- Email: [email protected]
Integrin expression in primary breast cancer and its relation to axillary nodal status
Integrin expression in primary breast cancer and its relation to axillary nodal status Gerald P. H. Gui, FRCS, Clive A. Wells, MRCPath, Peter D. Browne, MSc, Patricia Yeomans, BSc, Suzanne Jordan, MSc, John R. Puddefoot, MSc, Gavin P. Vimon, DSc, and Robert Carpenter, FRCS, London, U.K. Background. Integrins are transmembrane receptors that modulate cell adhesion. Each is a heterodimer of varying a and 13 subunits. In malignancy, loss of integrin expression may result in less adhesive cells more likely to metastasize. Our aim was to characterize the integrins in human breast tissue and to examine the relationship between integrin expression and nodal metastasis in breast cancer. Methods. Cryostat sections from 72 benign and 67 malignant (50 ductal and I I lobular) samples were stained by the avidin-biotin complex method with monoclonal antibodies to the 131, 133, 134, and 135 subfamilies. All slides were read by two independent assessors with consensus agreement. Integrin expression was compared to variables by using the chi-squared test with Yates" correction and multivariate analysis based on logistic regression. Results. All integrin subunits studied were significantly reduced on breast cancer compared with benign cells (chi-squared test) but were not related to tumor differentiation. Loss of a1~1, a2~/, a3131, a6131, avflI, and av135 were related to the presence of axillary metastasis. Independently the integrins were of limited clinical value as predictors of axillary spread. However, on multivariate analysis the combination of 131, av, a 1, tumor size, and vascular invasion gave a cumulative overall accuracy in predicting nodal disease of 97%. Conclusions. Integrin expression is reduced in breast cancer and may explain tumor progression. Measuring the integrins might thus provide a means of selection for aggressive axillary treatment. (SURGERY 7995; 17 7: 702-8.) From the Departments of Surgery, Pathology, and Computer Services, St. Bartholomew's Hospital, and Department of Biochemistry, Queen Mary ~r Westfield College, London University, London, U.K.
THE INTEGRINS ARE A major family of cell adhesion molecules that regulate cell-matrix and cell-cell interactions. By influencing cell migration, anchorage, and differentiation, they play an important part in a diverse range of biologic processes including organogenesis, inflammation, and coagulation) When normal regulation of the integrins is disturbed, they predispose to pathologic conditions including tumor invasion and metastasis. 2 The integrins are transmembrane proteins with a large extracellular domain containing the receptor binding site and a smaller intracellular domain that links the cytoskeleton. The intact molecule is a glycoSupportedby Joint ResearchBoardof Trusteesof St. Bartholomew's Hospital and South Essex Medical Educationand Research Trust. Presented at the Fifty-fifthAnnual Meeting and Ninth' Tripartite Meeting of the SocietyOfUniversitySurgeons,Jackson, Miss., Feb. %12, 1994. Reprint requests:R. Carpenter,FRCS, The BreastUnit, St. Bartho. lomew's Hospital, 2nd Floor King GeorgeV Block,LondonEC1A 7BE, U.K. Copyright 9 1995 by Mosby-YearBook,Inc. 0039-6060/95/$3.00 + 0 11/6/59186 102
SURGERY
protein heterodimer composed of an a and/3 subunit, classified into subfamilies named after the /3 chain. Fourteen a and eight/3 subunits have been discovered, and these associate to form 20 integrin heterodimers. Their ligands are found on matrix proteins and basement and cell membranes)' 4 Tumor metastasis is a complex multistep process that includes cell adhesion as an important prerequisite. At various times in tumor dissemination a cell might be required to alter its adhesive properties. 2 In breast cancer most studies show a reduction of integrin expression, 57 although some conflicting results are noted. The relationship between integrin expression and the presence of axillary nodal metastasis has never been established. Reduced cell adhesion and aggressive malignant phenotype might be related to down-regulation of the integrin receptor. We therefore studied expression of the individual subunits of the ill,/33,/34, and/35 subfamilies, which are the main integrins found on glandular epithelium. Our aim was to characterize integrin expression on normal, benign, and malignant human breast tissue and to see whether discrepancies between
Surgery Volume 117, Number 1
Gui et al.
103
Fig. 1. Benign breast epithelium shows dual cell type and stromal fll integrin staining (original magnification x400).
existing studies might be resolved by more detailed analysis. Loss of integrin expression might result in less adhesive cells that are more likely to metastasize. We tested this hypothesis by examining the relationship between integrin expression on primary breast cancer cells and the presence of axillary nodal metastasis. PATIENTS A N D M E T H O D S T h e patient sample and tumor characteristics are shown in Table I. All patients with cancer had stage I or II disease treated primarily by segmental resection or mastectomy with axillary clearance. Sections of normal breast were obtained from patients undergoing operation for a breast lump; the normal biopsy specimen was taken from sites well away from the primary tumor. Each tumor was collected fresh from the operating room, and a segment was snap frozen in liquid nitrogencooled isopentane. Immunohistochemistry was performed by the avidin-biotin peroxidase complex staining technique on 5 #m cryostat sections. Sections were fixed in acetone at - 1 0 ~ C for 10 minutes and air dried, and multiple washes were performed in phosphatebuffered saline solution. Eleven monoclonal antibodies 8-17 (Table II) were used at concentrations predetermined in a titration assay and incubated for 1 hour. Endogenous peroxidase activity was blocked by 99.8% methanol with 3 0 o weight in volume hydrogen peroxide ( B D H Laboratory Supplies, Leicestershire, U.K.). Relevant biotin conjugated secondary antibodies raised in rabbit against mouse (Dako Ltd., High Wycombe, Bucks, U.K.) or rat (Vector Laboratories Ltd., Peterborough, U.K.) were used at 1:400 dilution for 30 minutes. Streptavidin-peroxidase conjugated complex (Dakopatts A/S, Copenhagen, Denmark) was added as the third layer, and the reaction was developed by using
T a b l e I. Patient sample and tumor characteristics Malignant (n = 61) Mean age (range): 56.1 (29-82) yr Mean tumor size (range): 2.4 (0.7-5.2) cm Histologic type (%): ductal, n = 50 (82); lobular, n = 11 (18) Grade (%): I, n = 14 (23); II, n = 25 (41); III, n = 22 (36) Lymph node positive (%): rz = 26 (43) Mean number of nodes involved (range): 4.5 (1-25) Mean number of nodes obtained at operation (range): 14.3 (7-30) Vascular invasion present (%): n = 19 (31) Benign (n = I2) Mean age (range): 49.4 (29-62) yr Histologic type (%): normal, n = 6 (50); fibroadenoma, n = 3 (25); fibrocystic, n = 3 (25)
3,3' diaminobenzidine tetrahydrochloride (Sigma Chemical Company, St: Louis, Mo.) and hydrogen peroxide. All sections were stained with Harris' hematoxylin, taken through a graded alcohol series, and mounted in D P X ( B D H Laboratory Supplies). Grading was performed by two assessors with consensus agreement wherever necessary, and membrane staining was scored as - (negative), + (weak), + + (moderate), + + + (strong), and + + + + (very strong). All equivocal sections were considered negative. Where necessary, sections were also stained with antibodies to laminin (Bionuelear Sciences, Reading, U.K.) and collagen type IV (Bionuclear Sciences) to distinguish in situ from invasive disease by the presence or absence of basement membrane. In situ and invasive carcinomas were scored separately. Normal skin was used as positive external tissue controls, and integrin expression on
104
Gui et al.
