Automated and quantitative immunocytochemical assays of CD44v6 in breast carcinomas

Automated and Quantitative Immunocytochemical Assays of CD44v6 in Breast Carcinomas COLETTE CHARPIN, MD, STEPHANE GARCIA, MD, CORINNE BOUVIER, ME), BENEDICTE DEVICTOR, PHD, LUCILE ANDRAC, MD, REGINE CHOUX, MD, MARIE-NOI~LLE LAVAUT, MD, AND CLAUDE ALLASIA, PHD s

s

CD44 variants carrying sequences encoded by exon v6 are preferentially expressed in metastatic animal cancer cell lines. CD44v6 overexpression correlates tumor dedifferentiation and progression in some human carcinomas, but the relationship of CD44v6 overexpression with metastatic behavior of tumor observed in animal models is controversial, particularly in breast carcinomas. The discrepancies probably result from mlalytical bias. We investigated CD44v6 mad CD44s expression in 218 frozen samples of primary breast carcinomas. lmmunocytochemlcal procedure was performed under optimal technical conditions using commercially available 2F-10 monoclonal antibody (MAb), a microprocessor-controlled automated device (Ventana Medical Systems, Tucson, AZ), and quantitative evaluation of results by processing digitized-colored microscopic images (SAMBA, Grenoble, France). CD44v6 expression in tissue sections was shown to be independent of the patient age, tumor size, histological types and grades, and the lymph node status. CD44v6 expression was also independent of the expression of molecules endowed with poor prog-

nostic significance detected by MAbs (anti-p53, anti-c-erb B-2 protein, MIB1) on consecutive sections. No significant relationship could be evidenced either between CD44v6 expression, and CD31 involved

stromal angiogenesis and cathepsin D. Finally, CD44v6 was independent of markers of hormone dependence (estrogen and progesterone receptors, pS2) and of multidrug resistance (P-glycoprotein). Similar results were observed with anti-CD44s. We conclude that the true prognostic significance of CD44v6 overexpression still remains to be shown under rigorous technical conditions (frozen samples, welldocumented MAbs, and optimal standardization of procedure using automation and quantitative analysis) providing data appropriate for further correlation with long-term patient follow-up. HUM PATHOL 28:289-296. Copyright © 1997 by W.B. Saunders Company Key words: CD44v6, automated immunocytochemistry, computerized image analysis, breast carcinomas. Abbreviations: MAb, monoclonal antibody; MOD, mean optical density; QIC, quantitative immunocytochemical index.

During t u m o r progression, carcinomatous cells detach f r o m the primary tumor, penetrate the b a s e m e n t m e m b r a n e in the connective tissue, and invade adjacent structures including lymph and blood vessels, thus attaining metastatic outgrowth. Loss of adhesive function resulting f r o m adhesive molecule down-regulation plays a crucial role in the initial step of the metastatic cascade of events. Expression of cancer i n v a s i o n - r e l a t e d proteins is of particular clinical relevance because failure of cancer t r e a t m e n t results f r o m metastases development. CD44 is a cell surface adhesive molecule that is involved in cell-cell and cell-matrix interaction. Several CD44 isoforms that are variants of c o m m o n standard hematopoietic form, CD44s, has b e e n identified by sequencing analysis. 1'2 CD44 gene contains 20 exons. T e n of these exons can potentially be alternatively spliced. 1'2 Epithelial cells contain exons 12 to 14 inserted into the CD44s transcripts, and this isoform is designated as CD44E. 2 O t h e r CD44 variants (CV44v) differ in the middle region located on the external side of the cell From the Department of Pathology, Facult6 de M6decine Timone and H6pital Nord, Marseille, France. Accepted for publication July 10, 1996. Supported by grants from LNLCC (Ligues Nationale et D6partmentale pour la Lutte contre le Cancer) and Institute F6d~ratif de Recherche en Canc~rologie et Immunologie de Marseille. Address correspondence and reprint requests to Colette Charpin, MD, Anatomie Pathologique, EA 875 "Oncogenese des tumeurs solides," Facult~ de Mfidecine, 27 BdJean Moulin, 13385 Marseille, France. Copyright © 1997 by W.B. Sannders Company 0046-8177/97/2803-001255.00/0

