3'-diaminoben/idine tetrahydrochloride (Sigma) in 10 ml of 0.05 M Tris

buffer. pH 7.6. containing 10 ul of 30% H2O;. The reaction was stopped after8 min by washing in tap water. Sections were then briefly counterstained withMayer's hematoxylin, dehydrated in a graded alcohol series, cleared in methyl

cyclohexane, and mounted. Incubation with the different antibodies was always performed at room temperature in a humidified chamber.

Sections were considered as positive or negative according to the presenceor absence of specific staining when compared to the staining obtained withcontrol antibody MARK 1. The intensity of the staining was scored arbitrarilyas follows: very strong ( + + + ); strong ( + + ); weak ( + ); equivocal ( +/-): andnegative (-). Sections were evaluated on two different occasions by three of us

(P. C.. L. F., M. C.). In case of disagreement between examiners, slides werereviewed and a consensus opinion obtained.

In Situ Hybridization. Human TSP cDNA clone M9 was kindly providedby Dr. J. Lawler of Brigham and Women Hospital. Boston. MA. M9 codes forthe NH: terminus of TSPI (A(-15) to W (355)] (37). The cDNA was insertedinto the Eci>R\ site of pGEM-2 (Promega Biotec) and used to derived antisenseand sense probes. RNA probes were generated using 100 uCi of [a-'5S]UTP

1423

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BREAST CARCINOMAS

Fig. 3. Ductal carcinoma NOS immunostained for TSR In A. a striking increase instromal TSPis observed in the desmoplastic region of the tumor (arrow], while in B, distalstromal regions are negative (arrowhead). Original magnifications (or A and ßare X 160and X 63, respectively.

and either SP6 or T7 polymerases. The pGEM-2/TSP construct was linearized

with ßiimHland transcribed with SP6 RNA polymerase (Boehringer Mannheim. France) to obtain an antisense probe. Alternatively, the pGEM-2/TSP

construct was linearized with Spii I and transcribed with T7 RNA polymeraseto obtain a sense probe.

Breast tumors were fixed overnight at 4°Cin PBS, pH 7.2, containing 4c/c

paratbrmaldehyde. dehydrated by graded alcohols, and finally embedded in

paraffin. Tissue sections (5 urn) were cut. adhered to glass slides previouslytreated with 3-aminopropyltriethoxysilane (Sigma), and stored at 4°Cuntil

used. For in situ hybridization, sections were dewaxed and treated with pro-teinase K (5 ug/ml) for 30 min at 37°C.After proteinase K treatment, sections

were washed twice in PBS (2 min/wash) and dehydrated by graded concentrations of ethanol. Dried sections were covered with 10 to 20 ul of hybridization buffer containing the RNA probe (1.6 ug/ml; 1 x 1()6cpm/20 ul) in

4-fold SSC (0.06 M sodium citrate:0.6 MNaCl) buffer. 50% deionized forma-mide, 1% Denhardt's solution (0.05% polyvinylpyrrolidone 40:0.05% Ficoll

400:0.02% bovine serum albumin). 10% dextran sulfate. 100 ug/ml of yeasttRNA, 100 ug/ml of salmon sperm, and 0.1 M dithiolhreitol. After overnightincubation al 37°Cin a humid chamber, slides were washed once at 37°Cin

2-fold SSC buffer containing 50% formamide. once in SCC buffer containing10 ug/ml of RNase, once in SCC buffer, and once in 0.5-fold SCC buffer (30

min/wash). After washing, sections were dehydrated in graded ethanol andimmersed at 42°Cin K5 emulsion (Ilford) diluted 1:4 with distilled water. Afterimmersion, slides were exposed at 4°Cfor 15 to 30 days in a light-tight box.

Sections were incubated in developing solution (DI9. Kodak) for 3 min at18°Cand counterstained with toluidine blue. Epifluorescent UV-light micros

copy was performed with a Dialux 22 EB microscope.

RESULTS

Immunohistochemistrv

Tissue sections of normal, hyperplastic, and neoplastia breast wereexamined by immunohistochemistry using a negative control mousemonoclonal antibody (MARK-1) directed against rat K light chains,

which is isotype matched with the other antibodies used in this study.No immunostaining was observed with MARK-1 (Fig. 1).