Surgery January 1995
T a b l e II. Monoclonal antibodies directed against integrin subunits used in this study
Epitope
Animal
Clone
Source
Dilution
Reference no.
al a2 a3 a4 a5 a6 oev [31 /33 /34 [35
Mouse Mouse Mouse Mouse Mouse Rat Mouse Mouse Mouse Mouse Mouse
TS2/7 P1E6 P1 B5 L25.3 P 1D6 CLB-701 LM142 DF5 RUU-PL 7F12 AA3 P1F6
Dr. M. Hemler, USA Becton Dickinson Chemicon Becton Dickinson Chemicon Chemicon Chemicon Chemicon Becton Dickinson Becton Dickinson Gibco
1:200 1:400 1:200 1:100 1:100 1:100 1:200 1:400 1:200 1:100 1:50
8 9 9 10 11 12 13 14 15 16 17
T a b l e III. Integrin expression in benign breast tissue
Median integrin staining intensity
Myoepithelium Luminal cells Stroma Smooth muscle Endothelium
a1
a2
+++ + ++ ++
+++ + . +
a3 +++ + . . ++ -
a4
a5
a6
av
~1
/33
/34
[35
+ . . -
+ -
+++ +
++ + +
+++ ++ ++ ++ ++
+ +
++ +
++ -
. .
. -
. +
++
T a b l e IV. Integrin expression as a predictor of axillary nodal metastasis in breast cancer
Integrin subunits
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) Overall accuracy (%)
al
a2
a3
a6
av
El
[35
73 66 61 77 69
77 74 69 81 75
85 69 67 86 75
77 71 67 81 74
88 74 72 90 80
70 97 95 81 85
100 32 55 100 63
stromal elements and blood vessels within each section served as positive internal tissue controls. A mouse monoclonal antibody against Aspergillus niger glucose oxidase, an enzyme neither present nor inducible in human beings, was used as a negative control (Dako Ltd.). Integrin expression was compared with variables by using the chi-squared test with Yates' correction and also multivariate analysis based on logistic regression with a stepwise forward approach and cross-tabulation (SPSS for Windows; SPSS Inc., Chicago, Ill.).
RESULTS B e n i g n breast tissue. Normal breast parenchyma consists of a dual epithelial cell type with an outer myoepithelial layer and an inner luminal or secretory layer. All the integrin subunits studied were expressed on myoepithelium (moderate to strong intensity), but ex-
pression on luminal cells was confined to the a2, a3, a6, and av members of the/31 subfamily (weak to moderate intensity). Integrin expression on myoepithelium was distributed basatly in contact with the substratum, reflecting cell-matrix interaction (Fig. 1). Integrin expression on luminal cells, where present, was polarized to the basal and basolateral surfaces of the cell. Similar staining patterns and intensities were observed in fibroadenomas and fibrocystic change. A summary of staining patterns and intensities of benign breast tissue including its supporting stroma is shown in T a b l e III. Breast cancer. In invasive malignant disease three patterns of integrin staining were identified. Integrin expression could be diffuse on tumor cell membranes (Fig. 2) with loss of polarization to the basolateral surfaces previously seen in benign disease. This pattern of staining represents intercellular interaction and was
Surgery Volume /17, Number /
Gui et al.
105
Fig. 2. Diffuse c~3 integrin staining on tumor cell membranes (original magnification X400).
Fig. 3. Peripheral a6 integrin enhancement of tumor cell nests (original magnification •
weak to moderate in intensity. Second, integrin staining could be confined to the periphery of tumor cell nests in moderate intensity (Fig. 3). This peripheral pattern of staining represents cell-matrix interactions. T h e third pattern of expression was focal or heterogenous staining, with loss of polarization, and was likely to represent focal proliferation of cells. This pattern of staining was uncommon, usually weak in intensity, and seen in less than 20% of sections studied. Integrin staining was significantly reduced or absent on malignant cells compared with benign breast parenchyma on chi-squared testing with Yates' correction (Fig. 4). No relationship was found between integrin expression and tumor grade, size, tiistologic type, age at presentation, or the presence of vascular invasion. In this series 26 patients had associated in situ carcinoma (20 ductal and six lobular) in addition to invasive
T a b l e V. Multivariate analysis of variables relating to nodal status Variable
p Value
Cumulative overall accuracy (%)
/31 Vascular invasion Tumor size av al a6
<0.0001 0.0004 0.003 0.02 0.03 0.07
85 87 93 95 97 97
disease on the sections studied. In these patients staining was always peripheral around in situ tumor nests. In a proportion of these patients integrin expression was also diffuse on the in situ tumor cells themselves. T h e relevance of this diffuse staining on in situ cells is un-
106
Gui et al.
