The adamalysin-thrombospondin (ADAMTS) proteinases are a relatively newly described branch of the metzincin family that contain metalloproteinase, disintegrin, and thrombospondin motifs. They have been implicated in various cellular events, including cleavage of proteoglycans, extracellular matrix degradation, inhibition of angiogenesis, gonadal development, and organogenesis. However, in many cases, their normal physiological roles and their potential for dysregulation in malignancy remain to be established. The expression profile of ADAMTS1–20 in human breast carcinoma was undertaken by real-time PCR using RNA isolated from malignant tumors, nonneoplastic mammary tissue, and breast cancer cell lines to identify altered regulation that may have potential pathogenetic and prognostic significance. Our studies show that seven of the ADAMTS genes (ADAMTS1, 3, 5, 8, 9, 10, and 18) are consistently down-regulated in breast carcinomas with respect to nonneoplastic mammary tissue, irrespective of the heterogeneity of the samples and the tumor type or grade (Mann-Whitney U test, P < 0.0001 for each gene). Conversely, ADAMTS4, 6, 14, and 20 are consistently up-regulated in breast carcinomas (P = 0.005, P < 0.0001, P = 0.003, and P = 0.001, respectively). ADAMTS2, 7, 12, 13, 15, 16, 17, and 19 show no significant difference between the sample types. ADAMTS1, 2, 7, 8, 10, and 12 are expressed predominantly in stromal fibroblasts. ADAMTS3, 4, 5, 6, 9, and 13–20 inclusive are expressed predominantly in myoepithelial cells; all appear to be relatively poorly expressed in luminal epithelial cells. ADAMTS15 has emerged as being an independent predictor of survival, with RNA expression levels significantly lower (P = 0.007) in grade 3 breast carcinoma compared with grade 1 and 2 breast carcinoma.
The adamalysin-thrombospondin (ADAMTS) proteinases are a family of metalloproteinases with functions in extracellular matrix processing, organogenesis, and hemostasis (for review, see Ref. 1 ). To date, there are 19 published ADAMTS genes, numbered 1–10 and 12–20. The proteins they encode all comprise an NH2-terminal signal peptide, a prodomain, and metalloproteinase, disintegrin, cysteine-rich, and thrombospondin type 1 (TSP1) domains, followed by a spacer region and a variable number of TSP1-like motifs (for an overview of the domain organization, see Ref. 2 ); in some cases, there is a COOH-terminal protease and lacunin domain (3) or a complement subcomponent/embryonic sea urchin protein/bone morphogenic protein domain (4) . Unlike the adamalysins, the ADAMTSs are not membrane-anchored but bind to extracellular matrix components such as heparin and heparan sulfate after their secretion (1 , 5) ; this binding is mediated through their TSP1 motifs (6) .
The prototype, ADAMTS-1, was first described by Kuno et al. (6) and isolated on the basis of elevated expression in adenocarcinomas that induce cachexia in mice. It is an effective inhibitor of angiogenesis (7) and has a role in follicular rupture at ovulation (8) and urogenital organogenesis, as evidenced by the phenotype of ADAMTS1 knockout mice (9) . ADAMTS-2, -3, and -14 are all procollagen N-proteinases, involved in the processing of procollagen I (ADAMTS-2 and -14) and procollagen II (ADAMTS-2 and -3) to collagen by cleavage of the NH2 terminus propeptides (10, 11, 12, 13) . Deficiency in the ADAMTS2 gene causes Ehlers-Danlos syndrome type VIIc in humans (14) ; ADAMTS2 knockout mice develop severe skin fragility, and males are sterile, suggesting an additional role for ADAMTS-2 in spermatogenesis or fertility (15) . Five of the ADAMTSs have been identified as aggrecanases, which are pivotal for cartilage degradation in osteoarthritis: ADAMTS-1 (16) ; ADAMTS-4 and ADAMTS-5/TS-11 (17 , 18) ; ADAMTS-9 (19) ; and ADAMTS-15 (20) . ADAMTS6 and 7 have been described, but their protein function remains uncharacterized (21) . ADAMTS-8, like ADAMTS-1, is a potent inhibitor of angiogenesis (7) and was also described independently by Georgiadis et al. (22) . ADAMTS-9 is thought to function as an antiangiogenic agent (23) , and both it and ADAMTS-20 may possibly have a role in gonadal development because of its high sequence homology to the Caenorhabditis elegans ADAMTS gene gon-1 (19 , 23) . ADAMTS12 transcripts have been isolated in gastric carcinoma and various tumor cell lines, suggesting a role in tumor processes (24) . ADAMTS-13, which has been described very recently, has been characterized as the blood clotting protein von Willebrand factor cleaving protease (4 , 25, 26, 27) . Another recent paper describes seven novel ADAMTS genes, including ADAMTS15–19 (2) .