289

m e m b r a n e . 1'2 CD44v are variably expressed by epithelial cells 3'4 a n d have attracted considerable attention because they are involved in t u m o r progression. 5 Studies in animal models and on h u m a n cell lines have suggested a correlation between up-regulation of particular variants and metastasis. Transfection of CD44v6 (CD44 isoform containing v6) confers metastatic potential to nonmetastasizing pancreatic carcinoma cells in rats. ~'7 Coinjection of variant-specific m o n o c l o n a l antibodies with metastatic cells leads to retardation or complete blockage of metastatic spread in vivo) Spliced variants containing v6 are overexpressed by aggressive n o n - H o d g k i n ' s lymphomas 9 and also by some h u m a n carcinomas. CD44v6 is increasingly expressed during h u m a n colorectal 1° and gastric 1~ t u m o r progression. CD44 variants 4v, 6v, and 9v investigated by i m m u n o c y t o c h e m i c a l m e t h o d s are also, but variously, expressed by other types of h u m a n malignancies including lung, 12 ' 13 endometrial, 14 ovarian, 15 and urothelial 16 carcinomas and skin tumors, I7 whereas 5v is up-regulated in melanomas, is In breast carcinomas also, several recent immunocytochemical studies have shown an overexpression of CD44s 19 and of CD44v6 or of other variants. 2° However, these different reports have led to some controversies, and the relative contribution of each individual CD44 variant in different types of tumors still remains unclear. In endometrium 14and urothelia116 carcinomas, CD44 variants are down-regulated, whereas in colorectal, 1°'12'13 gastric, 11 lung, 12'*3 and ovarian carcinomas 15 CD44 variants are up-regulated. Morever, in studies of the same type of tumors, results are not always consistent. In colon carcino-

HUMAN PATHOLOGY Volume28, No. 3 (March 1997) mas, CD44v6 is overexpressed, 1° but in other reports CD44v6 expression in cancer is not significantly stronger than in normal tissue. 12'13'21 Similarly, discrepancies have been reported regarding CD44v6 expression of breast carcinomas and normal breast and CD44v6 prognostic significance. 12'13'22"25 All these controversies probably result from analysis bias 24 and shortcomings of the methods of investigation, particularly from lack of standardization of immunocytochemical procedures. Therefore, standardization of immunocytochemical techniques is required to document CD44v6 variations of expression in dssue and clarify the prognostic significance in breast cancer. The clinical relevance of CD44v6 expression in breast cancer depends on the reliability of the method of investigation of CD44v6 epitopes that may be a target for specific therapy. In the present study, we have investigated CD44v6 expression in 218 breast carcinomas by immunocytochemistry achieved u n d e r optimal technical conditions. A u t o m a t e d i m m u n o c y t o c h e m i c a l tests (Ventana device; Ventana Medical Systems, Tucson, AZ) were p e r f o r m e d on frozen sections, using well-documented c o m m e r cially available m o n o c l o n a l anti-CD44v6 and antiCD44s. I m m u n o p r e c i p i t a t e s on tissue sections were quantified by densitometry on digitized microscopic images processed by image analysis, and results were correlated with clinicopathologic data (univariate analysis) and with the expression of prognostic indicators evaluated according to the same immunocytochemical procedure.

MATERIALS AND METHODS Source of Tissue Samples The specimens were surgically obtained from 218 patients with breast carcinomas, from January 1993 to May 1994. Mean age was 56.7 years (range, 32 to 83 years). For all of them, surgical resection was the primary treatment, and none received irradiation or chemotherapy preoperatively. Surgical specimens were fixed in Bouin fixative, paraffin embedded, and hematoxylin-eosin and saffronin stained for routine microscopic diagnostic. Samples for immunodetection were taken from the representative cancerous lesions by pathologists, similar to the sample used for the intraoperative microscopic diagnostic on frozen sections. Tissue samples for immunodetections were promptly dipped in liquid nitrogen and stored frozen at -80°C in the tumor library of the laboratory.

of grade III, 27% of grade IV, and 4% of grade V. Tumors were also ranked according to the Nottingham prognostic index, 2s which ranged from 2.1 to 8 (mean = 4.1; SD = 1.22).

Immunohistological Staining Procedures Antibody Sources. Antihuman CD44 (80-95 KD) mouse rnonoclonal (immunoglobulin G1) antibody (clone J173) was purchased from Immunotech (Marseille, France) and antiCD44v6 (clone 2F10) (12,13) was purchased from RD Systems Europe (Abingdon, Oxon, United Kingdom) and used diluted at 1:1500 (~173) and 1:1000 (2F10). The other monoclonal antibodies used were all commercially available and used as previously described2g-~5:MIB1 and anti-CD31 (Immunotech SA, Marseille, France), anti-P53 clone DO-6 (Oncogene Science, Paris, France), anti-c-erb B-2 (Biogenex Menarini, Chevilly Larue, France), anti-cathepsin D (CisBio International, Gif sur Yvette, France), anti-P-glycoprotein (Pgp) cloneJSB1-N (Tebu, Le Perray en Yvelines, France), and anti-ER and PR (Abbott kits, Rungis, France). Automated Immunohistochemistry. Automated immunohistochemistry was performed on consecutive sections (4 # thick) with avidin-biotin-peroxidase complex on Ventana 320 device36 (Ventana) Ventana kits (Ventana Systems, Strasbourg France) including amino-ethyl-carbazol reagent. Sections were counterstained with hematoxylin, dehydrated, and mounted in glycergel.

