Imaging, Diagnosis, Prognosis

Analysis of Integrin B4 Expression in Human Breast Cancer: Association with Basal-like Tumors and Prognostic Significance Shaolei Lu,1Karl Simin,1Ashraf Khan,2 and Arthur M. Mercurio1

Abstract Purpose: The h4 integrin has been implicated in functions associated with the genesis and progression of carcinomas based on data obtained from cell lines and mouse models. Data on its expression and relevance to human carcinomas, however, are relatively scant. The aim of this study was to assess its expression and prognostic significance in human breast carcinomas. Experimental Design:We integrated data on h4 expression from multiple profiling studies of breast tumors of known clinical outcome with immunohistochemical analysis of 105 breast carcinomas, and we identified whose expression correlates with that of h4. Results: The expression of both h4 mRNA and is not homogeneous in breast cancer and it associates most significantly with the ‘‘basal-like’’ subtype of breast tumors (P = 0.008). No association between h4 and HER2 expression was evident from either gene profiling or immunohistochemical analysis. To gain insight into the relevance of h4expressiontohuman breast carcinomas, we generated a 65-gene‘‘h4 signature’’ based on integration of four published gene profiling studies that included the top 0.1% of genes that correlated with h4, either positively or negatively. This h4 signature predicted decreased time to tumor recurrence and survival of patients when applied to four data sets including two independent ones. Conclusions: These observations indicate that h4 expression in human breast cancer is restrict- ed and associated with basal-like cancers, and they support the hypothesis that h4mayfunction in concert with a discrete set of to facilitate the aggressive behavior of a subset of tumors.

The a6h4 integrin (referred to as ‘‘h4’’ because there is only h4 can regulate key signaling pathways associated with one h4 integrin) is expressed primarily in epithelial cells and in carcinoma progression and facilitate the migration, invasion, a few other cell types and is defined as an adhesion receptor for and survival of carcinoma cells (3, 13, 14). More recently, h4 most of the known laminins (1). The distinguishing structural has been implicated in the genesis of squamous and breast feature of h4 is the atypical intracellular domain of the h4 carcinomas using mouse models (9, 15, 16). A seminal finding subunit, which is distinct both in size (f1,000 amino acids) is that h4 is mobilized from hemidesmosomes as a conse- and structure from any other integrin subunit (2). A major quence of carcinoma progression and it localizes in filamentous function of this intracellular domain is to link h4to protrusions where it facilitates migration and invasion intermediate filaments in hemidesmosomes (3). Although the and functions as a signaling receptor (6, 10, 14, 17). This involvement of h4 in hemidesmosome organization and dichotomy of h4 function is summarized best by the hypo- function dominated the study of this integrin, our lab thesis that h4 switches from a mechanical adhesive device into pioneered studies that established that this integrin plays a a signaling-competent receptor during the progression from pivotal role in functions associated with carcinoma progression normal epithelium to invasive carcinoma (10, 13). (4–9), a role that is now widely accepted (3, 10–12). In brief, Given that the data implicating h4 in functions associated with breast and other carcinomas were obtained largely from studies on carcinoma cell lines and some mouse models, a critical issue that needs to be addressed is the relevance of these Authors’ Affiliations: Departments of 1Cancer Biology and 2Pathology, and data to human carcinomas. Surprisingly, relatively few studies Cancer Center, University of Massachusetts Medical School, Worcester, have assessed h4 expression in human breast tumors or Massachusetts Received 9/4/07; revised 10/26/07; accepted 10/29/07. correlated its expression with clinical outcome. One study that Grant support: NIHgrantCA80789andSusanG.KomenBreastCancer did address this issue using in situ hybridization observed a Foundation grant PDF0600265. correlation of h4 mRNA expression with tumor size and grade The costs of publication of this article were defrayed in part by the payment of page (18). To gain more insight into this important issue, we advertisement charges. This article must therefore be hereby marked in accordance integrated data from gene profiling studies of breast tumors of with 18 U.S.C. Section 1734 solely to indicate this fact. Note: Supplementary data for this article are available at Clinical Cancer Research known clinical outcome with immunohistochemical analysis of Online (http://clincancerres.aacrjournals.org/). h4 expression. The results obtained reveal that h4 expression is Requests for reprints: Arthur M. Mercurio, Department of Cancer Biology, not homogeneous in breast cancer and that it correlates most University of Massachusetts Medical School, LRB-408, 364 Plantation Street, significantly with the ‘‘basal-like’’ subtype of breast tumors Worcester, MA 01605. Phone: 508-856-8676; Fax: 508-856-1310; E-mail: @ (19). Moreover, cluster analysis of genes whose expression arthur.mercurio umassmed.edu. h h F 2008 American Association for Cancer Research. correlates with 4 generated a 4 ‘‘gene signature’’ that is doi:10.1158/1078-0432.CCR-07-4116 prognostic for tumor recurrence and decreased survival time.