Like matrix metalloproteinases (MMPs), certain ADAMTSs have matrix-degrading activities and might therefore be involved in pathological tissue remodeling and cell invasion. At least two family members have antiangiogenic capabilities; others may do so via their TSP1 motifs. Therefore, it is important to explore possible links between ADAMTS genes and cancer, particularly because ADAMTSs are potential targets of synthetic metalloproteinase inhibitors that have entered clinical cancer trials (for review, see Ref. 28 ). In this study, real-time PCR was used to evaluate the expression profile of ADAMTS1–20 in a series of human breast carcinomas, nonneoplastic mammary tissue, and purified populations of mammary tissue cell types to assess evidence of altered regulation with potential pathogenetic and prognostic significance.
MATERIALS AND METHODS
This study received the approval of the Norwich District Ethics Committee, the Norwich Research and Development Committee, and the Partners in Cancer Research Tissue Bank Committee.
Human Breast Tumor Samples.
Forty-eight women ages 40–88 years (median age, 58 years) provided samples of 50 primary breast neoplasms after informed consent. Two patients who had bilateral disease gave two samples, one from each side. Forty-eight of the neoplasms were invasive carcinomas, and two were high-grade ductal carcinoma in situ, one of which showed foci of microinvasion. All samples were taken from fresh, unfixed carcinomas removed surgically. The specimens were sent immediately after removal to the laboratory, where, after inking the margins and slicing the specimen, a small part of the tumor was removed by a histopathologist for this study. Specimens were prepared and snap-frozen in the department’s human tissue bank as described by Riddick et al. (29) . The rest of the surgical specimen was fixed in 10% formal saline and processed for routine histopathological diagnosis and establishment of prognostic factors.
Nonneoplastic Human Mammary Tissue.
Samples of nonneoplastic mammary tissue were obtained from 10 patients with informed consent at reduction mammoplasty. All samples were snap-frozen and stored as described for the cancers. An additional sample was obtained commercially as total RNA from normal human mammary tissue (Stratagene, La Jolla, CA).
Purified Samples of Nonneoplastic Human Mammary Cell Populations.
Normal breast tissue was obtained from reduction mammoplasty procedures following consent. Primary myoepithelial, luminal epithelial, and fibroblast populations were isolated according to a modified version of the protocol described by Gomm et al. (30) . Tissue was digested as described by Stampfer et al. (31) in 200 units/ml collagenase type IA and 125 units/ml hyaluronidase in DMEM containing 10% fetal bovine serum, 100 IU/ml penicillin, 100 μg/ml streptomycin, and 2.5 μg/ml fungizone for 12–18 h. The resulting mixed organoid and single cell preparation was subjected to a single sedimentation step of 30 min at 4°C, and the supernatant, rich in fibroblasts, was removed and passed sequentially through 2 × 40-μm filters and then transferred to coverslips for characterization. Organoids were washed in Enriched Medium (1:1 DMEM:Ham’s F-12 medium, 10% fetal bovine serum, 5 μg/ml insulin, 5 μg/ml hydrocortisone, 20 ng/ml epidermal growth factor, 100 IU/ml penicillin, 100 μg/ml streptomycin, and 2.5 μg/ml fungizone; all from Sigma) and then digested in 20 ml of trypsin/EDTA containing 400 μl of DNase, incubated at 37°C for 10 min. After further filtration, the single cell suspension was resuspended in Enriched Medium for isolation of myoepithelial cells or Luminal Medium (Enriched Medium lacking epidermal growth factor) for isolation of the luminal population. Myoepithelial cells were selected using Pan-Mouse IgG Dynabeads (Dynal United Kingdom Ltd., Bromborough, United Kingdom) labeled with CALLA (CD10; Dako), as described previously (32) . Luminal cells were isolated using the MiniMACS magnetic cell sorting system (Miltenyi Biotec Ltd.), according to the manufacturer’s instructions. Briefly, cells were incubated with antibody to epithelial membrane antigen [EMA (clone E29; Dako)] for 10 min at 4°C with rotation. After washing in EDTA/PBS, the cells were incubated with MACS goat antimouse IgG microbeads in EDTA/PBS for 15 min at 4°C and then passed through a MiniMACS column. After several washes to elute the nonlabeled fraction, the column was removed from the magnet, and the labeled cell fraction was eluted.