Image Processing and Statistical Analysis Immunoprecipitates were analyzed using an Axiophot microscope (Zeiss, Rueil Malmaison, France) and a 3CCD camera (Sony, Paris, France), and then processed by an image analysis system (SAMBA 2005, Alcatel-TITN, Grenoble, Fr ance). 37 The two parameters of densitometric analysis, percentage ofimmunostained surface (versus counterstained surface), and mean optical density (MOD), which reflects the staining intensity (on SAMBA arbitrary units scale of 0 to 255), were obtained as previously reported. 29-35Also a quantitative immunocytochemical index (QIC) combining the percentage of stained surface and MOD was computed (% of stained surface/MOD × 100). Statistical analysis was assessed using BMDP (Biomedical Data Package) software (University of California, Berkeley, CA). Various statistical tests were used depending on the type (nominal or ordinal) and the distribution (normal or not) of the variables. Consequently, parametric or nonparametric tests were applied including the chi-squared test, Student's t-test, Kruskas Wallis test, MannWhitney U test, and the computation of correlation coefficients (Spearman, Kendall, and Pearson).

RESULTS

Histopathologic Features Tumor sizes ranged from 4 to 80 mm (mean = 16.8; SD = 11.95). In 89.9% of the cases (198 of 218), axillary lymph node resection was performed, and 62% of patients (n = 122) were node positive, and 38% (n = 76) were node negative. Carcinomas were in sire in 17 of 218 cases (8%). Invasive ductal carcinomas accounted for 66% (143 of 218), lobular carcinomas for 17% (38 of 218), and invasive carcinomas of other types for 9% (20 of 218). Tumors were graded according to the Bloom grading system (Scarff-Bloom-Richardson).26 Grade I tumors accounted for 22% (45 of 201), grade II tumors for 54% (108 of 201), and grade III tumors for 24% (48 of 201). Tumors were also graded according to a modified Bloom grading system27 in five grades: 8% of grade I, 30% of grade II, 31%

290

Patterns of CD44 Distribution in Ceils and Tissues Patterns of i m m u n o r e a c t i o n are shown a n d described in detail in Figs 1 to 3. In brief, anti-CD44s reacted with b o t h epithelial t u m o r cells a n d lymphocytes in t u m o r stroma (Fig 1A to C), whereas antiCD44v6 reacted with only epithelial cells (Fig 1D). The positive i m m u n o r e a c t i o n s with anti-CD44v6 did not differ in ductal carcinomas f r o m lobular carcin o m a s (Figs 2 and 3) n o r f r o m carcinomas of o t h e r types (unshown). In intraductal and invasive ductal carcinomas (grade I to III), staining was variable, but variations in

CD44v6 IN BREAST CARCINOMAS (Charpin et al)

FIGURE I. Frozen sections, automated immunoperoxidase (Ventana) in breast carcinomas. (A) Strong epithelial staining in in situ ductal high-grade carcinoma (large cell) with MAb J 173 (CD44s), (B) Moderate positive immunoreaction in invasive ductal component (left side) with stronger positive reaction in intraductal component (right side) and positive stroma lymphocytes (middle), with MAb J 173 (CD44s). (C) Stronger positive reaction in stromal lymphocytes (right) than in intraductal carcinoma (left). (D) Strong heterogeneous positive immunoreaction in invasive ductal carcinoma and negative reaction in stromal lymphocytes with MAb 2F10 anti CD44v6 (in contrast to C), (Streptavidin peroxidase, amino-ethyl-carbazol stain; original magnification [A-C] x75, [D] ×125.)

positive immunoreactions observed in individual t u m o r was i n d e p e n d e n t of t u m o r dedifferentiation (Fig 2). CD44 Quantitative Immunodetection

All the tumors (n = 218) were positive with anti CD44s/J173 MAb. But 17.4% (n = 38/218) were antiCD44v6 negative and 82.6% (n = 180 of 218) antiCD44v6 positive. Distribution of anti-CD44v6 positive staining evaluated by densitometry on tissue sections is shown in Figs 4 and 5. Positive tumors surface stained by anti-CD44v6 ranged from 1% to 57% (mean = 18%; SD = 12.1) (Fig 4). The CD44v6 quantitative immunocytochemical index varied from 2.7% to 64.7% (mean = 17.3%; SO = 11.8) (Fig 5). C D 4 4 Expression a n d C l i n i c o p a t h o l o g i c Data

CD44v6 imnmnostained surface evaluated by image analysis and QIC was i n d e p e n d e n t of tile patients' age, t u m o r size, histological type, tumor differentiation and histoprognostic grades, lymph node status, and Nottingham prognostic index. No significant difference

between the preceding parameters and anti-CD44v6positive or negative reaction was observed. Moreover, there was no significant relationship between CD44v6 overexpression evaluated by quantitative immunocytochemistry and any of the clinicopathologic parameters studied (Table 1). C o r r e l a t i o n o f C D 4 4 Expression w i t h Quantitative Immunodetection