Clin Cancer Res 2008;14(4) February 15, 2008 1050 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Expression and Prognostic Value of b4Integrin

with a SD of 1.5; 51% of the tumors had local vessel invasion; 59% Materials and Methods had local lymph node metastases; 61% had necrotic foci; 47% were estrogen receptor positive (ER+); 61% were progesterone receptor Tumor samples. A total of 105 cases of invasive ductal breast positive (PR+); and 35% were HER2+. All of the slides were reviewed by carcinomas were diagnosed in the Department of Pathology at the a pathologist (A.K.) to confirm the diagnosis. University of Massachusetts Medical School from 1997 to 2004. Immunohistochemistry. Two monoclonal integrin h4 antibodies Clinical history and pathology reports on these cases were reviewed were used [a rat antibody 439-9b from Dr. Rita Falcioni, (Regina Elena to obtain the clinicopathologic information. All of the tumors were Cancer Institute, Rome, Italy) and a mouse antibody Elf-1 from invasive ductal cancers. The mean patient age at diagnosis was 59 years Novocastra]. The 439-9b antibody was used at a final concen- (range, 29-94 years); the mean size of the primary tumors was 2.6 cm tration of 2 Ag/mL and Elf-1 antibody was used at 1:50 dilution. The

Fig. 1. Analysis of h4 expression in two gene profiling data sets of breast cancer. A and C, h4 expression in basal-like and luminal tumors based on the data from two probe sets of the molecular portraits data (19). B and D, regression analysis assessing the relationship between h4andkeratin5(KRT5), and h4 and HER2 expression using the molecular portraits data. There are two, three, and four probe sets representing h4, 5, and HER2 expression, respectively. E, h4 levels in basal-like and luminal tumors of the combined data set (20). F, regression analysis of integrin h4 and HER2 expression in the combined data set. *, significant difference (P < 0.0001). The values at the X-axis and Y-axis are normalized data. The top and bottom boundaries of the boxes indicate mean (FSE).The mean is denoted by a horizontal line in the box.The whiskers above and below extend to mean (FSD).

www.aacrjournals.org 1051 Clin Cancer Res 2008;14(4) February 15, 2008 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Imaging, Diagnosis, Prognosis