Surgery January 1995
Fig. 4. Significant reduction of integrin expression was observed in breast cancer compared with benign cells by use of chi-squared test with Yates' correction (*, p < 0.05; **, p < 0.01; ***,p (0.001). On multivariate analysis the combination of integrin heterodimers most likely to be reduced in breast cancer were the ~2/31, a3/31, c~6/31, c~vl31, and ~v/33.
Fig. 5. Patients with axillary nodal metastasis are more likely to exhibit complete loss of integrin expression in primary breast cancer. Loss of seven of 11 subunits studied was significantly related to positive nodal status by use of chisquared test with Yates' correction (*,p < 0.01; **,p < 0.001; ns, not significant).
Certain, but it may represent a progressive loss of integrin expression in the change from in situ to invasive disease. No difference in expression between ductal and lobular in situ types was observed. Loss of integrin expression a n d axillary n o d a l metastasis. Studies of loss of c~1, o~2, o~3, o~6, ~v, 131, and /35 integrin subunits were significantly related to the presence of axillary nodal metastasis on chi-squared testing with Yates' correction (Fig. 5). When the r subunit was considered, 70% of patients who had positive lymph nodes had completely lost integrin expression on their breast cancer cells compared with 3% of patients who had negative lymph nodes. Test characteristics of each of these integrin subunits when considered individually are shown in Table IV. T o determine whether integrin expression was independent of other predictors of axillary nodal metastasis, multivariate analysis with a stepwise forward approach was performed. T h e ill, c~v, and a l subunits continued to be important independent predictors of axillary spread
(Table V). When the nonintegrin variables were incorporated, vascular invasion and tumor grade were found to be important predictors of axillary disease. In this model, when t31, c~v, and cd were considered together with vascular invasion and tumor size, the cumulative overall accuracy of positive and negative tests was 97%. Integrin expression of metastatic cells in axillary nodes was measured in 10 patients. Eight women had negative integrin expression in the primary tumor with corresponding negative expression in the nodes. In one patient, weak ~2, cd, and/31 expression was noted in the primary tumor with negative expression in the nodal metastasis. In the final patient the converse was true with negative expression in the primary but weak o~2 and f31 expression in the nodes. Distant metastatic tissue was not included in this study.
DISCUSSION T h e integrins provide a molecular basis for the mechanisms underlying tumor progression. M a n y tu-
Surgery Volume 717, Number 1 mor cell lines with decreased integrin receptor levels are associated with increased malignant phenotype) s Transfecting the a5/31 integrin back into fibronectin receptor-deficient transformed Chinese hamster ovary cells results in a loss of tumorigenic potential. 19 Loss of cell-cell and cell-matrix adhesion is an important prerequisite step in the metastatic cascade. 2 An understanding of thes e interactions may allow their exploitation as targets for therapeutic agents and as prognostic markers. Breast parenchyma possesses a dual epithelial cell type. We have found members of the/31 subfamily to be expressed on both myoepithelial and luminal cells but the/33,/34, and/35 subfamilies to be confined to myoepithelial cells only. Previous studies on breast tissue report conflicting results on the distribution and expression of certain integrin subunits. Expression of the a5 subunit on breast parenchyma is disputed, with some studies showing a reduction on malignant cells compared with normal, 6 whereas others suggest this subunit is absent on normal breast but present on ductal and mucinous carcinoma. 5 Our results support the evidence that a5 is present on normal and benign breast epithelium but is decreased or absent in malignancy. The/34 subunit has been reported to be present on luminal cells, 7 a finding that we are unable to confirm. The/34 subfamily in epithelial cells is currently thought to be associated with a6 as part of the hemidesmosome. 16 It is thus unlikely to be found on epithelial cells that are not in contact with the matrix; this is in keeping with our findings that/34 expression is confined to the myoepithelium. We also found/33 to be expressed on myoepithelium, unlike Koukoulis et al., 5 who failed to demonstrate /33 staining in breast parenchyma. The a4 subunit has not previously been described on breast tissue. We have shown that a4 is expressed on normal and benign breast epithelium but is absent on tumor cells. Other authors have shown a relationship between reduction of the a2, 20 a6, 21 and av/336 integrins with poor tumor differentiation. We were unable to confirm this in our study, although we found a weak association between reduction of/33 integrin expression and poor tumor grade that did not reach statistical significance. The av/33 integrin is important in cell spreading on matrix proteins and may work in conjunction with a5/31 and av/35.17' 22 It follows that the intact integrin heterodimer is required to mediate effects on normal cellular interactions. Loss of integrin expression would thus predispose to disordered cellular function and might account for the relation to poor tumor differentiation. A simple relationship to membrane receptor expression is clearly not the case because an overexpression of av/33 has been associated with aggressive malignant melanomas. 23 This study shows that loss of the a l , a2, a3, a6, av,
Gui et al.
107
ill, and/35 subunits is significantly related to axillary nodal metastasis. All these a subunits form heterodimers with/31, and av, in addition, associates with/35. These integrins collectively recognize laminin, type IV collagen, fibronectin, and vitronectin as their ligands. The interaction of tumor cells with these specific components of basement membrane and matrix proteins has been found to be of great importance in animal and in vitro experiments. 3 It is essential that studies investigating axillary metastasis such as ours use patients who have undergone an axillary clearance operation because sampling of the axilla might provide incomplete information with regard to true nodal status. 24 We found integrin expression in axillary metastasis to generally mirror that of the primary tumor. Although we did not examine the integrin expression of distant metastatic cells, Natali et al. 25 reported a reduction of a6/34 expression in extranodal compared with nodal metastasis. No relationship between the proportion of involved axillary nodes and loss of integrin expression in the primary tumor was found. This suggests that once loss of integrin expression occurs, the metastatic process becomes independent of integrin control. Vascular invasion predisposes to local recurrence and is a precursor of distant metastasis. We found no relationship between integrin expression and vascular invasion, which suggests that carcinoma cells invade vessels by alternative mechanisms, possibly by proteolytic enzyme activity. The a4/31 integrin that recognizes the vascular cell adhesion molecule on endothelial cells 1 was found to be absent on breast cancer cells. In breast cancer the surgical management of the axilia is a subject of great debate. 24 As yet, pathologic assessment of lymph nodes obtained by formal dissection is the only reliable method of staging the axilla. Therefore a proportion of women undergo axillary operation solely as a staging procedure for prognostic purposes with debatable therapeutic benefit. This study shows that the integrins may provide a means of predicting node positivity, thereby sparing axillary operations for those at low risk without loss of prognostic information. The test characteristics of the integrin subunits independently are clearly of limited clinical value because of the false-positive and -negative investigations (Table IV). However, when the/31, av, and a l integrin subunits are considered in conjunction with tumor size and vascular invasion (Table V), the overall accuracy is very high. A scoring index based on these variables may allow the accurate prediction of nodal status and may be a means of selecting patients for axillary surgery. We are currently evaluating this, as well as the role of the integrins as prognostic markers, in a prospective series. In this new and exciting area of breast cancer tumor biology, the integrins appear to be important determinants of aggressive malignant phenotype. Loss of inte-
108
Gui et al.