Results of quantitative evaluation of growth fraction (MIB1), and immunoreaction with anti-p53, a n t i c-erb B-2 protein, anti-cathepsin D, anti-CD31, anti Pgp, anti-ER and -PR antigenic sites, and anti-pS2 are shown in Table 2. No significant correlation was observed between CD44v6 or CD44s expression and the tumor tissue contents of these antigens evaluated by the same p r o c e d u r e on consecutive frozen sections from the same tissue blocks (Table 2). DISCUSSION H u m a n carcinomas CD44s and CD44 variants, particularly CD44v6 expression, have been d o c u m e n t e d in

291

HUMAN PATHOLOGY

Volume 28, No. 3 (March 1997)

FIGURE 2. Frozen sections, MAb 2FI0 (anti-CD44v6) automated immunoperoxidase (Ventana), in breast carcinomas. (A) Moderate positive staining heterogeneously distributed in cancerous duct in intraductal low-grade carcinoma (cribriform, small cells). (B) Moderate and strong positive staining in invasive ductal carcinoma grade 1, with prominent staining on cell membrane contrasting with mixed cytoplasm-cell membrane pattern of staining shown in Fig 1D. (C) Strong but heterogeneously distributed positive staining in invasive ductal carcinoma grade II. (D) Strong staining in ductal carcinoma grade III, in which most cells are positive, but membranes are only focally stained in individual cells. (Streptavidin peroxidase amino-ethyl carbazol; original magnification [A, B] ×75, [C, D] ×350.)

FIGURE 3. Frozen sections, MAbs 2F10 (anti CD44v6), automated immunoperoxidase (Ventana) in invasive, lobular breast carcinomas. (A) All the tumor cells are positive, and cell membranes are either strongly immunoreactive, or weakly immunoreactive and barely visible. (B) Only some cells are strongly positive along cell membrane or cytoplasm, whereas others are negative or focally faintly positive. (Streptavidin peroxidase, amino-ethyl carbazol; original magnification ×350.)

292

CD44v6 IN BREASTCARCINOMAS (Charpin et al) Breast ca. ( % )

25.0 22.5

o.o

L

17.5 15.0 12.5

!

10.0

7.5 L 5.0 0

!

0

]0

20

!

g

u

30

40

u

u

50

positive CD 44 - 6v surface ( % )

FIGURE 4. Histogram of immunoreactive tumor surface positively stained in the CD44v6 positive breast carcinomas (n = 180 of 218) evaluated by processing (SAMBA device) of digitized microscopic images (frozen section net significant; 2F10 MAb, Ventana automated immunoperoxidase procedure).

normal breast and have b e e n shown to correlate variously t u m o r differentiation and progression, and prognostic. Normal breast expresses CD44v63']2']~'2°'25 like various other normal glandular structure epithelia, 12-]6' 2] although o t h e r reports showed no CD44v6 expression in normal breast gland 2~ or only in myoepithelial cells. 25 Also in h u m a n carcinomas, CD44v6 expression has b e e n shown to decrease with t u m o r dedifferentiation in urothelial carcinomas ]6 and also with lymph vascular space involvement in endometrial carcinomas. ~4 Conversely, CD44s overexpression correlates p o o r prognostic and t u m o r progression in breast c a r c i n o m a s ] 9 In breast, colon, and gastric carcinomas, it was claimed that CD44v61°'H'2°'23 and o t h e r variants 22 were up-regulated, and up-regulation correlates a o o r prognostic . • 20'22'23 1n with metastases d e v e l o p m e n t and low Psurvival skin tumors, CD44v6 expression increases in metastatic tumors a7 but not in melanomas, TM in which only CD44v5 overexpression correlates a p o o r prognostic, as In contrast, in o t h e r reports, although CD44v6 could be strongly expressed in carcinoma cells, particularly in aggressive tumors, its overexpression could not significantly correlate .tum°r P12~91r° ression,, low15survival, and • status m colon, ' ' ovarmn, l u n - g, ~2 ' ~3 and metastanc breast carcinomas, ~2'~3'24n o r its down-regulation, in urothelial carcinoma, a6 The discrepancies observed in these recent studies can be explained. 24 by bias of patients selecnon" in the series stuthed a n d the lack of standardization immunocytochemical assays including different properties of the antibodies used, 24 the variations in the immunocytochemical procedures p e r f o r m e d , a n d the shortcomings of subjective semiquantitative evaluation of the results. I m m u n o c y t o c h e m i c a l assays have some m a j o r advantages regarding o t h e r m e t h o d s because they require a small a m o u n t of tissue, are rapidly processed (especially on frozen tissue samples), and enable accurate 293