Fig. 2. Integrin h4 staining patterns in breast cancer. A, cell surface staining pattern. B, epithelial-stroma boundary pattern. C, mixed cell surface and cytosolic pattern. D and E, coexpression of h4(D) and basal /6 (E) in an invasive ductal cancer. F, h4 expression in a squamous metaplastic cancer. paraffin-embedded cancer sections were stained using corresponding expression of the h4 signature based on the first two branches of tumor streptavidin-biotin systems (Vectastain Elite avidin-biotin complex kit, clustering. The survival analyses were then done to compare the clinical Vector Laboratories). Antigen retrieval was done by steaming the slides outcome of these two tumor groups. for 30 min in a citrate buffer (pH 6.0; Zymed/Invitrogen). Endogenous peroxidase was quenched by 3% hydrogen peroxide (Sigma-Aldrich). Results Endogenous biotin was blocked using the avidin/biotin blocking system (Vector Laboratories), and casein (Vector Laboratories) was used to dimi- Analysis of gene profiling data reveals that b4 is associated nish nonspecific staining. In addition, keratin 5/6 immunostaining was with basal-like breast cancers. In the original study on the done on all tumors using a mouse monoclonal antibody from Dako- molecular portraits of breast cancer molecular profiling, breast Cytomation at 1:50 dilution and pretreatment with EDTA at pH 8.0. cancers are classified into four subsets, including basal-like and Horseradish peroxidase was developed using 3,3¶-diaminobenzidine luminal tumors, and h4 was identified in one of the two basal tetrahydrochloride (DakoCytomation), and the specimens were counter- gene clusters (19). We reanalyzed the original data and found stained with hematoxylin. that h4 expression in both probe sets is significantly higher Statistical analysis. The Student’s t test was used to compare in basal-like tumors than it is in luminal tumors (both integrin h4 gene expression levels in tumor subtypes whose data were P < 0.0001; Fig. 1A and C). A more recent analysis that fused extracted from the molecular portraits data (19) and the combined data three independent microarray studies into a single data set set (20). Pearson correlation was used to correlate integrin h4 with identified b4 as one of 306 ‘‘intrinsic’’ genes that are common m2 keratin 5 and HER2 expression in tumors. analysis was used to assess to all three microarray studies and that can be used to classify h the associations between integrin 4 and other characteristics. The breast tumor subtypes (luminal A and B, and basal like; survival curves were estimated and compared using Kaplan-Meier h P P ref. 20). We compared 4 expression levels among these dif- estimates and log-rank test. All values were two sided and values of h <0.05 were considered statistically significant. These analyses were ferent tumor subtypes and observed once more that 4 expres- carried out using JMP 6.0 (SAS Institute). sion in basal-like tumors is significantly higher than it is in the Generation of integrin b4 gene signature gene. Four breast cancer luminal subtypes (both P < 0.0001; Fig. 1E). gene profiling data sets (19–22) were reanalyzed using GeneSpring Basal-like tumors are characterized by their expression of basal GX 7.3.1 (Agilent Technologies). Pearson correlations between integrin , including keratins 5, 6, and 14 (25). We calculated the h4 and other genes in these data sets were calculated and ranked. The correlation between h4 and keratin 5 expression in the molecular top 0.1% of genes that correlated positively and the bottom 0.1% of portraits data. In a set of six Pearson correlation studies based on h h genes that correlated negatively with 4 were collected, and 65 4 two h4 probes and three keratin 5 probes (Fig. 1B and D, top), signature genes were identified that are common to at least two data there is a clear positive association between h4 and keratin 5 sets (Table 3). Other h4-related genes that are present in only one data expression (R2 values range from 0.18 to 0.32). set are listed in Supplementary Table S1. To classify tumors based on h4 gene signature, expression data of Studies using breast cancer cell lines and mouse models have h these genes were extracted from published data sets and tumors were shown both a physical and functional association between 4 clustered hierarchically by Cluster 3.0 (23) in an unsupervised way and erbB2 (HER2) expression (16, 26, 27). One inference of (centered average linkage). Java TreeView (24) was used to display the these data is that the expression of these two receptors dendrogram and to identify tumors that exhibited high and low correlates in human breast tumors. Surprisingly, however, there