g r i n e x p r e s s i o n i n b r e a s t c a n c e r is likely to r e p r e s e n t t h e final c o m m o n p a t h w a y of m o r e c o m p l e x c h a n g e s o c c u r r i n g at a n i n t r a c e l l u l a r m o l e c u l a r level. N e v e r t h e l e s s , t h e e n d r e s u l t is a l t e r e d c e l l u l a r f u n c t i o n w i t h a g r e a t e r p r o p e n s i t y to t u m o r m e t a s t a s i s . M e a s u r i n g the i n t e g r i n s m a y t h u s b e of c o n s i d e r a b l e v a l u e as a p r e d i c t o r of a x illary disease a n d m a y a l l o w t h e t a i l o r i n g of a x i l l a r y s u r g e r y to suit t h e i n d i v i d u a l n e e d s of e a c h p a t i e n t . We thank Dr. M a r t i n Hemler of the D a n a - F a b e r Cancer Institute in Boston for providing the monoclonal antibody TS2/7. REFERENCES
t. Albelda SM, Buck CA. Integrins and other ceil adhesion molecules. FASEB J 1990;4:2868-80. 2. Hart IR, Saini A. Biology of tumour metastasis. Lancet 1992; 329:1453-61. 3. Coopman PJ, Bracke ME, Lissitzky JC, et al. Influence of basement membrane molecules on directional migration of human breast cell lines in-vitro. J Cell Sci 1991;98:395-401. 4. Gehlsen KR, Argraves WS, Pierschbacher MD, Ruoslahti E. Inhibition of in vitro tumor cell invasion by Arg-Gly-Asp-containing synthetic peptides. J Cell Biol 1988;106:925-30. 5. Koukoulis GK, Virtanen I, Korhonen M, Laitenen L, Quaranta V, Gould VE. Immunohistochemieal localisation of integrins in the normal, hyperplastic, and neoplastic breast. Am J Pathol 1991;139:787-99. 6. Zutter MM, Mazoujian G, Santoro SA. Decreased expression of integrin adhesive protein receptors in adenocarcinomas of the breast. Am J Pathol 1990;137:863-70. 7. Pignatelli M, Cardillo MR, Hanby A, Stamp GWH. Integrins and their accessory adhesion molecules in mammary carcinomas: loss of polarization in poorly differentiated tumours. Hum Pathol 1992;23:1159-66. 8. Hemler ME, Sanchez-Madrid F, Flotte T J, et al. Glycoproteins of 210,000 and 130,000 M.W. on activated T cells: cell distribution and antigenic relation to components on resting cells and T cell lines. J Immunol 1984;132:3011-8. 9. Carter WG, Wayner EA, Bouchard TS, Kaur P. The role of integrins a2fll and oe3fll in cell-cell and cell-substrate adhesion of human epidermal cells. J Cell Biol 1990;110:1387-404. 10. Clayberger C, Krensky AM, McIntyre BW, et al. Identification and characterisation of two novel lymphocyte-associated antigens, L24 and L252 J Immunol 1987;138:1510-4. 11. Wayner EA, Carter WG, Piotrowicz RS, Kunicki TJ. The function of multiple extracellular matrix receptors in mediating cell adhesion to extracellular matrix: preparation of monoclonal
Surgery January 7995 antibodies to the fibronectin receptor that specifically inhibit cell adhesion to fibroneetin and react with platelet glycoproteins IeIIa. J Cell Biol 1988;107:1881-91. 12. Sonnenberg A, Janssen H, Hogervorst F, Calafat J, Hilgers J. A complex of platelet glycoproteins Ic and lIa identified by a rat monoclonal antibody. J Biol Chem 1987;262:10376-83. 13. Cheresh DA. Human endothelial cells synthesize and express an Arg-Gly-Asp-directed receptor involved in attachment to fibrinogen and yon Willebrand factor. Proc Natl Acad Sci USA 1987;84:6471-5. 14. Ylanne J, Virtanen I. The Mr 140,000 fibronectin receptor complex in normal and virus-transformed human fibroblasts and in fibrosareoma cells: identical localisation and function. Int J Cancer 1989;43:1126-36. 15. Metzelaar M J, Korteweg J, Sixma J J, Nieuwenhaus HK. Biochemical characterization of PECAM-1 (CD31 antigen) on human platelets. Thromb Haemost 1991;6&700-7. 16. Jones JCR, Kurpakus MA, Cooper HM, Quaranta V. A function for the integrin a6f14 in the hemidesmosome. Cell Regul 1991;2:427-38. 17. Wayner EA, Orlando RA, Cheresh DA. Integrins avfl3 and eevfl5 contribute to cell attachment to vitronectin but differentially distribute on the cell surface. J Cell Biol 1991;113:919-29. 18. Schreiner C, Fisher M, Hussein S, Juliano RL. Increased tumorigenicity of fibroneetin receptor deficient chinese hamster ovary cell variants. Cancer Res 1991;51:1738-40. 19. Giancotti FG, Ruoslahti E. Elevated levels of the a5/31 fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell 1990;60:84%59. 20. Pignatelli M, Hanby AM, Stamp WH. Low expression of/31, a2 and a3 subunits of VLA integrins in malignant mammary tumours. J Pathol 1991;165:25-32. 21. D'Ardenne AJ, Richman PI, Horton MA, McCaulay AE, Jordan S. Co-ordinate expression of the alpha-6 integrin laminin receptor sub-unit and laminin in breast cancer. J Pathol 1991;165:213-20. 22. Charo IF, Nannizi L, Smith JW, Cheresh DA. The vitronectin receptor o~vfl3 binds fibronectin and acts in concert with a5fll in promoting cellular attachment and spreading on fibronectin. J Cell Biol 1990;111:2795-800. 23. Albelda SM, Mette SA, Eider DE, et al. Integrin distribution in malignant melanoma: association of the f13 subunit with tumour progression. Cancer Res 1990;50:6757-64. 24. Kissin MW, Thompson EM, Price AB, Slavin G, Klark AE. The inadequacy of axillary sampling in breast cancer. Lancet 1982;1:1210-11. 25. Natali PG, Nicotra MR, Botti C, Mottolese M, Bigotti A, Segatto O. Changes in expression of ee6/fl4 integrin heterodimer in primary and metastatic breast cancer. Br J Cancer 1992; 66:318-22.