localization of i m m u n o r e a c t i o n s in tissue, particularly for evaluation of the differences in the antigens distribution between normal, hyperplastic, and cancer tissue. Immunohistochemistry can easily be used to evaluate the expression in cancer tissue of antigens that are endowed with clinical relevance because they are prognostic indicators or may be potential targets for specific therapy, provided (1) that immunocytochemistry sensitivity is sufficient and correlates other current techniques of measurement, and (2) that immunocytochemical assays are standardized and, therefore, reliable for patients therapy monitoring. In this respect, several previous reports have shown that CD44 immunocytochemical assays correlate results of RT-PCR, although the latter is, as expected, m o r e sensitive. 3,~z'13,ls'2°'~2,2~ So, basically, i m m u n o c y t o c h e m i cal assays can be validated if achieved according to optimal technical conditions for standardization. Discrepancies regarding the prognostic significance of CD44 expression in breast cancer can be explained by the fact that the different antibodies used recognize different CD44 isoforms or variants. The studies showing CD44 overexpression with t u m o r progression in breast carcinomas were achieved either with m o n o c l o n a l anti CD44s ~9 or with polyclonal antiCD44v6 VFF 7, D I I I . 23-25 Studies showing no significant relationship between p o o r prognostic t u m o r progression or low survival were achieved in breast cancer samples with 2F10 MAb ~2'~3 as in this study, and with FW 11.9 and 11.31 M A b s 3'24 in m e l a n o m a s with VFF 7 and VFF 18 is and in colon carcinomas with F10.44. 21 Studies showing that conversely CD44v6 down-regulation correlated with t u m o r dedifferentiation in urothelial carcinomas were assessed with individually developed m o n o clonal antibody. 16 Such heterogeneity of the antibodies is probably responsible for the variations of the CD44 epitopes de-

Breast c a . ( % )

0

5

10 15 20 25 30 35 40 45 50 55 60 65 70

CD 44 - 6 v Q I C

FIGURE 5. Histogram of quantitative immunocytochemical index (QIC) distribution (% of positive surface/mean optical density ×100) in the CD44v6 positive (n = 180 of 218) breast carcinomas evaluated by processing (SAMBA device) ef digitized microscopic images (frozen sections; 2F10 MAb, Ventana automated immunoperoxidase procedure).

HUMAN PATHOLOGY TABLE 1. Groups Age (yr) ~<50 (n = 7 2 ) >~50 (n = 108) Tumor size (cm) ~<1.5 (n = 83) ~>1.5 (n = 97) Histological type In situ (n = 14) Ductal invasive (n = 119) Lobular invasive (n = 31) Histoprognostic grade 26 Grade 1 (n = 37/166) Grade 2 (n = 89/166) Grade 3 (n = 40/166) Modified histoprognostic grade 2~ Group I (n = 124/180) Group II (n = 56/180) Prognostic index NP128 ~3.4 (n = 19/163) >3.4-~<5.4 (n = 106/163) >5.4 (n = 38/163) Lymph node invasion O (n = 110/198) 1-3 (n = 48/198) > 3 (n = 20/198)

Volume 28, No. 3 (March 1997)

Expression of CD44v6 in 180 of 218 Positive Breast Carcinomas in Various Subgroups of Patients and Tumors Positive No. (%)

CD44v6 Mean Surface %*

2F 10 MAb QICJ"

Statistical Tests

57 (79) 91 (84)

19.4 16.6

17.7 16.5

Not significant

61 (74) 88 (90)

15.7 20.3

20 15.3

Not significant

10 (73) 109 (92) 25 (80)

17.1 22.4 15,2

16.2 19.2 16

Not significant

26 (71) 74 (83) 37 (92)

15.5 20,5 18

16.4 17 20

Not significant

97 (78) 48 (86)

16,1 19.9

16.8 "17.6

Not significant

13 (70) 88 (83) 35 (93)

15.7 I8.8 19,5

16.7 17.4 17.5

Not significant

62 (75) 39 (81) 18 (90)

17 14,4 22.6

15.4 16.9 19.1

Not significant

Abbreviations: MAb, monoclonal antibody; QIC, quantitative immunocytochemical index; NPI, Nottingham prognostic index; MOD, mean optical density. * Mean positive surface evaluated by image analysis. t QIC cmnputed: % stained surface/MOD ;<100.

tected, and no practical conclusion can be drawn from the literature survey. In this study, using commercially available well-documented MAb 2F-10, we observed, like others, 12'13 no significant association of CD44v6 overexpression with any other histoprognostic indicators or t u m o r progression. Another clue for immunocytochemical assays standardization concerns the use of frozen tissue sections

TABLE 2.