Characterization of Isolated Cell Populations by Immunohistochemistry.
Aliquots of isolated cells were plated onto poly-d-lysine-coated coverslips and grown for 48 h in appropriate media. The cells were then fixed in acetone for 10 min at 4°C, and immunohistochemistry was performed using antibodies to cytokeratin (CK) 14 and CK18 (LL001 and LE061, respectively; both received as gifts from Prof. E. B. Lane; Cancer Research UK Cell Structure Research Laboratories, Dundee, Scotland, United Kingdom), EMA, and vimentin (Dako) with a standard streptavidin/biotin detection system, as described previously (32) . For analysis of RNA from isolated mammary cells to assess cell purity using primers for the desmosomal glycoproteins Dsg2 and Dsg3, see Refs. 32 and 33 .
Human Breast Cancer Cell Lines.
The breast cancer cell lines MCF-7, T47D, MDA-MB-231, and BT549 were obtained from the American Type Culture Collection (Manassas, VA) and cultured as directed. All media and supplements were obtained from Life Technologies, Inc. (Paisley, United Kingdom).
Total RNA was isolated using a modification of the SV Total RNA Isolation System (Promega, Madison, WI; Ref. 34 ): 50–100 mg of tumor tissue were homogenized in 1 ml of RNazol B reagent (Biogenesis Ltd., Poole, United Kingdom) using an UltraTurrax T8 homogenizer (IKA). The homogenate was stored at −80°C pending the following step of the protocol: the samples were allowed to thaw completely at room temperature, then centrifuged at 14,000 rpm for 10 min and the clear supernatants were collected into 200 μl of chloroform. They were shaken vigorously for 15 s, incubated for 3 min at room temperature, and centrifuged for 15 min at 14,000 rpm. The upper phase was collected into 200 μl of 95% ethanol, mixed, and transferred into a spin basket assembly. The Promega protocol was then followed from step 7 to the end according to the manufacturer’s instructions.
One μg of total RNA was primed with 0.5 μg of random hexameric primers (Promega) and reverse transcribed into cDNA in a 20-μl reaction volume containing 200 units of SuperScript II reverse transcriptase, 5× first-strand buffer [final concentrations: 50 mm Tris-HCl (pH 8.3), 75 mm KCl, and 3 mm MgCl2], 10 mm DTT, 0.5 mm of each deoxynucleotide triphosphate (dATP, dCTP, dGTP, and dTTP; Life Technologies, Inc.), and 40 units of RNasin RNase inhibitor (Promega). Each priming reaction was carried out at 70°C for 10 min and stopped by placing samples on ice; the reverse transcription was carried out at 42°C for 1 h followed by incubation at 70°C for 10 min.
The integrity of the RNA in each sample was checked by PCR amplification of the reverse-transcribed cDNA as described above, using primers for the glyceraldehyde-3-phosphate dehydrogenase gene. The sequence of the forward primer was 5′-CGGAGTCAACGGATTTGGTCGTAT-3′ (nucleotides 78–101), and that of the reverse primer was 5′-AGCCTTCTCCATGGTGGTGAAGAC-3′ (nucleotides 384–361) to give a 307-bp amplicon (Marc Lafleur; University of East Anglia, Norwich, United Kingdom). The reactions contained 100 ng of cDNA, 20 pmol of each primer, 1× PCR buffer [final concentrations: 67 mm Tris-HCl (pH 8.8), 16 mm (NH4)2SO4, and 0.01% Tween 20], 1.5 mm MgCl2, 2.5 units of Taq DNA polymerase (Bioline, London, United Kingdom), and 0.2 mm of each deoxynucleotide triphosphate (dATP, dCTP, dGTP, and dTTP; Life Technologies, Inc.) in a 50-μl reaction volume. The amplification reaction was carried out over 35 cycles with the following parameters: 94°C for 60 s; 60°C for 60 s; and 72°C for 90 s. The 72°C step was extended to 10 min for the final cycle to obtain full-length PCR products.