Clin Cancer Res 2008;14(4) February 15, 2008 1052 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Expression and Prognostic Value of b4Integrin is no correlation in their expression based on data from two h4 h probes and four HER2 probes with R2 values ranging from Table 2. Integrin 4 is associated with basal-like 0.01 to 0.08 (Fig. 1B and D, bottom). In addition, regression tumors analysis of the data retrieved from the combined data set of Classification Integrin B4 P 315 tumors substantiated a lack of correlation between h4 and HER2 (R2 = 0.0245; Fig. 1F). Negative, Positive, n (%) n (%) Immunohistochemical analysis of b4 expression confirms an association with basal-like tumors. To validate our analysis of Luminal A h4 gene expression and confirm that h4 is associated with Yes 27/36 (75) 9/36 (25) 0.19 h No 38/62 (61.29) 24/62 (38.7) basal-like tumors, we assessed 4 protein expression in 105 Luminal B cases of primary invasive ductal breast cancers by immunohis- Yes 22/29 (75.86) 7/29 (24.1) 0.245 tochemistry. Data on the expression of the ER, PR, HER2, and No 43/69 (62.3) 26/69 (37.7) + - 5/6 were available for all of these tumors. Two HER2 /ER h Yes 4/6 (66.7) 2/6 (33.3) 1 different 4 monoclonal antibodies (Elf-1 and 439-9b) were No 61/92 (66.3) 31/92 (33.7) used for the immunohistochemistry that yielded similar Basal like staining patterns. Both monoclonal antibodies stained the Yes 12/27 (44.4) 15/27 (55.6) 0.008 basal layer of skin intensely, as well as the myoepithelium of No 53/71 (74.65) 18/71 (25.35) normal mammary gland (Supplementary Fig. S1). Interestingly, only 32% (34 of 105) of the tumors exhibited h4 staining that NOTE: Classification of breast cancer subtypes was modified from was predominantly localized to the cell surface and seen in Carey et al. (29), which used a panel of immunohistochemical markers to define subtypes obtained from microarray analysis. discrete clusters of cells within tumors (Fig. 2A and B). Some Luminal A tumors are ER+ and/or PR+, HER2-; luminal B tumors h + + + + - - - tumors (6 of 34) exhibited additional cytosolic 4 staining are ER and/or PR , HER2 ; HER2 /ER tumors are ER ,PR,and (Fig. 2C). We noted that cytokeratin 5/6–positive basal-like HER2+; and basal-like tumors are ER-,PR-, and HER2-. The tumor cells tended to express h4 (Fig. 2D and E), and h4 ex- associations were calculated using m2 analysis. pression was evident in areas of squamous metaplasia (Fig. 2F), which is also classified as basal like (28). Analysis of the immunohistochemical data revealed an with a basal-like phenotype, a finding that substantiates the h P gene expression analysis described above. We also classified inverse association between 4 expression and both ER ( = + P h these tumors into four subtypes (luminal A, luminal B, HER2 / 0.006) and PR ( = 0.009) and no association between 4 and - HER2 (P = 0.511; Table 1). In contrast, a significant association ER , and basal like) based on the immunohistochemical criteria between h4 and ER-,PR-, and HER2- (triple negative) tumors established by Carey et al. (29). This analysis also revealed a h was evident (P = 0.008). A significant association also exists significant correlation between 4 and the basal-like subtype between h4 and cytokeratin 5/6 expression (P = 0.026). but not with the luminal A and luminal B subtypes (Table 2). h + - Collectively, these data indicate that h4 expression is associated A link between 4 and the HER2 /ER subtype could not be established because there were only six such tumors. Integrin b4 signature genes are associated with poor progno- sis. To test the relationship between clinical outcome and the h Table 1. Integrin 4 is associated with basal-like expression of genes highly correlated with h4, four independent features gene profiling data sets were analyzed (19, 21, 22, 30). This analysis is based on the hypothesis that genes whose expression Prognostic Integrin B4 P correlates significantly may function in common mechanisms to marker Negative, Positive, affect tumor behavior. Using Pearson correlations, a 65-gene ‘‘h4 n n (%) (%) signature’’ was generatedthat iscomposed ofthe top 0.1%of genes ER that correlate either positively or negatively with h4inatleasttwo Positive 39/48 (81.25) 9/48 (18.75) 0.006 of the four data sets (Fig.3A; Table 3). Genes that correlated in only Negative 30/54 (55.6) 24/54 (44.4) PR one data set are listed in Supplementary Table S1. Positive 48/62 (77.4) 14/62 (22.6) 0.009 To test the prognostic power of the h4 signature, the expression Negative 20/39 (51.3) 19/39 (46.7) data of the 65 signature genes were used to cluster 295 breast ER and PR cancer samples from Chang et al. (Fig. 3A; ref. 30). The two groups Double (-) 16/33 (48.5) 17/33 (51.5) 0.012 defined by the first branch of the dendrogram were designated as Others 50/66 (75.8) 16/66 (24.2) h h HER2 ‘‘ 4-high’’ and ‘‘ 4-low’’ expressing tumors. Comparison of the Positive 26/36 (72.2) 10/36 (27.8) 0.511 Kaplan-Meier survival curves showed that h4-high tumors had Negative 43/67 (64.2) 24/67 (35.8) worse relapse-free survival (P = 0.0229) and overall survival ER, PR, and HER2 (P = 0.0447) than did h4-low tumors (Fig. 3B and C). Triple (-) 12/27 (44.4) 15/27 (55.6) 0.008 Others 53/71 (74.65) 18/71 (25.35) We next analyzed three other data sets of available clinical Cytokeratin 5/6 data: one that was used to generate the signature (21), one Positive 22/41 (53.7) 19/41 (46.3) 0.026 independent data set from Sotiriou et al. (31), and another Negative 41/53 (77.4) 12/53 (22.6) independent 106-tumor data set from the University of North Carolina Microarray Database.3 The integrin h4 signature in all NOTE: The association between h4 status and other pathologic markers was calculated using m2 analysis. 3 http://genome.unc.edu/pubsup/breastGEO/

www.aacrjournals.org 1053 Clin Cancer Res 2008;14(4) February 15, 2008 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Imaging, Diagnosis, Prognosis