Immunoreactions Evaluated by Image Analysis by Densitometry of Digitized Microscopic Preparations (Frozen Sections From 218 Breast Carcinomas/SAMBA System)

Growth fraction (Ki67/MIB1) p53 c-erb B-2 protein Cathepsin D CD31 P-gp ER PR pS2

in which antigens are optimally preserved, in contrast to paraffin sections of fixed tissue. In most of the recent studies, CD44 expression was evaluated in frozen samples as in this study, and so the CD44v6 antigen damage c a n n o t be responsible for inconsistent results. However, difficulties regarding the collection and storage of tum o r frozen samples in a t u m o r library for a long time may explain that the n u m b e r of tissue samples investi-

Positive Cases (%)

Mean Surface Stained (%)

n = 216/218 (99) n = 71/218 (32.5) n = 151/218 (69.3) n = 218/218 (100) n = 218/218 (100) n = 83/218 (30) n = 153/218 (70.2) n = 138/218 (63.3) n = 99/218 (45.4)

14.2 (SD = 11) 13.5 (SD = 12.4) 32.3 (SD = 24.7) 10,6 (SD = 6.9) 11.2 (SD = 6.5) 5.2 (SD = 4.7) 13.6 (SD + 24.7) 33.5 (SD = 32.8) 6.4 (SD = 6.9)

294

Range (rain/ max) (%)

CD44v6 Correlation

2.1-62.4

Not significant

0.7-72.3

Not significant

3.2-83

Not significant

2-53

Not significant

6-31

Not significant

0.5-16

Not significant

1-57

Not significant

0.5-68

Not significant

1-29

Not significant

CD44v6 IN BREAST CARCINOMAS (Charpin et al)

gated is relatively small, as already re]0orted. 24 Also, investigations of antigens in frozen samples usually concerns prospective studies, in contrast to retrospective studies assessed on archival blocks of paraffin-embedded tissues. Both large series and long-term patient follow-up are required to establish the true prognostic significance of CD44v6 expression in breast carcinomas. In the present study we first aimed at developing CD44v6 optimal m e t h o d of detection and of evaluation of CD44 expression using (1) commercially available and well-documented antibody, and (2) automated, and (3) quantitative immunohistochemistry. We shall further apply this m e t h o d of detection and correlate the results of CD44v6 detection with patient follow-up (work in preparation). Standardization of ICAs also involves the technique itself. Inconsistent results in CD44v6 colon carcinomas10, 12, is, 2l may result from the variations in the m e t h o d used (eg, immunoperoxidase or immunofluorescence. ~5 Also, the results rely on the quality of the technique, and automation ideally provides for a better quality control, particularly regarding reproducibility, in addition to the fact that many tissue samples may be identically p r o b e d at the same time. We, therefore, investigated the 218 tumor samples in this series using the same automated p r o c e d u r e and commercially available device, 36 to eliminate bias because of manual procedure flaws and to ensure that variations in immunoreactions with anti-CD44v6 reflect as accurately as possible the variations in tumor CD44v6 antigen content. This is a major point of interest, especially if the aim of the study is to compare CD44v6 distribution in various groups of tumors. Sharp statistical analyses are meaningless if variations reflect, even partly, differences in processing the batch of slides. No previous study using an automated m e t h o d of immunodetection of CD44v6 has previously been reported. Similarly, shortcomings may also result from semiquantitative analysis of immunoprecipitates. Semiquantitative evaluation of results of immunodetections is convenient, rapidly assessed, and is a cost-effective m e t h o d of analysis. However, its reliability depends on observer experience and reproducibility. Although semiquantitative evaluation is sufficient to differentiate negative versus positive reaction, or weak versus strong immunoreaction, it is not accurate enough to evaluate intermediate patterns of staining quantitatively. This must be pointed out, particularly when the quantitative variation of the distribution of antigen is correlated with prognostic parameters,~such as survival, metastasis, and recurrence, or various histoprognostic indicators. Discrepancies in prognostic significance of CD44v6 expression in major h u m a n carcinomas, such as colon or breast, likely result from this type of analysis bias. In the present study, we have evaluated the results of CD44v6 expression in tissue by computerized processing of digitized microscopic image, using the same hardware and software as in our previous studies. 293~ The results were obtained more objectively, variations of staining more accurately evaluated, and numerical values of parameters more appropriate to statistical analysis.