TaqMan Real-Time PCR.
Forward and reverse primers and fluorescence-labeled oligonucleotide probes [using 6-carboxy-fluorescein (FAM) as the reporter dye and 6-carboxy-tetramethyl rhodamine (TAMRA) as the quencher dye] were designed for each of the human ADAMTS genes and the c-erbB-2 gene using Primer Express 1.0 software (PE Applied Biosystems, Foster City, CA) and synthesized by PE Applied Biosystems (Warrington, United Kingdom). The primer and probe sequences for MMP-9 and -11, designed in our laboratory, are now the intellectual property of Applied Biosystems. The primer and probe set for cyclin D1 was obtained directly from Applied Biosystems (as before). BLASTN searches were undertaken on all our designed primer and probe sequences to confirm gene specificity. To ensure against genomic DNA amplification, the forward and reverse primers were placed in separate adjacent exons. Since the genomic sequences of the human ADAMTS genes were not published at the time of primer design, the mouse Adamts1 genomic sequence (35) was used to predict their intron-exon boundaries using the CLUSTALW program (Biology Workbench 3.2).6 The correct identity of the PCR products obtained was confirmed by direct sequencing of the amplicons. Details of the primers and probes are given in Table 1⇓ . For each human ADAMTS gene, a standard curve with concentrations ranging from 25 ng to 1 ng was produced using human placental cDNA dissected from fetal villi as the template. An XY scatter plot was for each gene was produced using Microsoft Excel Chart Wizard software (Microsoft Office 2001), and values for the equation y = mx + b (where m = the slope of the standard curve and b = the y intercept of that line) and R2 were obtained. The 18S rRNA gene was used as an endogenous control to normalize for differences in the amount of total RNA in each sample. The cyclin D1 gene was also used as a further control to confirm that differences in gene expression between sample types were not due merely to the proliferation rate of the tumors. PCR reactions for all samples were performed in duplicate in 96-well optical plates with 5 ng of cDNA (1 ng of cDNA for the 18S gene), 100 nm probe, 200 nm each primer, and 12.5 μl of TaqMan Universal 2× PCR Master Mix (PE Applied Biosystems, Warrington, United Kingdom) in a 25-μl reaction volume. The amplification reaction was carried out over 40 cycles with the following parameters: an initial holding stage of 2 min at 50°C and then 10 min at 95°C, followed by a two-step cycling program of 15 s at 95°C and 1 min at 60°C.
The data were found not to be normally distributed (Kolmogorov-Smirnov test; data not shown). Significant differences in the normalized input values, representing RNA expression levels, of invasive carcinoma and normal mammary tissue for each ADAMTS gene were determined using the Mann-Whitney U test. Correlations between the normalized input values of the invasive carcinoma samples and various clinical characteristics were undertaken using Spearman’s rank correlation. Kaplan-Meier survival functions were plotted to evaluate the expression levels of each ADAMTS gene from the normalized input values of the invasive carcinoma samples against event-free survival for those patients. Pvalues < 0.05 were deemed statistically significant in all analyses. All data were analyzed using SPSS 11.0; the normalized input values are in arbitrary units.
Expression levels for ADAMTS RNA were analyzed in invasive breast carcinoma (n = 50), nonneoplastic mammary tissue (n = 11), and breast cancer cell lines (n = 4). Normalized input values are presented graphically in Fig. 1⇓ , and the grouped median values of the three sample groups for each gene are given in Table 2⇓ . Comparison of RNA expression levels in invasive breast carcinomas with those of nonneoplastic mammary tissue showed that gene expression levels of ADAMTS1, 3, 5, 8, 9, 10, and 18 were significantly lower in the malignant samples (P < 0.0001 for each gene). By contrast, the same evaluation showed that expression of ADAMTS4, 6, 14, and 20 was significantly higher in the malignant samples as compared with normal mammary tissue (P = 0.005, P < 0.0001, P = 0.003, and P = 0.001, respectively). No significant difference was seen between the two sample types for ADAMTS2, 7, 12, 13, 15, 16, 17, or 19.