Anti-TSP mouse monoclonal antibodies PIO, PI2, MA-I, andMA-II showed a similar immunostaining profile for a given tissue

section. However, the intensity of the staining was always weaker withmonoclonal antibodies P12 and MA-II. In addition, the immunostaining with PI2 and MA-II was always equivocal (+/-) or negative (-)on normal breast tissue. Anti-CD36 mouse monoclonal antibodiesFA6-152 and OKM5 gave a similar immunoreactivity for a given

tissue section.

Normal Breast. Benign Lesions, and Borderline Lesions

The expression patterns of TSP1, CD36, and CD51 in normalbreast, benign lesions (cyst disease, fibroadenoma, adenoma), and

Table 2 Breast carcinomas

In situ ductalandlobularcarcinomaInvasive

ductalcarcinomaNOSInvasive

lobularcarcinomaIn

situ comedo-carcinoma''Invasive

comedocarcinoma''Mucinous

carcinomaMedullary

carcinomaInvasive

ductal carcinnrna withsquamous metaplasiaTSP1+++

in basementmembrane;-in tumorcells+

in 10% oftumorcells+/++

in 40 to 80%oftumorcells++

in basementmembrane;

- in tumor cells+/- in tumorcells+/-

in tumorcells+

al periphery ofclusters+

at periphery ofclustersCD36"__+/++

in 30 to40%oftumorcells-

in tumor cells;

+++ in fibroblastssurroundingducts--+

in some tumor cells at periphery ofclustersCD5I+/-

in tumor cells, enhancedatperiphery+

in tumorcells.enhancedat periphery of

clusters+/++

in 40 to 80%oftumorcells+

in tumor cells enhanced atperiphery+

at periphery ofclusters+

at periphery ofclusters+

in tumor cells, enhanced atperiphery of

clusters+

in tumor cells, enhanced atperiphery of

clusters" A moderate to strong staining (+/++) was always observed in endothelial cells and in membranes of adipocytes.h Ductal carcinoma with dominant intraductal comedocarcinoma.

1424

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BREAST CARCINOMAS

Fig. 4. A and B, in situ ductal carcinoma NOS immunostained for TSF; a strong immunostaining is observed in the basement membrane surrounding ducts containing in situcarcinoma (arrow}, while no staining is noted in tumor cells (arrowhead}. Nonspecific staining is also observed in necrotic cells present in the middle of the duct (asterisk}. Originalmagnifications. X 160 and x 500, respectively. C and D. invasive ductal carcinoma NOS immunostained for TSP; few invasive cells are immunoreactive (urrow}. Originalmagnifications, X 160 and X 400. respectively. £and F. in situ ductal carcinoma NOS immunostained for CD51; staining was weak in the malignant population (arrowhead) andslightly enhanced at the peripheral layer (arrow}. Original magnifications. X 400 and X 500, respectively. G and H. invasive duclal carcinoma NOS immunostained for CD5I;immunoreactivity was moderately expressed by neoplastic cells, the staining being more prominent at the cell's periphery (arrow). Original magnifications. X 160 and X SIX),

respectively. / and J, in situ ductal carcinoma NOS immunostained for CD36; immunoreactivity is only noted in endothelial cells (arrow} and adipocytes {arrowhead}. Originalmagnifications, X 160 and x 500. respectively. K and L, invasive ductal carcinoma NOS immunostained for CD36; the immunostaining pattern is similar to that observed for in \itiiduclal carcinomas. Original magnifications, x 160 and x 500, respectively.

borderline lesions (atypical ductal hyperplasia, cystosarcoma phyl-

lode) are summarized in Table 1.Normal Breast Tissue. TSP1 was weakly but consistently found

as a broad pale line in the basement membrane of normal ductules(Fig. 2A). No evidence of TSP1 was seen in the stroma. Immunostaining for CD51 (the o\ subunit integrin) was consistently localizedin the basal surfaces of myoepithelial cells, whereas the reactivity wasweaker in luminal epithelial cells of normal ducts (Fig. 2B). No

reactivity in either myoepithelial or luminal epithelial cells was observed in any of our samples when using two different antibodiesagainst CD36 (Fig. 2C). By contrast. CD36 was consistently expressed with variable intensity in endothelial cells and in membranesof adipocytes (Fig. 2C).