Fig. 3. Integrin h4 signature genes predict clinical outcomes. A, classification of 295 breast tumors of Chang et al. (30) by 65-gene h4signature.B and C, Kaplan-Meier relapse-free and overall survival curves showing the outcomes of h4 signature ^ high and h4 signature ^ low groups from the 295 tumors. D, overall survival curves of h4 signature ^ high and h4 signature ^ low groups from a total of 162 tumors of Korkola et al. (21). E, relapse-free curves of h4 signature ^ high and h4 signature ^ low groups from a total of 99 tumors of Sotiriou et al. (31). F and G, relapse-free and overall curves of two h4 signature ^ high and h4 signature ^ low groups from a total of 106 tumors from the University of North Carolina Microarray Database.

Clin Cancer Res 2008;14(4) February 15, 2008 1054 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Expression and Prognostic Value of b4Integrin

Table 3. List of 65 h4 signature genes

Presence in ID Gene symbol Gene name Maximal no. of data sets R value 4 3691 ITGB4 Integrin, h41 4 3855 KRT7 0.548 3 3918 LAMC2 Laminin, g2 0.795 3 3872 KRT17 0.751 3 2195 FAT FAT tumor suppressor homologue 1 (Drosophila) 0.52 3 2296 FOXC1 Forkhead box C1 0.435 2 3852 KRT5 Keratin 5 0.75 2 23650 TRIM29 Tripartite motif-containing 29 0.716 2 7136 TNNI2 I type 2 (skeletal, fast) 0.707 2 1956 EGFR Epidermal growth factor receptor 0.704 2 728113 LOC728113 Similar to Annexin A8 0.681 2 1001 CDH3 Cadherin 3, type 1, P-cadherin (placental) 0.668 2 6273 S100A2 S100 calcium-binding protein A2 0.636 2 2049 EPHB3 EPH receptor B3 0.595 2 3860 KRT13 0.585 2 9540 TP53I3 Tumor protein p53-inducible protein 3 0.571 2 4000 LMNA A/C 0.569 2 10610 ST6GALNAC2 ST6 (a-N-acetyl-neuraminyl-2,3-h-galactosyl-1,3)- 0.553 N-acetylgalactosaminide a-2,6-sialyltransferase 2 2 1824 DSC2 Desmocollin 2 0.552 2 3675 ITGA3 Integrin, a3 0.551 2 3898 LAD1 Ladinin 1 0.549 2 2771 GNAI2 Guanine nucleotide-binding protein (G protein), 0.547 a inhibiting activity polypeptide 2 2 6455 SH3GL1 SH3 domain GRB2-like 1 0.544 2 3678 ITGA5 Integrin, a5 0.523 2 23432 GPR161 G protein – coupled receptor 161 0.522 2 2022 ENG Endoglin 0.484 2 57513 CASKIN2 CASK interacting protein 2 0.482 2 2316 FLNA A, a 0.478 2 3783 KCNN4 Potassium intermediate/small conductance 0.449 calcium-activated channel, subfamily N, member 4 2 3914 LAMB3 Laminin, h3 0.447 2 10848 PPP1R13L Protein phosphatase 1, regulatory (inhibitor) subunit 13 like 0.444

NOTE: Genes that associate with h4 in at least two of the four data sets analyzed. Values were calculated based on Pearson correlation. The maximal R value of each gene is listed.