To define the cellular origins of the ADAMTS genes expressed in normal mammary gland, we isolated fibroblasts, epithelial and myoepithelial cells using a combination of collagenase digestion and affinity separation procedures. Immunohistochemistry demonstrated that CALLA-extracted cells were uniformly positive for CK14 and vimentin but negative for CK18 and EMA, in keeping with a myoepithelial phenotype. EMA-isolated cells were positive for CK18 and EMA but negative for CK14 and vimentin, consistent with a luminal epithelial phenotype, and fibroblasts displayed uniform positivity for vimentin but were negative with the other markers. The reverse transcription-PCR analysis on isolated populations showed that EMA-isolated cells expressed DSg2 but not DSg3, whereas CALLA-isolated cells expressed both genes, consistent with the myoepithelial cell-restricted expression of DSg3 reported previously (33) . Fibroblasts did not express either desmosomal gene. Together, these results indicate substantially pure populations of different breast cell types.
Gene expression localization studies using the sorted cell populations from pooled nonneoplastic mammary tissue showed that expression levels of ADAMTS1, 2, 7, 8, 10, and 12 were highest in stromal fibroblasts, whereas expression levels of ADAMTS3, 4, 5, 6, 9, and 13–20 inclusive were highest in myoepithelial cells. All of the ADAMTS genes appeared to be relatively poorly expressed in luminal epithelial cells (Fig. 2)⇓ .
Whereas the data in Fig. 1⇓ provide a quantitative comparison between samples for each ADAMTS gene normalized to 18S RNA levels, the analyses did not indicate the relative expression levels of each gene. To resolve this, cycle threshold results and their ranges are also presented in Table 2⇓ . These data indicate that, of the gene family, ADAMTS1, 5, and 15 show the highest levels of expression.
To characterize further the tissues that were analyzed in this study, expression analysis of several genes that have previously been reported to be linked with malignancy in mammary neoplasms was undertaken. The results obtained for MMP-9, MMP-11, and c-erbB-2 (Fig. 1)⇓ were consistent with previously published expression patterns for these genes in both invasive breast carcinoma and nonneoplastic mammary tissue (36, 37, 38, 39, 40) . All three genes showed significant overexpression at the RNA level in carcinomas compared with nonneoplastic mammary tissue. A control analysis using cyclin D1 as a proliferation marker of mammary neoplasia showed that the altered regulation seen with the ADAMTS genes was independent of tumor cell proliferation (data not shown).
Table 3⇓ summarizes the clinical and pathological features of patients with invasive breast carcinoma, and the clinical features of the nonneoplastic breast tissue patients are given in Table 4⇓ . Possible correlations between RNA expression levels in invasive carcinoma and tumor grade, estrogen receptor status (Spearman’s rank correlation), and differences with respect to axillary lymph node involvement (Mann-Whitney U test) were evaluated for each ADAMTS gene. ADAMTS15 expression levels were significantly lower in patients with grade 3 tumors compared with grade 1 or grade 2 tumors (P = 0.007). Analysis of event-free survival of the invasive breast carcinoma patient set was undertaken for each ADAMTS gene using Kaplan-Meier survival functions. The median normalized input value for each gene was used to divide the patients into two groups: those with values at or above the median versus those with values below the median. This showed that only ADAMTS15 expression in the tumors was an independent predictor of survival (P = 0.0024; shown graphically in Fig. 3⇓ ). When combined with Nottingham Prognostic Index values, prediction of those patients who have died or relapsed increased from 17% to 75% (data not shown). Survival data of the patients with breast carcinoma are presented in Table 5⇓ .
The data were also assessed by coexpression analysis to look for potential patterns of coregulation of the genes in mammary tissues. The mRNA expression of each ADAMTS gene in the entire tissue collection (i.e., with malignant and nonneoplastic cases grouped, but excluding cell lines) was compared with each of the other ADAMTSs and with MMP-9, MMP-11, and c-erbB-2. All statistically significant correlations are shown in Fig. 4⇓ . This analysis revealed particular ADAMTS gene pairs that showed strong association of their expression. These pairs can be seen to form two principal groupings that show positive pairwise correlations of mRNA expression at significance levels of P < 0.0001: the first set includes ADAMTS1, 3, 5, 8, 9, 10, and 18; and the second set comprises ADAMTS2, 4, 6, 7, 12, and 14. Expression of c-erbB-2 did not correlate strongly (i.e., P < 0.0001) with any of the other genes assessed, but MMP-11 expression showed strong negative correlation with ADAMTS1, 5, 8, and 9.