Benign Lesions (Fibroadenoma, Cystic Disease, Adenoma). Immunostaining forTSPl in fibroadenoma showed a strong immunoreactivity in the basement membrane immediately adjacent to the basal

1425

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BREAST CARCINOMAS

\ir•¿�*i

' " ' jjl•¿�f

< /

¿.f

Fig. 5. In situ comedocarcinoma immunostained for CD36. CD36 is strongly expressed in the demarcating layer of fibroblasts surrounding in situ comedocarcinomu (arrow), whilethe distribution patterns of TSP and CD51 are similar to those noted in other in situ carcinomas (see Fig. 4). A, original magnification. X 160: B, original magnification, x MX).

surface of myoepithelial cells (Fig. 2D). CD51 was moderately ex- in endothelial cells and adipocytes (Fig. 2F). Asimilar immunoprofile

pressed in both myoepithelial and epithelial cells of ducts associated pattern for TSP1, CD51. and CD36 was retained in cystic diseasewith fibroadenoma (Fig. 2E). No immunoreactivity in either myoep- (Table 1). By contrast, in cases of adenomas, TSP1 was strongly

ithelial or luminal epithelial cells was observed with CD36 (Fig. 2F). expressed at the apices of secretory epithelial cells in lactating ducts,However, as observed for normal breast tissue, CD36 was expressed while the basement membrane was completely negative (Fig. 2G). In

Fig. 6. Invasive lobular carcinoma immunostained for TSP, CD51, and CD36, respectively. Invasive cells in "Indian file" strands are strongly stained for TSP. CDS I. and CD36.A. C, and £',original magnifications, X 160; B, D, and F. original magnifications, X 400.

1426

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BR1-AST CARCINOMAS

" tf •¿�••¿�'*

•¿�---,

Fig. 7. Specific localization of TSP1 mRNA in normal and neoplastic human breast by in situ hybridization using an antisense probe. A, bright-field lieft) and dark-field (right)photomicrographs of a section of normal duct demonstrating the silver grains (open arrows) localized over myoepitheliat/epithelial cells. Original magnification, x 250. B, bright-field(left) and dark-field (ri^/jr) photomicrographs of a section of in situ carcinoma demonstrating the silver grains <upen arrows) localized over myoepithelial cells. Original magnification.X 250. C, bright-field (left) and dark-field (right) photomicrographs of a section of invasive ductai carcinoma showing the silver grains (upen arnws) localized over fibroblasls. Nohybridization is observed over lumor cells (black arrttws). Original magnification. X 250. D. bright-field (left) and dark-field (right) photomicrographs tifa section of invasive lobularcarcinoma showing the silver grains (open arrows) localized over tumor cells. Original magnification, x 400.

addition, immunostaining for CD51 was negative (Fig. 2H), and amoderate-to-strong immunoreaction for CD36 was noted at the apices

of luminal epithelial cells (Fig. 21).Borderline Lesions (ADH, Cystosarcoma Phyllode). The

amount of TSPI surrounding ducts associated with ADH and cystosar-

coma phyllode was consistently increased when compared to normalbreast tissue (Table 1). CD51 was moderately expressed in both myoepithelial and epithelial cells of ducts associated with ADH, whereasno staining was observed in cystosarcoma phyllode (Table 1). Inaddition to endothelial cells and adipocytes, a strong immunoreaetiv-

ity for CD36 was observed in the demarcating layer of fibroblastssurrounding hyperplastic ducts associated with ADH (Table 1). On the

other hand, CD36 was only expressed in endothelial cells and inmembranes of adipocytes in cystosarcoma phyllode (Table 1).

Breast Carcinomas

Most of the invasive breast carcinomas studied, predominantly ofthe ductai types, were characterized by a desmoplasia which frequently remained circumscribed. In desmoplasia, a striking increase inTSPI immunoreactivity was observed in the stroma surrounding negatively stained clusters of malignant cells (Fig. 3/Õ).while TSPIimmunoreactivity disappeared along the invasive front of the tumor

1427

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS i IN BREAST CARCINOMAS

(Fig. 3ß).By contrast, invasive lobular carcinomas were rather poorlydemarcated and did not show any convincing stromal immunoreac-

tivity for TSP1.The expression of TSP1, CD36, and CD51 in breast carcinomas

was therefore studied in distal desmoplastic regions where invasive-

ness of normal breast tissue occurs. Results are summarized in Table2.