295

In the present study, the CD44v6 overexpression did not correlate patient age, tumor size, current or histoprognostic indicators including histological types and grades, and lymph node status, in accordance to some, but not all 19' 24 previous reports. However, these results cannot be compared with those of other studies because, as m e n t i o n e d earlier, the antibodies and methods used in those studies differed from the present study. Few studies, available in the literature, have correlated CD44v6 with the expression of other molecular prognostic indicators. CD44v6 expression, in this study, was shown to be i n d e p e n d e n t of the expression of molecules involved in other steps of tumor invasion, such as angiogenesis (CD31) or extracellular matrix digestion (cathepsin D). Similarly, CD44v6 expression was unrelated to that of p53 or c-erb B-2 protein tumor content and growth fraction (MIB1), suggesting that potential CD44v6 role in tumor invasion or progression is indep e n d e n t of cell proliferation deregulation. Similar findings for c-erb B-2 have been reported previously. 23 CD44v6 overexpression correlated strong Ki67 immunolabeling (growth fraction) in colon carcinomas. 21 Our results are consistent with the lack of relationship of CD44v6 expression and the tumor size, grades, node status, and prognostic index. The lack of correlation of CD44v6 overexpression with h o r m o n e receptor antigens and e n c o d e d pS2 protein in tissue evaluated by immunocytochemical procedures, with P-glycoprotein, suggests that CD44v6 is not involved in h o r m o n e response or multidrug resistance. However, our results are not consistent with some of those previously reported because an inverse relationship between CD44v6 expression and h o r m o n e receptors has been shown in breast c a n c e r . 2a In conclusion, we showed that CD44v6 overexpression, evaluated by automated and quantitative immunocytochemical assays on frozen tissue samples of breast carcinomas, is not significantly related to tumor size, types and grades, lymph node status, and markers of high degree of proliferation (p53, c-erb B-2, protein MIBI antigen). These data suggest that CD44v6 overexpression in tissue accurately evaluated u n d e r optimal technical conditions, relieved of potential major analysis bias, is not significantly involved in tumor cell proliferation and differentiation. Also, the fact that CD44v6 expression does not show relationships with the lymph node status CD31 and cathepsin D suggests that a CD44v6 potential role in tumor invasion and metastasis is i n d e p e n d e n t of angiogenesis and extracellular matrix digestion. In addition, specific therapy, including a blockage of CD44v6 as shown in in vitro models, would not act through mechanisms of h o r m o n o d e p e n d e n c e or multidrug resistance. However, the true prognostic significance of CD44v6 expression in breast cancer tissue remains to be shown, and results of quantitative and automated immunocytochemistry have to be futher correlated with patient follow-up (work in preparation).

REFERENCES 1. Arch R, Wirth K, Ho£mann M, et al: Participation in normal i m m u n e responses of a metastasis-inducing splicing variant of CD44. Science 257:682-685, 1992

HUMAN PATHOLOGY

Volume 28, No. 3 (March 1997)

2. Screaton GR, Bell MV, Jackson DG, et al: Genomic structure of DNA coding the lymphocyte homing reception CD44 reveals at least 12 alternatively spliced exons. Proc Natl Acad Sci U S A 89:1216012164, 1992 3. Mackay CR, Terpe HJ, Stauder R, et al: Expression and modulation of CD44 variant isoforms in humans. J Cell Biol 124:71-82, 1994 4. Terpe HJ, Stark H, Prehm P, et al: CD44 variant isoforms are preferentially expressed in basal epithelia of non-malignant human fetal and adult tissues. Histochemistry 101:79-89, 1994 5. Sy MS, Guo YJ, Stamenkovic I: Distinct effects of two CD44 isoforms on tumor growth in vivo. J Exp Med 174:859-866, 1991 6. G~mthert U, Hofmann M, Rudy W, et al: A new variant of glycoprotein CD44 confers metastatic potential to rat carcinoma cells. Cells 65:13-24, 199I 7. Rudy W, Hofn-tann R, Schwartz-Albiez M, et al: Two major CD44 proteins expressed on a metastatic rat tumor cell line are derived from different splice variants: Each one individually suffices to confer metastatic behavior. Cancer Res 53:1262-1268, 1993 8. Seiter S, Arch R, Reber S, et al: Prevention of tumor metastasis formation by anti-variant CD44. J Exp Med 177:443-455, 1993 9. Koopman G, Heider KH, Horst E, et al: Activated human lymphocytes and aggressive non-Hodgkin's lymphomas express a homologue of the rat metastasis-associated variant of CD44. J Exp Med 177:897-904, 1993 10. Wielenga VJ, Heider K, Offerhaus JA, et al: Expression of CD44 variant proteins in human colorectal cancer is related to tumor progression. Cancer Res 53:4754-4756, 1993 11. Heider KH, Dammrich J, Skroch-Angel P, et al: Differential expression of CD44 splice variants in intestinal and diffuse-type human gastric carcinomas and normal gastric mucosa. Cancer Res 53:41974203, 1993 I2. Fox SB, Gatter KC, Jackson DG, et al: CD44 and cancer screening. Lancet 342:548-549, 1993 13. Fox SB, Fawcett J, Jackson DG, et al: Normal human tissues, in addition to some tumors, express multiple different CD44 isoforms. Cancer Res 54:4539-4546, 1994 14. Fujita N, Yaegashi N, Ide Y, et al: Expression of CD44 in normal human versus tumor endometrial tissues: Possible implication of reduced expression of CD44 in lymph-vascular space involvement of cancer cells. Cancer Res 54:3922-3928, 1994 15. Cannistra SA, Abu-Jawdeh G, NiloffJ, et al: CD44 variant expression is a common feature of epithelial ovarian cancer: Lack of association with standard prognostic factors. J Clin Oncol 13:19121921, 1995 16. Hong RL, Pu YS, Hsieh TS, et al: Expressions of E-cadherin and exon V6-containing isoforms of CD44 and their prognostic values in human transitional cell carcinoma. J Urol 153:2025-2028, 1995 17. Guttinger M, Sutti F, Barnier C, et al: Expression of CD44 standard and variant forms in skin tumors loss of CD44 correlates with aggressive potential. EurJ Dermatol 5:398-406, 1995 18. Manten-Horst E, Danen EHJ, Smit L, et al: Expression of CD44splice variants in human cutaneous melanoma and melanoma cell lines is related to tumor progression and metastatic potential. IntJ Cancer 64:182-188, 1995 19. Joensuu H, Klemi PJ, Toikkanen S, et al: Glycoprotein CD44 expression and its association with surviwal in breast cancer. Am J Pathol 143:687-874, 1993