This study is the first detailed investigation of the ADAMTS genes in human breast cancer. The most striking observation is that 11 of the 19 ADAMTS genes studied show altered expression in malignant compared with nonneoplastic tissue. ADAMTS1, 3, 5, 8, 9, 10, and 18 are down-regulated in human breast carcinoma; ADAMTS4, 6, 14, and 20 are up-regulated; and ADAMTS2, 7, 12, 13, 15, 16, 17, and 19 show no significant difference in mRNA expression levels between malignant and nonneoplastic tissue. Whereas all of the ADAMTSs that are up-regulated in breast carcinoma are expressed predominantly in myoepithelial cells in nonneoplastic mammary tissue, those that are down-regulated are not confined to one cell type in nonneoplastic mammary tissue; some are expressed predominantly in myoepithelial cells, and some are expressed predominantly in stromal fibroblasts. It could be argued that genes that show a myoepithelial-restricted expression pattern in nonneoplastic tissue could be subject to down-regulation in tumors due to the absence of a myoepithelial population in invasive carcinomas. Whereas we cannot rule this out for genes such as ADAMTS3 and 18 (and possibly ADAMTS9), it is clearly not the case for the other largely myoepithelial-restricted genes, including ADAMTS4, which is profoundly up-regulated in tumors. Thus, it is clear that both the level and location of expression of several members of the ADAMTS gene family become dysregulated during mammary tumorigenesis.
Although members of the ADAMTS family show on average only between 20% and 40% homology to each other (1) , two groups, ADAMTS2, 3, and 14 and ADAMTS4 and 5, share very similar functions and substrates (13 , 18) . It was anticipated that these groups of genes might display the same overall trend with respect to relative gene expression levels and distribution in breast tissue, but this was shown not to be the case. Instead, pairwise analysis of expression levels provided evidence for potential coregulation of ADAMTS1, 3, 5, 8, 9, 10, and 18, all of which were down-regulated in cancer specimens. There was also evidence for coexpression of ADAMTS2, 4, 6, 7, 12, and 14. None of the ADAMTS genes showed coexpression with c-erbB-2, which was overexpressed in 22% of carcinoma samples, consistent with previous reports (36 , 37) . However, several ADAMTS genes that are down-regulated in malignancy showed significant negative correlation of expression with that of MMP-11, a protease that is up-regulated in stromal fibroblasts in mammary carcinomas (38) . Thus, there may be some common mechanisms that achieve reciprocal regulation of the ADAMTS genes and MMP-11.
The predominance of myoepithelial cell expression of the ADAMTSs suggests a possible role for some of these genes in the control of basement membrane remodeling in the normal breast. Furthermore, it has been reported that myoepithelial cells exhibit an antiangiogenic phenotype (41) , and the expression of ADAMTSs may contribute to this. Because ADAMTS proteins have catalytically intact metalloproteinase domains, disintegrin domains (that may interact with integrins), and regions resembling the antiangiogenic thrombospondins, their contributions to tumorigenesis may be complex. Indeed, the best-characterized family member, ADAMTS-1, is an aggrecanase (16) that has also been shown to inhibit endothelial cell proliferation (7) . In the corneal micropocket and chorioallantoic membrane assays, both ADAMTS-1 and ADAMTS-8 display antiangiogenic activities that are more potent than that of thrombospondin-1 at the same molar concentration (7) .