Invasive Ductal Carcinomas. In the in situ areas of invasive duc-tal carcinomas, strong immunostaining for TSP1 was observed in thebasement membrane surrounding ducts containing in situ carcinoma,while no staining was noted in tumor cells (Fig. 4, A and B). In allcases of invasive ductal carcinomas studied (n = 20), a weak staining

for TSP1 was observed in 10% of invasive cells (Fig. 4, C and D).Immunostaining for CD51 in most in situ areas of invasive ductal

carcinomas was weak in the malignant population and enhanced at theperipheral layer, which could be interpreted as remnants of myoepi-

thelium (Fig. 4, E and F). In invasive ductal carcinomas, CD51 wasmoderately expressed by some neoplastic clusters, the immunostaining being more prominent at the cell's periphery (Fig. 4, G and //).

No reactivity for CD36 was observed in both in situ and invasiveductal carcinomas (Fig. 4, / to L). However, as observed in benignlesions and cystosarcoma phyllode, CD36 immunoreactivity was consistently noted in endothelial cells and in membranes of adipocytes.

Invasive Ductal Carcinomas with Dominant Intraductal Come-

docarcinomas. The distribution patterns of TSP1 and CD51 weresimilar to those noted in others in situ and invasive ductal carcinomas(Table 2). By contrast, CD36 was expressed in the demarcating layerof fibroblasts surrounding in situ comedocarcinomas (Fig. 5), as it hasbeen previously observed for ADH (Table 1). In invasive comedocarcinomas, CD36 imunoreactivity was only observed in endothelialcells and in membrane of adipocytes (Table 2).

Invasive Lobular Carcinomas. In the in situ areas, the distribution patterns of TSP1, CD51, and CD36 were similar to those observed in their ductal counterparts (Table 2). However, in the fivecases of invasive lobular carcinomas studied, most of the tumor cells(40 to 80%) in "Indian file" strands were strongly stained for TSP1

(Fig. 6, A and B) and CD51 (Fig. 6, C and D). Less invasive tumorcells (30 to 40%) were immunostained for CD36 (Fig. 6, E and F).

Other Invasive Carcinomas (Mucinous Carcinoma, MedullaryCarcinoma, and Invasive Ductal Carcinoma with SquamousMetaplasia). In mucinous and medullary carcinomas, the distributionpatterns of TSP1, CDS 1, and CD36 were similar to those observed forinvasive ductal carcinomas NOS (Table 2). A similar immunoprofilepattern was also retained in invasive ductal carcinoma with squamousmetaplasia (Table 2). However, a weak staining for CD36 wasalso noted in some invasive tumor cells at the periphery of clusters(Table 2).

In Situ Hybridization

To further define the source(s) of TSP1 in normal and neoplasticbreast tissues, specific localization of TSP1 mRNA was performed byin situ hybridization (Fig. 7). With the specific antisense probe, in bothlight- and dark-field photomicrographs, there was high-intensity hy

bridization over myoepithelial cells (and luminal epithelial cells?) ofnormal ducts (Fig. 1A) and over myoepithelial cells surrounding insitu carcinomas (Fig. IB). Comparison with hybridization using thesense probe indicated minimal hybridization over tumor cells in in situcarcinomas. In invasive ductal carcinomas, there was high-intensity

hybridization over fibroblasts present in desmoplastic areas, whileinvasive tumor cells were negative (Fig. 1C). Nonspecific hybridization gave hybridization to a level similar to that observed for invasivetumor cells (not shown). In invasive lobular carcinomas, silver grainswere localized specifically over tumor cells (Fig. ID). Because of thehigh-intensity hybridization observed over invasive lobular carcinoma

cells, a higher magnification (X400) photomicrograph was used tolocalized tumor cells. Nonspecific hybridization with the sense probeshowed several grains over the stroma (not shown).

DISCUSSION

The aim of this study was to investigate the expression of TSP andtwo of its cell surface receptors (CD36, CD51) in normal, hyperplas-tic, and neoplastic human breast. For this purpose, four distinct anti-TSP monoclonal antibodies (PIO, PI2, MA-I, MA-II) were used inimmunohistochemistry, one of them (MA-II) being specific forTSPl