THE INTEGRINS ARE A major family of cell adhesion molecules that regulate cell-matrix and cell-cell interactions. By influencing cell migration, anchorage, and differentiation, they play an important part in a diverse range of biologic processes including organogenesis, inflammation, and coagulation) When normal regulation of the integrins is disturbed, they predispose to pathologic conditions including tumor invasion and metastasis. 2 The integrins are transmembrane proteins with a large extracellular domain containing the receptor binding site and a smaller intracellular domain that links the cytoskeleton. The intact molecule is a glycoSupportedby Joint ResearchBoardof Trusteesof St. Bartholomew's Hospital and South Essex Medical Educationand Research Trust. Presented at the Fifty-fifthAnnual Meeting and Ninth' Tripartite Meeting of the SocietyOfUniversitySurgeons,Jackson, Miss., Feb. %12, 1994. Reprint requests:R. Carpenter,FRCS, The BreastUnit, St. Bartho. lomew's Hospital, 2nd Floor King GeorgeV Block,LondonEC1A 7BE, U.K. Copyright 9 1995 by Mosby-YearBook,Inc. 0039-6060/95/$3.00 + 0 11/6/59186 102
SURGERY
protein heterodimer composed of an a and/3 subunit, classified into subfamilies named after the /3 chain. Fourteen a and eight/3 subunits have been discovered, and these associate to form 20 integrin heterodimers. Their ligands are found on matrix proteins and basement and cell membranes)' 4 Tumor metastasis is a complex multistep process that includes cell adhesion as an important prerequisite. At various times in tumor dissemination a cell might be required to alter its adhesive properties. 2 In breast cancer most studies show a reduction of integrin expression, 57 although some conflicting results are noted. The relationship between integrin expression and the presence of axillary nodal metastasis has never been established. Reduced cell adhesion and aggressive malignant phenotype might be related to down-regulation of the integrin receptor. We therefore studied expression of the individual subunits of the ill,/33,/34, and/35 subfamilies, which are the main integrins found on glandular epithelium. Our aim was to characterize integrin expression on normal, benign, and malignant human breast tissue and to see whether discrepancies between
Surgery Volume 117, Number 1
Gui et al.
103
Fig. 1. Benign breast epithelium shows dual cell type and stromal fll integrin staining (original magnification x400).
existing studies might be resolved by more detailed analysis. Loss of integrin expression might result in less adhesive cells that are more likely to metastasize. We tested this hypothesis by examining the relationship between integrin expression on primary breast cancer cells and the presence of axillary nodal metastasis. PATIENTS A N D M E T H O D S T h e patient sample and tumor characteristics are shown in Table I. All patients with cancer had stage I or II disease treated primarily by segmental resection or mastectomy with axillary clearance. Sections of normal breast were obtained from patients undergoing operation for a breast lump; the normal biopsy specimen was taken from sites well away from the primary tumor. Each tumor was collected fresh from the operating room, and a segment was snap frozen in liquid nitrogencooled isopentane. Immunohistochemistry was performed by the avidin-biotin peroxidase complex staining technique on 5 #m cryostat sections. Sections were fixed in acetone at - 1 0 ~ C for 10 minutes and air dried, and multiple washes were performed in phosphatebuffered saline solution. Eleven monoclonal antibodies 8-17 (Table II) were used at concentrations predetermined in a titration assay and incubated for 1 hour. Endogenous peroxidase activity was blocked by 99.8% methanol with 3 0 o weight in volume hydrogen peroxide ( B D H Laboratory Supplies, Leicestershire, U.K.). Relevant biotin conjugated secondary antibodies raised in rabbit against mouse (Dako Ltd., High Wycombe, Bucks, U.K.) or rat (Vector Laboratories Ltd., Peterborough, U.K.) were used at 1:400 dilution for 30 minutes. Streptavidin-peroxidase conjugated complex (Dakopatts A/S, Copenhagen, Denmark) was added as the third layer, and the reaction was developed by using
T a b l e I. Patient sample and tumor characteristics Malignant (n = 61) Mean age (range): 56.1 (29-82) yr Mean tumor size (range): 2.4 (0.7-5.2) cm Histologic type (%): ductal, n = 50 (82); lobular, n = 11 (18) Grade (%): I, n = 14 (23); II, n = 25 (41); III, n = 22 (36) Lymph node positive (%): rz = 26 (43) Mean number of nodes involved (range): 4.5 (1-25) Mean number of nodes obtained at operation (range): 14.3 (7-30) Vascular invasion present (%): n = 19 (31) Benign (n = I2) Mean age (range): 49.4 (29-62) yr Histologic type (%): normal, n = 6 (50); fibroadenoma, n = 3 (25); fibrocystic, n = 3 (25)
3,3' diaminobenzidine tetrahydrochloride (Sigma Chemical Company, St: Louis, Mo.) and hydrogen peroxide. All sections were stained with Harris' hematoxylin, taken through a graded alcohol series, and mounted in D P X ( B D H Laboratory Supplies). Grading was performed by two assessors with consensus agreement wherever necessary, and membrane staining was scored as - (negative), + (weak), + + (moderate), + + + (strong), and + + + + (very strong). All equivocal sections were considered negative. Where necessary, sections were also stained with antibodies to laminin (Bionuelear Sciences, Reading, U.K.) and collagen type IV (Bionuclear Sciences) to distinguish in situ from invasive disease by the presence or absence of basement membrane. In situ and invasive carcinomas were scored separately. Normal skin was used as positive external tissue controls, and integrin expression on
104
Gui et al.