As with the situation in breast cancer, the expression of ADAMTS1 has also been found to be lower in pancreatic cancer compared with noncancerous tissue (42) . However, among the pancreatic cancer cases, those with higher levels of ADAMTS1 showed poorer prognosis, with evidence of increased local invasion and lymph node metastasis, suggesting that ADAMTS1 may play different roles in the normal organ compared with its involvement in tumor progression. There was no association between ADAMTS1 and microvessel density in the pancreatic cancer cases, and ADAMTS8 was expressed at very low levels in both normal and cancer tissue, indicating that neither gene is closely linked with regulation of angiogenesis at this site. Analysis of clinical data revealed no strong links of ADAMTS1 expression with clinicopathological features of the breast cancer cases in our study set; however, interestingly, ADAMTS15 expression showed correlations with tumor grade and also correlated with event-free survival: normalized input values below 0.5 identify a subset of patients with a poorer prognosis; and values above 1.0 suggest a better prognosis. When evaluated together with two of the standard prognosticators, tumor grade and size, 75% of those patients who have died or relapsed were identified as falling into a poor risk category, compared with the Nottingham Prognostic Index, which (where available) identified only 17% of cases who have died or relapsed as poor risk.
The regulation of the ADAMTS genes in breast is not known. However, it might be reasonable on present evidence to speculate that progesterone may have some regulatory effect. One study has shown that ADAMTS1 is a direct target of progesterone receptor-mediated activation in murine ovarian granulosa cells in follicles destined to ovulate, suggesting a possible role for ADAMTS1 in follicular rupture (8) . TSP1 itself has been shown to be regulated by progesterone in the human endometrium (43) . It is possible, therefore, that certain ADAMTS genes expressed in mammary tissue may also be progesterone-regulated.
A further speculation is that some ADAMTS genes that are down-regulated in breast cancer may have been silenced by methylation, although this remains to be confirmed. Although two ADAMTS genes that are down-regulated in breast carcinoma, ADAMTS8 and ADAMTS9, map to two regions known to be frequently deleted in breast cancer and other malignancies [11q25 and 3p14.2, respectively (44 , 45)] , we tested one gene, ADAMTS8, and found no evidence of loss of heterozygosity in several breast carcinoma cell lines assayed by Southern blotting and no evidence of a gross deletion of the gene locus by cytogenetic G-banded analysis.7 The cytogenetic evaluation is also relevant with regard to ADAMTS15, approximately 20 kb distal to ADAMTS8 at 11q25. Several authors postulate the existence of a hitherto undiscovered tumor suppressor gene at 11q25 in breast carcinoma (44 , 46 , 47) . These observations, coupled with our findings of marked down-regulation of ADAMTS8 in human breast carcinoma, indicate the value of further investigations to demonstrate whether or not ADAMTS8 or 15 functions as a tumor suppressor gene in mammary tissue.
In summary, our data demonstrate for the first time that there is down-regulation at the level of RNA expression of several members of the ADAMTS family in breast cancer compared with nonneoplastic mammary tissue. This is of potential interest in the context of breast carcinomas because they are frequently vascular tumors, and their angiogenic index, or ratio of microvessels to tumor volume or tumor cell number, has emerged as a reliable indicator of early metastasis and thereby prognosis, especially in small tumors (48 , 49) . Thus, disease progression in breast cancer likely involves suppression or abrogation of antiangiogenic molecules, and the down-regulation of ADAMTS1 and 8 in breast carcinoma may be significant. We note that, at the end of their paper on endogenous regulators of angiogenesis, Carpizo and Iruela-Arispe (50) mention their unpublished work on ADAMTS1 and breast carcinoma. Human T47D breast carcinoma cells, stably transfected with ADAMTS1 cDNA and injected s.c. into nude mice, show marked inhibition of tumor growth compared with tumors derived from control T47D cells transfected with an empty vector and concurrently injected. We await their further observations with interest, especially because they correspond closely with our ongoing studies of ADAMTS1 and 8 in human breast carcinomas.
We thank Nick Palmer and Gavin Pettigrew (Specialist Registrars to Stuart D. Scott) and Paul Morris (Specialist Registrar to Elaine M. Sassoon) for assistance in providing breast tissue for this study. We are grateful for the additional support of the Norwich and Norfolk Big C Appeal and the Medical Research Council.
Grant support: Grant QLG1–2000-01131 from the European Union Framework Programme 5 and Grant 2001:232 from the Breast Cancer Campaign.
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 with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Dylan R. Edwards, School of Biological Sciences, University of East Anglia, Norwich NR4 7TJ, United Kingdom. Phone: 44-1603-592184; Fax: 44-1603-592250; E-mail:
↵7 S. Porter, A. Hogan, and M. Weber, unpublished observations.
- Received March 6, 2003.
- Revision received January 8, 2004.
- Accepted January 8, 2004.