(33, 37). In addition, a probe specific for TSP1 mRNA (37) was usedby in situ hybridization. We have shown by both immunohistochemistry and in situ hybridization that, in normal breast tissue, TSP1mRNA is expressed by mammary epithelial cells and that, despite aweak immunostaining, TSP1 is incorporated in the basement membrane of nonlactating ducts. As previously observed by Wong et al.(25), immunostaining of normal breast tissue was found negative withanti-TSP monoclonal antibody PI2. However, in this study, otheranti-TSP antibodies such as PIO give a positive staining for TSP1 in

the basement membrane of nonlactating ducts. Such results might beexplained by the affinity of antibody PI2 for TSP1 which is muchlower than that of PIO (l UMand 10 nM, respectively) (32). Ourfindings regarding CD51 expression in normal and hyperplastic ductsare similar to those of other investigators (30, 31). Its colocalizationwith TSP1 in nonlactating ducts suggests that CD51 could serve as aTSP receptor in myoepithelial cells as it has been reported for endothelial and smooth muscle cells (13). In lactating adenomas, bothTSP1 and CD51 disappear from the myoepithelial-stromal junction of

ducts. However, TSP1 becomes selectively expressed at the apices ofsecretory epithelial cells of lactating ducts, confirming our previousfindings that TSP is present in breast secretions during initiation oflactation in humans (23). CD36, which is closely related to mammaryepithelial cell surface protein PAS-IV (26, 28), is also selectively

expressed in secretory epithelial cells. These findings are in agreementwith the presence of PAS-IV in breast secretions during lactation (27).The fact that both CD36 and PAS-IV bind TSP1 in vitro (28) strongly

suggests that CD36 is a TSP cell surface receptor in lactating ducts. Itseems therefore that the distribution of TSP 1 and the appearance of itsreceptors are dependent on the secretory activity of human mammaryducts.

Immunohistochemical studies performed by Wong et al. (25) andby ourselves (this study) clearly demonstrate a strong staining forTSP1 in the basement membrane surrounding in situ carcinomas andin desmoplastic areas of invasive ductal carcinomas. In addition, TSP1mRNA has been localized by in situ hybridization in myoepithelialcells surrounding in situ carcinomas and in fibroblasts present indesmoplastic areas. On the other hand, few invasive ductal carcinomacells (10%) express TSP1, while CD51 is moderately expressed bysome neoplastic clusters, and no immunoreactivity is observed withCD36. In this regard, it is most conceivable that fibroblasts are responsible forTSPl incorporation in the desmoplastic areas of invasiveductal carcinomas. This is consistent with the fact that fibroblastssynthesize and secrete TSP1 and that TSP1 promotes both adhesionand growth of fibroblasts (2). It has been suggested that desmoplasiain invasive ductal carcinomas could have a protective effect on tumorinvasiveness and metastasis (38). Increased deposits of fibronectin indesmoplasia of invasive breast carcinomas are observed, and thisoverexpression of fibronectin appears to reflect a low metastatic potential of the tumor (39). These findings (2, 25, 38, 39), taken togetherwith the fact that TSP1 is poorly effective in promoting migration ofhuman breast ductal carcinoma cells in vitro (16), promptly suggestthat TSP1 in desmoplasia may restrain migration and proliferation ofinvasive ductal carcinoma cells in vivo. As opposed to invasive ductal

1428

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMHOSPONDIN. CTO6. AND CD51 IN BRKAST CARCINOMAS

carcinoma cells, TSPI is codistributed with CDS I in most of theinvasive lobular carcinoma cells (40 to 80%) and with CD36 in asubpopulation (30 to 40%) of these invasive tumor cells. Our findingsregarding CD51 expression in invasive ductal and lobular carcinomasare similar to those of other investigators (30, 31). As discussed abovein the case of lactating adenomas, it is likely that the coexpression ofTSPI and CD36 by invasive lobular carcinoma cells is related to thesecretory activity of these cells. This is consistent with the fact thatlobular carcinomas exhibit a more or less constant and intensivesynthesis of milk proteins in invasive tumor cells (40). However, ourfindings are in sharp contrast with those of Wong et al. (25). Usinganti-TSF antibody PI2, the authors reported the absence of TSP im-

munostaining in invasive tumor cells of 2 cases of lobular carcinoma(25). As discussed above, the extent of TSP immunoreactivity may besubjected to great variation, depending on the affinity of the anti-TSPantibodies used. However, whatever the anti-TSP antibody may be, we

always observed an immunostaining in most of the invasive tumorcells (40 to 80%) (5 cases of lobular carcinoma studied). Moreover,using in situ hybridization, TSPI mRNA has been clearly located ininvasive lobular carcinoma cells. Interestingly, invasive ductal carcinoma cells do not express CD36, while CD36 and TSPI are codistributed in a subpopulation (30 to 40%) of invasive lobular carcinomacells. Subsets of invasive lobular carcinoma cells show a conspicuousincrease in expression of a,ß,and a6ß,integrins (29, 31). It has beensuggested that the overexpression of these integrins may increase thecapability of these neoplastic subpopulations to bind and to invadethrough the surrounding strema (29, 31). Besides its role as a TSPmembrane receptor (21, 22, 28), the function of CD36 has recentlybeen extended to that of a signal transduction molecule (41). CD36-