Surgery January 1995
T a b l e II. Monoclonal antibodies directed against integrin subunits used in this study
Epitope
Animal
Clone
Source
Dilution
Reference no.
al a2 a3 a4 a5 a6 oev [31 /33 /34 [35
Mouse Mouse Mouse Mouse Mouse Rat Mouse Mouse Mouse Mouse Mouse
TS2/7 P1E6 P1 B5 L25.3 P 1D6 CLB-701 LM142 DF5 RUU-PL 7F12 AA3 P1F6
Dr. M. Hemler, USA Becton Dickinson Chemicon Becton Dickinson Chemicon Chemicon Chemicon Chemicon Becton Dickinson Becton Dickinson Gibco
1:200 1:400 1:200 1:100 1:100 1:100 1:200 1:400 1:200 1:100 1:50
8 9 9 10 11 12 13 14 15 16 17
T a b l e III. Integrin expression in benign breast tissue
Median integrin staining intensity
Myoepithelium Luminal cells Stroma Smooth muscle Endothelium
a1
a2
+++ + ++ ++
+++ + . +
a3 +++ + . . ++ -
a4
a5
a6
av
~1
/33
/34
[35
+ . . -
+ -
+++ +
++ + +
+++ ++ ++ ++ ++
+ +
++ +
++ -
. .
. -
. +
++
T a b l e IV. Integrin expression as a predictor of axillary nodal metastasis in breast cancer
Integrin subunits
Sensitivity (%) Specificity (%) Positive predictive value (%) Negative predictive value (%) Overall accuracy (%)
al
a2
a3
a6
av
El
[35
73 66 61 77 69
77 74 69 81 75
85 69 67 86 75
77 71 67 81 74
88 74 72 90 80
70 97 95 81 85
100 32 55 100 63
stromal elements and blood vessels within each section served as positive internal tissue controls. A mouse monoclonal antibody against Aspergillus niger glucose oxidase, an enzyme neither present nor inducible in human beings, was used as a negative control (Dako Ltd.). Integrin expression was compared with variables by using the chi-squared test with Yates' correction and also multivariate analysis based on logistic regression with a stepwise forward approach and cross-tabulation (SPSS for Windows; SPSS Inc., Chicago, Ill.).
RESULTS B e n i g n breast tissue. Normal breast parenchyma consists of a dual epithelial cell type with an outer myoepithelial layer and an inner luminal or secretory layer. All the integrin subunits studied were expressed on myoepithelium (moderate to strong intensity), but ex-
pression on luminal cells was confined to the a2, a3, a6, and av members of the/31 subfamily (weak to moderate intensity). Integrin expression on myoepithelium was distributed basatly in contact with the substratum, reflecting cell-matrix interaction (Fig. 1). Integrin expression on luminal cells, where present, was polarized to the basal and basolateral surfaces of the cell. Similar staining patterns and intensities were observed in fibroadenomas and fibrocystic change. A summary of staining patterns and intensities of benign breast tissue including its supporting stroma is shown in T a b l e III. Breast cancer. In invasive malignant disease three patterns of integrin staining were identified. Integrin expression could be diffuse on tumor cell membranes (Fig. 2) with loss of polarization to the basolateral surfaces previously seen in benign disease. This pattern of staining represents intercellular interaction and was
Surgery Volume /17, Number /
Gui et al.
105
Fig. 2. Diffuse c~3 integrin staining on tumor cell membranes (original magnification X400).
Fig. 3. Peripheral a6 integrin enhancement of tumor cell nests (original magnification •
weak to moderate in intensity. Second, integrin staining could be confined to the periphery of tumor cell nests in moderate intensity (Fig. 3). This peripheral pattern of staining represents cell-matrix interactions. T h e third pattern of expression was focal or heterogenous staining, with loss of polarization, and was likely to represent focal proliferation of cells. This pattern of staining was uncommon, usually weak in intensity, and seen in less than 20% of sections studied. Integrin staining was significantly reduced or absent on malignant cells compared with benign breast parenchyma on chi-squared testing with Yates' correction (Fig. 4). No relationship was found between integrin expression and tumor grade, size, tiistologic type, age at presentation, or the presence of vascular invasion. In this series 26 patients had associated in situ carcinoma (20 ductal and six lobular) in addition to invasive
T a b l e V. Multivariate analysis of variables relating to nodal status Variable
p Value
Cumulative overall accuracy (%)
/31 Vascular invasion Tumor size av al a6
<0.0001 0.0004 0.003 0.02 0.03 0.07
85 87 93 95 97 97
disease on the sections studied. In these patients staining was always peripheral around in situ tumor nests. In a proportion of these patients integrin expression was also diffuse on the in situ tumor cells themselves. T h e relevance of this diffuse staining on in situ cells is un-
106
Gui et al.
Surgery January 1995
Fig. 4. Significant reduction of integrin expression was observed in breast cancer compared with benign cells by use of chi-squared test with Yates' correction (*, p < 0.05; **, p < 0.01; ***,p (0.001). On multivariate analysis the combination of integrin heterodimers most likely to be reduced in breast cancer were the ~2/31, a3/31, c~6/31, c~vl31, and ~v/33.
Fig. 5. Patients with axillary nodal metastasis are more likely to exhibit complete loss of integrin expression in primary breast cancer. Loss of seven of 11 subunits studied was significantly related to positive nodal status by use of chisquared test with Yates' correction (*,p < 0.01; **,p < 0.001; ns, not significant).