data sets divided tumor samples into two groups of distinct the progression from normal epithelium to carcinoma outcome. High expression of the h4 signature predicts signi- (reviewed in ref. 32). For the most part, such studies have not ficantly decreased overall survival compared with low expres- considered the complex pathology of specific carcinomas or a sion in 162 tumors from Korkola et al. (P = 0.0008; Fig. 3D; potential association of h4 with clinical outcome. Using a two- ref. 21). The h4 signature also predicted the clinical outcome of pronged approach that involved analysis of gene profiling data 99 tumors from Sotiriou et al. (P = 0.0336; Fig. 3E; ref. 31) and from human breast tumors and immunohistochemical analysis 106 tumors from the University of North Carolina with respect of these tumors, we determined that h4 expression is not to relapse-free survival (P = 0.0818) and overall survival homogeneous in breast cancer but that its expression is (P = 0.0296; Fig. 3F and G). restricted to a limited number of tumors and correlates most significantly with basal-like tumors. Moreover, we identified a Discussion cluster of genes whose expression correlates with that of h4in four gene profiling data sets of breast tumors. The value of this Although the h4 integrin has been implicated in the genesis h4 signature is that it can predict clinical outcome when and progression of carcinomas based on data from cell lines applied to independent data sets of human breast tumors. and mouse models (reviewed in refs. 3, 13, 14), data on its Importantly, the hypothesis that can be derived from these relationship to the clinical properties of human carcinomas are findings is that the genes present in the h4 signature function in relatively scant, although interesting data exist on the correla- concert with h4 to affect tumor behavior and affect the clinical tion between h4 mRNA expression and breast tumor size and progression of human breast cancer. grade (18). In general, however, existing assumptions on h4 A major finding in this study is that both h4geneand expression in carcinomas are rather generic in this regard: h4 protein expression correlates most significantly with basal- expression can be maintained (e.g., breast and squamous), like tumors. These are aggressive tumors identified initially increased (e.g., thyroid), or suppressed (e.g., prostate) during by gene profiling that account for 15% to 30% of all breast

www.aacrjournals.org 1055 Clin Cancer Res 2008;14(4) February 15, 2008 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Imaging, Diagnosis, Prognosis

Table 3. List of 65 h4 signature genes (Cont’d)

Presence in Locus ID Gene symbol Gene name Maximal no. of data sets R value 2 1048 CEACAM5 Carcinoembryonic antigen-related cell adhesion molecule 5 0.443 2 22883 CLSTN1 Calsyntenin 1 0.426 2 7407 VARS Valyl-tRNA synthetase 0.426 2 55357 TBC1D2 TBC1 domain family, member 2 0.419 2 29927 SEC61A1 Sec61 a1 subunit (Saccharomyces cerevisiae) 0.417 2 4851 NOTCH1 Notch homologue 1, translocation-associated (Drosophila) 0.412 2 4240 MFGE8 Milk fat globule-EGF factor 8 protein 0.411 2 22839 DLGAP4 Disks, large (Drosophila) homologue-associated protein 4 0.41 2 6663 SOX10 SRY (sex-determining region Y)-box 10 0.385 2 2099 ESR1 ER1 -0.238 2 528 ATP6V1C1 ATPase, H+ transporting, lysosomal 42 kDa, V1 subunit C1 -0.24 2 23658 LSM5 LSM5 homologue, U6 small nRNA associated (S. cerevisiae) -0.246 2 5828 PMP35 Perxin 2 -0.277 2 9443 CRSP9 Cofactor required for Sp1 transcriptional activation, subunit 9 -0.279 2 178 AGL Amylo-1, 6-glucosidase, 4-a-glucanotransferase -0.287 2 10564 ARFGEF2 ADP-ribosylation factor guanine nucleotide exchange factor 2 -0.296 2 5684 PSMA3 Proteasome (prosome, macropain) subunit, a type, 3 -0.333 2 10438 C1D Nuclear DNA-binding protein -0.334 2 7013 TERF1 Telomeric repeat-binding factor (NIMA-interacting) 1 -0.337 2 6880 TAF9 TAF9 RNA polymerase II -0.347 2 5204 PFDN5 Prefoldin subunit 5 -0.351 2 9147 SDCCAG1 Serologically defined colon cancer antigen 1 -0.356 2 5440 POLR2K Polymerase (RNA) II (DNA directed) polypeptide K, 7.0 kDa -0.367 2 10063 COX17 COX17 cytochrome c oxidase assembly -0.373 homologue (S. cerevisiae) 2 28981 IFT81 Intraflagellar transport 81 homologue (Chlamydomonas) -0.38 2 11103 KRR1 KRR1, small subunit processome component, homologue (yeast) -0.381 2 27258 LSM3 LSM3 homologue, U6 small nRNA associated (S. cerevisiae) -0.385 2 6860 SYT4 Synaptotagmin IV -0.385 2 51626 DYNC2LI1 , cytoplasmic 2, light intermediate chain 1 -0.387 2 6902 TBCA folding cofactor A -0.398 2 9738 CP110 CP110 protein -0.407 2 10935 PRDX3 Peroxiredoxin 3 -0.408 2 54536 EXOC6 Exocyst complex component 6 -0.411 2 9521 EEF1E1 Eukaryotic translation elongation factor 1q1 -0.419