mediated signal transduction occurs through autophosphorylation ofprotein tyrosine kinases of the src gene family, and the functions of theCD36-associated ire-related kinases clearly involve regulation of cell

growth and differentiation (41, 42). On the basis of these findings (29,31,41, 42) and the results presented in this study, the codistribution ofTSP 1 and CD36 in subpopulations of invasive lobular carcinoma cellscould be related to the capability of these tumor cells to invadethrough the surrounding stroma. In addition, the absence of expressionof CD36 by invasive ductal carcinoma cells could well explain whyTSPI is poorly effective in promoting migration of these tumor cellsin vitro (16). The significance of CD36 expression in fibroblastssurrounding ducts associated with ADH and in situ comedocarcino-

mas remains obscure. However, these results further confirm the needto separate invasive ductal comedocarcinomas from other breast carcinomas of the ductal type.

Our data do not address the questions of whether other TSPs (TSP2and TSP3) and other receptors are expressed in normal and neoplastichuman breast. It is likely that a complex set of interactions betweenTSPs and their cell surface receptors is required for invasiveness ofbreast carcinoma cells. The different distribution of TSPI in invasiveductal versus lobular carcinomas may well reflect biological differences between these two main types of breast carcinomas. In thisregard, the coexpression of TSPI and CD36 may, in part at least,account for the variably invasive behavior of lobular carcinoma cells.By contrast, in the absence of CD36, excessive TSPI deposits indesmoplasia may restrain invasiveness of ductal carcinoma cells.

ACKNOWLEDGMENTS

P.C. acknowledges J. Lawler for monoclonal antibodies MA-I and MA-II

and cDNA clone M9 and D. Cheresh for monoclonal antibody LMI42.

REFERENCES

I. Baenziger, N. L.. Brodie, G. N., and Majerus. P. W. A thrombin-sensitive protein ofhuman platelet membranes. Proc. Nati. Acad. Sci. USA, 68: 240-243, 1971.

2. Lawler. J. The structural and functional properties of thrornbospondin. BlixHl. 67.1197-1209. 1986.

3. Clezardin, P., Jouishomme, H.. Chavassieux, P.. and Marie. P. J. Thrombospondin issynthesized and secreted by human osteoblasts and osteosarcoma cells. A model tostudy the different effects of thrornbospondin in cell adhesion. Eur. J. Biochem., 181:721-726. 1989.

4. Clezardin, P.. Malaval, L.. Ehrensperger, A. S., Delrnas. P. D.. Dechavanne, M., andMcGregor. J. L. Complex formation of human thromhospondin with osteoneclin. Eur.J. Biochem., 175: 275-284, 1988.

5. O'Shea, K. S., and Dixit, V. M. Unique distribution of the extracellular matrix

component thrombospondin in the developing mouse embryo. J. Cell Biol., 107:2737-2748. 1988.

6. Raugi. G. J., Olerud. J. E., and Gown, A. M. Thrombospondin in early human woundtissue. J. Invest. Dermatol.. 89: 551-554, 1987.

7. Raugi, G. J.. Mullen. J. S.. Bark, D. H.. Okada. T. and Maybcrg, M. R. Thrombospondin deposition in rat carotid artery injury. Am. J. Pathol.. 137: 179-185, 1990.

8. Good. D. J.. Polverini. P. J.. Rastinejad. F. Le Beau, M. M.. Lemons, R. S.. Frazier,W. A., and Bouck. N. P. A lumor suppressor-dependent inhibitor of angiogenesis is

immunologically and functionally indistinguishable from a fragment of thrombospondin. Proc. Nati. Acad. Sci. USA. 87: 6624-6628. 1990.

9. Castle, V.. Varani. J., Fligiel, S., Prochownik. E. V.. and Dixit. V. M. Antisense-

medialed reduction in thrombospondin reverses the malignant phenotype of a humansquamuus carcinoma. J. Clin. Invest.. 87: 1883-1888, 1991.

10. Tuszynski, G. P.. Gasic. T. B.. Rothman. V. L.. Knudsen. K. A., and Gasic, G. J.Thrombospondin, a potentialor of tumor cell metastasis. Cancer Res.. 47: 4130-4133.