Certain, but it may represent a progressive loss of integrin expression in the change from in situ to invasive disease. No difference in expression between ductal and lobular in situ types was observed. Loss of integrin expression a n d axillary n o d a l metastasis. Studies of loss of c~1, o~2, o~3, o~6, ~v, 131, and /35 integrin subunits were significantly related to the presence of axillary nodal metastasis on chi-squared testing with Yates' correction (Fig. 5). When the r subunit was considered, 70% of patients who had positive lymph nodes had completely lost integrin expression on their breast cancer cells compared with 3% of patients who had negative lymph nodes. Test characteristics of each of these integrin subunits when considered individually are shown in Table IV. T o determine whether integrin expression was independent of other predictors of axillary nodal metastasis, multivariate analysis with a stepwise forward approach was performed. T h e ill, c~v, and a l subunits continued to be important independent predictors of axillary spread
(Table V). When the nonintegrin variables were incorporated, vascular invasion and tumor grade were found to be important predictors of axillary disease. In this model, when t31, c~v, and cd were considered together with vascular invasion and tumor size, the cumulative overall accuracy of positive and negative tests was 97%. Integrin expression of metastatic cells in axillary nodes was measured in 10 patients. Eight women had negative integrin expression in the primary tumor with corresponding negative expression in the nodes. In one patient, weak ~2, cd, and/31 expression was noted in the primary tumor with negative expression in the nodal metastasis. In the final patient the converse was true with negative expression in the primary but weak o~2 and f31 expression in the nodes. Distant metastatic tissue was not included in this study.
DISCUSSION T h e integrins provide a molecular basis for the mechanisms underlying tumor progression. M a n y tu-
Surgery Volume 717, Number 1 mor cell lines with decreased integrin receptor levels are associated with increased malignant phenotype) s Transfecting the a5/31 integrin back into fibronectin receptor-deficient transformed Chinese hamster ovary cells results in a loss of tumorigenic potential. 19 Loss of cell-cell and cell-matrix adhesion is an important prerequisite step in the metastatic cascade. 2 An understanding of thes e interactions may allow their exploitation as targets for therapeutic agents and as prognostic markers. Breast parenchyma possesses a dual epithelial cell type. We have found members of the/31 subfamily to be expressed on both myoepithelial and luminal cells but the/33,/34, and/35 subfamilies to be confined to myoepithelial cells only. Previous studies on breast tissue report conflicting results on the distribution and expression of certain integrin subunits. Expression of the a5 subunit on breast parenchyma is disputed, with some studies showing a reduction on malignant cells compared with normal, 6 whereas others suggest this subunit is absent on normal breast but present on ductal and mucinous carcinoma. 5 Our results support the evidence that a5 is present on normal and benign breast epithelium but is decreased or absent in malignancy. The/34 subunit has been reported to be present on luminal cells, 7 a finding that we are unable to confirm. The/34 subfamily in epithelial cells is currently thought to be associated with a6 as part of the hemidesmosome. 16 It is thus unlikely to be found on epithelial cells that are not in contact with the matrix; this is in keeping with our findings that/34 expression is confined to the myoepithelium. We also found/33 to be expressed on myoepithelium, unlike Koukoulis et al., 5 who failed to demonstrate /33 staining in breast parenchyma. The a4 subunit has not previously been described on breast tissue. We have shown that a4 is expressed on normal and benign breast epithelium but is absent on tumor cells. Other authors have shown a relationship between reduction of the a2, 20 a6, 21 and av/336 integrins with poor tumor differentiation. We were unable to confirm this in our study, although we found a weak association between reduction of/33 integrin expression and poor tumor grade that did not reach statistical significance. The av/33 integrin is important in cell spreading on matrix proteins and may work in conjunction with a5/31 and av/35.17' 22 It follows that the intact integrin heterodimer is required to mediate effects on normal cellular interactions. Loss of integrin expression would thus predispose to disordered cellular function and might account for the relation to poor tumor differentiation. A simple relationship to membrane receptor expression is clearly not the case because an overexpression of av/33 has been associated with aggressive malignant melanomas. 23 This study shows that loss of the a l , a2, a3, a6, av,
Gui et al.
107
ill, and/35 subunits is significantly related to axillary nodal metastasis. All these a subunits form heterodimers with/31, and av, in addition, associates with/35. These integrins collectively recognize laminin, type IV collagen, fibronectin, and vitronectin as their ligands. The interaction of tumor cells with these specific components of basement membrane and matrix proteins has been found to be of great importance in animal and in vitro experiments. 3 It is essential that studies investigating axillary metastasis such as ours use patients who have undergone an axillary clearance operation because sampling of the axilla might provide incomplete information with regard to true nodal status. 24 We found integrin expression in axillary metastasis to generally mirror that of the primary tumor. Although we did not examine the integrin expression of distant metastatic cells, Natali et al. 25 reported a reduction of a6/34 expression in extranodal compared with nodal metastasis. No relationship between the proportion of involved axillary nodes and loss of integrin expression in the primary tumor was found. This suggests that once loss of integrin expression occurs, the metastatic process becomes independent of integrin control. Vascular invasion predisposes to local recurrence and is a precursor of distant metastasis. We found no relationship between integrin expression and vascular invasion, which suggests that carcinoma cells invade vessels by alternative mechanisms, possibly by proteolytic enzyme activity. The a4/31 integrin that recognizes the vascular cell adhesion molecule on endothelial cells 1 was found to be absent on breast cancer cells. In breast cancer the surgical management of the axilia is a subject of great debate. 24 As yet, pathologic assessment of lymph nodes obtained by formal dissection is the only reliable method of staging the axilla. Therefore a proportion of women undergo axillary operation solely as a staging procedure for prognostic purposes with debatable therapeutic benefit. This study shows that the integrins may provide a means of predicting node positivity, thereby sparing axillary operations for those at low risk without loss of prognostic information. The test characteristics of the integrin subunits independently are clearly of limited clinical value because of the false-positive and -negative investigations (Table IV). However, when the/31, av, and a l integrin subunits are considered in conjunction with tumor size and vascular invasion (Table V), the overall accuracy is very high. A scoring index based on these variables may allow the accurate prediction of nodal status and may be a means of selecting patients for axillary surgery. We are currently evaluating this, as well as the role of the integrins as prognostic markers, in a prospective series. In this new and exciting area of breast cancer tumor biology, the integrins appear to be important determinants of aggressive malignant phenotype. Loss of inte-
108
Gui et al.