cancers (33). These tumors express basal and The relationship between h4 and erbB2 merits discussion other basal markers but often lack expression of the ER, PR, based on our findings and previously published data. Com- and HER2. As a consequence, they are refractory to tamoxifen pelling data, which were obtained using breast carcinoma cell and Herceptin therapy. The association of h4expression lines, showed a physical association between h4anderbB2on with basal-like tumors is consistent with the fact that h4is the cell surface and established that these two receptors can expressed in the basal myoepithelial cells of the normal cooperate to activate phosphatidylinositol 3-kinase and pro- mammarygland.Most,ifnotall,breasttumorsthatcontain mote h4-dependent invasion (26, 27). More recently, a mouse inactivating mutations in the BRCA1 gene, which is a major model of mammary carcinoma was used to investigate the determinantofhereditarybreastcancer,exhibitabasal role of h4 in breast cancer and its relationship to erbB2 (16). phenotype (34). Interestingly, the frequency of basal pheno- For this purpose, mice that express a constitutively active type breast carcinomas in premenopausal African-American erbB2 in the mammary gland that results in mammary carci- women is twice that for Caucasian women and this statistic nogenesis (MMTV-neu) were crossed with mice that express may contribute to the fact that these women have a 77% a cytoplasmic domain deletion of the h4 integrin subunit higher mortality from breast cancer than do Caucasian that is purported to be signaling deficient. The major finding women (29). An important issue that arises from these in this study was that deletion of the h4 cytoplasmic domain findings is the functional role of h4 in basal-like tumors. This impeded MMTV-neu–driven tumorigenesis and invasive issue should be amenable to analysis using transgenic mouse growth, and it was inferred that h4 is a potential target for models because such models have been described recently for breast tumors with amplified expression of erbB2. Our basal-like tumors (35). Although our data reveal a significant finding, however, that the expression of h4 does not correlate correlation between h4 expression and the basal phenotype, with that of erbB2 in human breast tumors questions the they also indicate that there are some h4-positive tumors that relevance of this bigenic mouse model to HER2+ human are not basal like and a smaller number of basal-like tumors breast cancer. Aside from our findings, the fact that h4was that are h4 negative (Table 2). It will be informative to assess linked to basal-like tumors in the original profiling studies the prognosis of patients with such tumors as a function of of human breast cancer diminishes the therapeutic potential h4 expression. of h4 and erbB2 because such tumors are defined by their

Clin Cancer Res 2008;14(4) February 15, 2008 1056 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Expression and Prognostic Value of b4Integrin

lack or minimal expression of HER2 (19). It should be noted, laminin (45, 46). Indeed, these earlier studies provided the first however, that the epidermal growth factor receptor (erbB1) is evidence that h4 expression could be correlated with clinical a basal marker in breast cancer (19) and its expression clusters outcome and they identified a ‘‘suprabasal’’ phenotype for with that of h4 (Table 3). These observations are of interest in squamous cell carcinomas. light of the studies that have shown a functional cross-talk The fact that the h4 signature contains actin-binding proteins between epidermal growth factor receptor and h4thatmay (filamin A and fascin) is intriguing in this regard because facilitate carcinoma progression (17, 36, 37). For these these proteins share in common with h4 the ability to facilitate reasons, targeted therapies aimed at both epidermal growth the formation and stabilization of actin protrusions and actin- factor receptor and h4 may prove effective for some types of membrane interactions that mediate cell migration and breast cancer. invasion (47, 48). Fascin is of particular interest in this regard The most novel aspect of this study is the generation of the because its expression in breast cancer is associated with basal- h4 gene signature that is prognostic for reduced survival and like tumors (49). Sox9 also merits further investigation because tumor recurrence. The hypothesis that derives from this it exhibits a strong correlation with h4 in the molecular por- cluster analysis is that genes whose expression correlates traits data (19), has been implicated in the maintenance of a significantly with h4 may function in common mechanisms basal/progenitor cell population in other epithelia (50, 51), to affect tumor behavior. This hypothesis is supported by and is associated with recurrent prostate cancer (51). At the very strong evidence from the analysis of gene expression in yeast least, the h4 gene signature provides a wealth of information and emerging data in mammalian cells (23). The h4signature for investigating functional interactions that may contribute to (Table 3), as well as the h4-related gene list (Supplementary the aggressive behavior of a subset of human breast tumors. The Table S1), includes known markers of basal-like breast can- h4 signature could also affect our understanding and treatment cers, such as K5 and K14 (38), Sox9 (19), P-cadherin (39), of basal-like breast tumors. The biology of these tumors is not laminin (40), fascin (40), epidermal growth factor receptor well understood and there is a need to identify novel and spe- (41), and Annexin A8 (42), and genes that are expressed pri- cific therapeutic targets because current treatment modalities marily in basal myoepithelial cells of the mammary gland, are limited. Defining the functional roles of these basal markers such as WT-1 (43) and S100A2 (Table 3; Supplementary in breast cancer and assessing their relationship to h4 will Table S1; ref. 44). Some of these gene products have been facilitate our understanding of basal-like tumors and the design shown to have prognostic value for breast cancer, such as WT-1 and development of therapeutic approaches for these aggressive (43) and P-cadherin (39), but others need to be evaluated tumors. more rigorously, and all of these genes need to be evaluated within the context of h4 function. It merits mentioning in this context that previous studies on squamous cell carcinomas Acknowledgments linked h4 to poor clinical outcome and found that its expression associated with some of the basal markers that we We thank Dr. Rita Falcioni for providing the h4 antibody, Dr. Chung-Cheng Hsieh identified in our h4 signature, including a3 integrin and for review of the statistical analysis, and Karen Dresser for technical assistance.