1987.11. Sun. X.. Mosher, D. F.. and Rapraeger. A. Heparan sulfate-mediated binding of

epithelial cell surface proteoglycan to thrombospondin. J. Biol. Chem., 264: 2885-

2889. 1989.12. Roberts. D. D. Interactions of thrombospondin with sulfated glycolipids and pro-

teoglycans of human melanoma cells. Cancer Res.. 48: 6785-6793, 1988.13. Lawler. J-, Weinslein, R., and Hynes. R. O. Cell attachment to thrombospondin: the

role of Arg-Gly-Asp. calcium, and integrin receptors. J. Cell Biol.. 107: 2351-2361.

1988.14. Yabkowitz. R.. and Dixit. V. M. Human carcinoma cells bind thrombospondin through

a M, 80.000/105/XX) receptor. Cancer Res., 5/: 3648-3656. 1991.

15. Asch. A. S., Tepler. J.. Silbiger. S., and Nachman. R. L. Cellular attachment tothrombospondin. Cooperative interactions between receptor systems. J. Bini. Chem..266: 1740-1745, 1991.

16. Taraboletti. G., Roberts. D., and Liona. L. A. Thrombospondin-induced tumor cellmigration: haptotaxis and chemotaxis are mediated by different molecular domains. J.Cell Biol., 105: 2409-2415. 1987.

17. Bornstein. P., O'Rourke, K.. Wikstrom. K., Wolf. F. W., Katz. R.. Li, P.. and Dixit, V.

M. A second, expressed thrombospondin gene (Thbx 2} exists in the mouse genome.J Biol. Chem.. 266: I282I-I2824. 1991.

18. LaBell, T. L.. McGookey. D. J.. Disteche. C. M., and Byers. P. H. Thrombospondin[I: partial cDNA sequence, chromosome hx'ation, and expression of a second memberof the thrombospondin gene family in humans. Genomics, 12: 421—429,1992.

19. Vos. H. L.. Devarayalu, S., de Vries. Y, and Bornstein. P. Thrombospondin 3 (7Vi/>.v3), a new member of the thrombospondin gene family. J. Biol. Chem.. 267: 12192-12196. 1992.

20. Bomstein. P. Thrombospondins: structure and regulation of expression. FASEB J.. 6:3290-3299. 1992.

21. McGregor. J. L.. Catimel. B.. Parmentier. S.. Clezardin. P., Dechavanne. M.. andLeung. L. L. K. Rapid purification and partial characterization of human plateletglycoprotein 1Mb. Interaction with thrombospondin and its role in platelet aggregation. J. Biol. Chem.. 264: 501-506, 1989.

22. Catimel. B.. Leung. L.. Ghisassi. H.. Mercier. N.. and McGregor. J. L. Human plateletglycoprotein Illb binds to thrombospondin fragments bearing the C-terminal region,and/or the type I repeals (CSVTCG motif), but not the N-terminal heparin-bindingregion. Biochem. J.. 284: 231-236. 1992.

23. Dawes, J.. Clezardin. P.. und Pratt. D. A. Thrombospondin in milk, other breastsecretions, and breast tissue. Semin. Thromb. Hemostasis, 13: 378-384, 1987.

24. Pratt, D. A.. Miller. W. R.. and Dawes. J. Thrombospondin in malignant and non-malignant breast (issue. Eur. J. Cancer Clin. Oncol.. 25: 343-350. 1989.

25. Wong, S. Y.. Purdie, A. T. and Han. P. Thrombospondin and other possible relatedmatrix proteins in malignant and benign breast disease. An immunohisiochemicalstudy. Am. J. Pathol.. 140: 1473-1482, 1992.

26. Greenwalt. D. E.. Wall. K. W. K.. So, Y. O.. and Jiwani, N. PAS IV, an integralmembrane protein of mammary epithelial cells, is related to platelet and endothelialcell CD36 (GPIV). Biochemistry, 29: 7054-7059. 1990.

27. Greenwalt. D. E.. and Mather. I. H. Characterization of an apically derived epithelialglycoprolein from bovine milk, which is expressed in capillary endothelia in diversetissues. J. Cell. Biol.. 100: 397^108. 1985.