296

20. Matsumura Y, Tarin D: Significance of CD44 gene products for cancer diagnosis and disease evaluation. Lancet 340:1053-1058, 1992 21. Abbasi AM, Chester KA, Talbot IC, et al: CD44 is associated with proliferation in normal and neoplastic human colorectal epithelial cells. EurJ Cancer 29:1995-2002, 1993 22. Iida N, Bourguignon LYW: New CD44 splice variants associated with human breast cancers. J Cell Physiol 162:127-133, 1995 23. Kaufmann M, Heider KH, Sinn HP, et al: CD44 variant exon epitopes in primary breast cancer and length of survival. Lancet 345:615-619, 1995 24. Friedrichs K, Kggler G, Franke F, et al: CD44 isoforms in prognosis of breast cancer. Lancet 345:1237, 1995 25. Dall P, Heider KH, Sinn HP, et al: Comparison of immunohistochemistry and RT-PCR for detection of CD44v-expression: A new prognostic factor in human breast cancer. I n t J Cancer 60:471-477, 1995 26. Bloom HJG, Richardson WW: Histological grading and prognosis in breast cancer. BrJ Cancer 11:359-377, 1957 27. Le Doussal V, Tubiana-Hulin M, Friedman S, et al: Prognostic value of histologic grade nuclear components of Scarff-BloomRichardson (SBR): An improved score modification based on a multivariate analysis of 1262 invasive ductal breast carcinomas. Cancer 64:1914-1921, 1989 28. Galea MH, Blarney RW, Elston CE, et al: The Nottingham Prognostic Index in primary breast cancer. Breast Cancer Res Treat 22:20%219, 1992 29. Charpin C, Martin P, Devictor B, et al: Multiparametric study (SAMBA 200) of estrogen receptor immunocytochemical assay in 400 human breast carcinomas analysis of estrogen receptor distribution heterogeneity in tissues and correlations with dextran coated charcoal assays and morphological data. Cancer Res 48:1578-1586, 1988 30. Charpin C, Jacquemier J, Andrac L, et al: Multiparametric analysis (SAMBA 200) of the progesterone receptor immunocytochemical assay in nonmalignant and malignant breast isorders. Am J Pathol 132:199-211, 1988 31. Charpin C, Andrac L, Vacheret H, et al: Multiparametric evaluation (SAMBA) of growth fraction (monoclonal Ki67) in breast carcinoma tissue sections. Cancer Res 48:4368-4374, 1988 32. Charpin C, Bonnier P, Devictor B, et al: Immunodetection of Her-2/neu protein in frozen sections evaluated by image analysis: Correlation with overall and disease-free survival in breast carcinomas. Anticancer Res 13:603-612, 1993 33. Charpin C, Devictor B, Andrac L, et al: p53 Quantitative immunocytochemical analysis in breast carcinomas. HuM PATHOL 26:159-166, 1995 34. Charpin C, Vielh P, Duffaud F, et al: Quantitative immunocytochemical assays of P-glycoprotein in breast carcinomas: Correlation to messenger RNA expression and to immunohistochemical prognostic indicators. J Natl Cancer Inst 86:1539-1545, 1994 35. Charpin C, Pellissier JF: Marqueurs mol~culaires dans les cancers du sein: Aspect pratique et fivaluation morphologique. Bull Acad Natl Mild 178:475-493, 1994 36. Grogan TM, Casey TT, Miller PC: Automation of immunohistochemistry. Adv Pathol Lab Med 6:253-283, 1993 37. Brugal G, Garbay C, Giroud-Adhel D: A double scanning microphotometer for image analysis: Hardware, software and biomedical application.J Histochem Cytochem 27:144, 1979