g r i n e x p r e s s i o n i n b r e a s t c a n c e r is likely to r e p r e s e n t t h e final c o m m o n p a t h w a y of m o r e c o m p l e x c h a n g e s o c c u r r i n g at a n i n t r a c e l l u l a r m o l e c u l a r level. N e v e r t h e l e s s , t h e e n d r e s u l t is a l t e r e d c e l l u l a r f u n c t i o n w i t h a g r e a t e r p r o p e n s i t y to t u m o r m e t a s t a s i s . M e a s u r i n g the i n t e g r i n s m a y t h u s b e of c o n s i d e r a b l e v a l u e as a p r e d i c t o r of a x illary disease a n d m a y a l l o w t h e t a i l o r i n g of a x i l l a r y s u r g e r y to suit t h e i n d i v i d u a l n e e d s of e a c h p a t i e n t . We thank Dr. M a r t i n Hemler of the D a n a - F a b e r Cancer Institute in Boston for providing the monoclonal antibody TS2/7. REFERENCES
t. Albelda SM, Buck CA. Integrins and other ceil adhesion molecules. FASEB J 1990;4:2868-80. 2. Hart IR, Saini A. Biology of tumour metastasis. Lancet 1992; 329:1453-61. 3. Coopman PJ, Bracke ME, Lissitzky JC, et al. Influence of basement membrane molecules on directional migration of human breast cell lines in-vitro. J Cell Sci 1991;98:395-401. 4. Gehlsen KR, Argraves WS, Pierschbacher MD, Ruoslahti E. Inhibition of in vitro tumor cell invasion by Arg-Gly-Asp-containing synthetic peptides. J Cell Biol 1988;106:925-30. 5. Koukoulis GK, Virtanen I, Korhonen M, Laitenen L, Quaranta V, Gould VE. Immunohistochemieal localisation of integrins in the normal, hyperplastic, and neoplastic breast. Am J Pathol 1991;139:787-99. 6. Zutter MM, Mazoujian G, Santoro SA. Decreased expression of integrin adhesive protein receptors in adenocarcinomas of the breast. Am J Pathol 1990;137:863-70. 7. Pignatelli M, Cardillo MR, Hanby A, Stamp GWH. Integrins and their accessory adhesion molecules in mammary carcinomas: loss of polarization in poorly differentiated tumours. Hum Pathol 1992;23:1159-66. 8. Hemler ME, Sanchez-Madrid F, Flotte T J, et al. Glycoproteins of 210,000 and 130,000 M.W. on activated T cells: cell distribution and antigenic relation to components on resting cells and T cell lines. J Immunol 1984;132:3011-8. 9. Carter WG, Wayner EA, Bouchard TS, Kaur P. The role of integrins a2fll and oe3fll in cell-cell and cell-substrate adhesion of human epidermal cells. J Cell Biol 1990;110:1387-404. 10. Clayberger C, Krensky AM, McIntyre BW, et al. Identification and characterisation of two novel lymphocyte-associated antigens, L24 and L252 J Immunol 1987;138:1510-4. 11. Wayner EA, Carter WG, Piotrowicz RS, Kunicki TJ. The function of multiple extracellular matrix receptors in mediating cell adhesion to extracellular matrix: preparation of monoclonal
Surgery January 7995 antibodies to the fibronectin receptor that specifically inhibit cell adhesion to fibroneetin and react with platelet glycoproteins IeIIa. J Cell Biol 1988;107:1881-91. 12. Sonnenberg A, Janssen H, Hogervorst F, Calafat J, Hilgers J. A complex of platelet glycoproteins Ic and lIa identified by a rat monoclonal antibody. J Biol Chem 1987;262:10376-83. 13. Cheresh DA. Human endothelial cells synthesize and express an Arg-Gly-Asp-directed receptor involved in attachment to fibrinogen and yon Willebrand factor. Proc Natl Acad Sci USA 1987;84:6471-5. 14. Ylanne J, Virtanen I. The Mr 140,000 fibronectin receptor complex in normal and virus-transformed human fibroblasts and in fibrosareoma cells: identical localisation and function. Int J Cancer 1989;43:1126-36. 15. Metzelaar M J, Korteweg J, Sixma J J, Nieuwenhaus HK. Biochemical characterization of PECAM-1 (CD31 antigen) on human platelets. Thromb Haemost 1991;6&700-7. 16. Jones JCR, Kurpakus MA, Cooper HM, Quaranta V. A function for the integrin a6f14 in the hemidesmosome. Cell Regul 1991;2:427-38. 17. Wayner EA, Orlando RA, Cheresh DA. Integrins avfl3 and eevfl5 contribute to cell attachment to vitronectin but differentially distribute on the cell surface. J Cell Biol 1991;113:919-29. 18. Schreiner C, Fisher M, Hussein S, Juliano RL. Increased tumorigenicity of fibroneetin receptor deficient chinese hamster ovary cell variants. Cancer Res 1991;51:1738-40. 19. Giancotti FG, Ruoslahti E. Elevated levels of the a5/31 fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell 1990;60:84%59. 20. Pignatelli M, Hanby AM, Stamp WH. Low expression of/31, a2 and a3 subunits of VLA integrins in malignant mammary tumours. J Pathol 1991;165:25-32. 21. D'Ardenne AJ, Richman PI, Horton MA, McCaulay AE, Jordan S. Co-ordinate expression of the alpha-6 integrin laminin receptor sub-unit and laminin in breast cancer. J Pathol 1991;165:213-20. 22. Charo IF, Nannizi L, Smith JW, Cheresh DA. The vitronectin receptor o~vfl3 binds fibronectin and acts in concert with a5fll in promoting cellular attachment and spreading on fibronectin. J Cell Biol 1990;111:2795-800. 23. Albelda SM, Mette SA, Eider DE, et al. Integrin distribution in malignant melanoma: association of the f13 subunit with tumour progression. Cancer Res 1990;50:6757-64. 24. Kissin MW, Thompson EM, Price AB, Slavin G, Klark AE. The inadequacy of axillary sampling in breast cancer. Lancet 1982;1:1210-11. 25. Natali PG, Nicotra MR, Botti C, Mottolese M, Bigotti A, Segatto O. Changes in expression of ee6/fl4 integrin heterodimer in primary and metastatic breast cancer. Br J Cancer 1992; 66:318-22.