28. Catimel. B.. McGregor. J. L.. Hasler. T.. Greenwalt. D. E.. Howard, R. J.. and Leung.L. L. K. Epithelial membrane glycoprotein PAS-IV is related to platelet glycoprotein1Mbbinding to thrombospondin but not It) malaria infected erythrocytes. Blood, 77:2649-2654. 1991.

29. Gould. V. E., Koukoulis. G. K., and Virtanen. I. Extracellular matrix and theirreceptors in the normal, hyperplastic. and neoplastic breast. Cell Differ. Dev., 32:409^(16. 1990.

30. Zutter. M. M.. Mazoujian, G.. and Santoro. S. A. Decreased expression of integritiadhesive protein receptors in adenocarcinoma of the breast. Am. J. Pathol., 137:863-870. 1990.

31. Koukoulis. G. K.. Virtanen. !.. Korhonen. M.. Laitinen. L.. Quaranta. V, and Gould.V. E. Immunohistochemical localization of integrins in the normal, hyperplastic. andneoplastic breast. Am. J. Pathol., 139: 787-799. 1991.

32. Clezardin. P.. McGregor, J. L.. Lyon. M.. Clemetson. K.. and Huppen. J. Character-

1429

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

THROMBOSPONDIN. CD36. AND CDS I IN BREAST CARCINOMAS

ization of two murine monoclonal antibodies (PIO, PI2) directed against different 38. Barsky, S. H., and Gopalakrishna, R. Increased invasion and spontaneous metastasisdeterminants on human blood platelet thrombospondin. Eur. J. Biochem., 154: 95- of BL6 melanoma with inhibition of the desmoplastic response in C57 BL/6 mice.102, 1986. Cancer Res., 47: 1663-1667. 1987.

33. Lawler. J.. Derick. L. H., Connolly, J. E., Chen. J. H.. and Chao. F. C. The structure 39 Christensen. L.. Nielsen. M., Andersen. J., and Clemmensen, I. Stremai fibronectinof human platelet thrombospondin. J. Biol. Chem.. 260: 3762-3772. 1985. stamlng pattern and melaslaslzing abilit of human hreast cardnoma. Cancer Res..

34. Legrand. C.. Pidard. D.. Beiso, P.. Tenza, D.. and Edelman. L. Interaction of a ^. {p^-oTJI |cggmonoclonal antibody to clycoprotein IV (CD36) with human platelets and its effect ... . ... „¿�'â„¢¿�'.._on platelet function Platelets. 2: 9»-105. 1991. 40' £?*"* ,' r ClaSSlfica"On °fPnmary Breasl Carc'"oma- '«•Prohlems '" Breas<

35. Asch. A. S.. Bamwell. J.. Silverstein. R. L.. and Nachman. R. L. Isolation of the Pathology. J. G. Azzopard, (ed.). Saunders Company. Ltd.. Philadelphia. Vol. II,thrombospondin membrane receptor. J. Clin. Invest.. 79: 1054-1061. 1987. -4D-.55. 1979.

36. Cheresh. D. A., and Harper. J. R. Arg-Gly-Asp recognition by a cell adhesion 4I Greenwalt. D. E.. Lipsky, R. H.. Ockenhouse. C. F.. Ikeda. H.. Tandon. N. N., andreceptor refuses its 130-kd a subunil. J. Biol. Chem.. 262.- 1434-1437. 1987. Jamieson. G. A. Membrane glycoprotein CD36: a review of its role in adherence.

37. Lawler, J.. and Hynes, R. O. The structure of human thrombospondin. an adhesive signal transduction, and transfusion medicine. Blood, 80: 1105-1115. 1992.glycoprotein with multiple calcium-binding sites and homologies with several differ- 42. Carpenter, G. Receptor tyrosine kinase substrates: src homology domains and signalent proteins. J. Cell Biol.. 103: 1635-1648. 1986. transduction. FASEB J.. d: 3283-3289. 1992.

1430

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from

1993;53:1421-1430. Cancer Res   Philippe Clezardin, Lucien Frappart, Magali Clerget, et al.   Breastand CD51) in Normal, Hyperplastic, and Neoplastic Human Expression of Thrombospondin (TSP1) and Its Receptors (CD36

  Updated version

  http://cancerres.aacrjournals.org/content/53/6/1421

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

  .pubs@aacr.orgDepartment 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://cancerres.aacrjournals.org/content/53/6/1421To request permission to re-use all or part of this article, use this link

Research. on August 16, 2019. © 1993 American Association for Cancercancerres.aacrjournals.org Downloaded from