References 1. Mercurio A. Receptors for the laminins. Achieving mellae formation and the chemotactic migration of 16. Guo W, PylayevaY, Pepe A, et al. h4Integrinampli- specificity through cooperation. Trends Cell Biol invasive carcinoma cells. J Cell Biol 1998;143: fies ErbB2 signaling to promote mammary tumorigen- 19 9 5;5 :419 ^ 23. 174 9 ^ 6 0 . esis. Cell 2006;126:489 ^ 502. 2. Hemler ME, Crouse C, Sonnenberg A. Association of 9. Lipscomb EA, Simpson KJ, Ring JE, Dugan AS, 17. Rabinovitz I,TokerA, MercurioAM. Protein kinase C- the VLA a6 subunit with a novel protein. A possible Mercurio AM. The a6h4 integrin maintains the survival dependent mobilization of the a6h4 integrin from alternative to the common VLA h1subunitoncertain of human breast carcinoma cells in vivo. Cancer Res hemidesmosomes and its association with actin-rich cell lines. J Biol Chem 1989;264:6529^ 35. 2005;65:10970 ^ 6. cell protrusions drive the chemotactic migration of 3. Wilhelmsen K, Litjens SH, Sonnenberg A. Multiple 10. Santoro MM, Gaudino G, Marchisio PC. The MSP carcinoma cells. J Cell Biol 1999;146:1147 ^ 60. functions of the integrin a6h4 in epidermal homeo- receptor regulates a6h4anda3h1integrinsvia14-3- 18. Diaz LK, Cristofanilli M, Zhou X, et al. h4 Integrin stasis and tumorigenesis. Mol Cell Biol 2006;26: 3 proteins in keratinocyte migration. Dev Cell 2003;5: subunit gene expression correlates with tumor size 2877 ^ 86. 257 ^ 71. and nuclear grade in early breast cancer. Mod Pathol 4. Chao C, Lotz MM, Clarke AC, Mercurio AM. A func- 11. Trusolino L, Bertotti A, Comoglio PM. A signaling 2005;18:1165^75. tion for the integrin a6h4 in the invasive properties of adapter function for a6h4 integrin in the control of 19. Perou CM, SorlieT, Eisen MB, et al. Molecular por- colorectal carcinoma cells. Cancer Res 1996;56: HGF-dependent invasive growth. Cell 2001;107: traits of human breast tumours. Nature 2000;406: 4811^9. 643^54. 747 ^52. 5. Shaw LM, Rabinovitz I, Wang HH, Toker A, Mercurio 12. Jauliac S, Lopez-Rodriguez C, Shaw LM, Brown LF, 20. Hu Z, Fan C, Oh DS, et al.The molecular portraits of AM. Activation of phosphoinositide 3-OH kinase by Rao A,Toker A. The role of NFAT transcription factors breast tumors are conserved across microarray plat- the a6h4 integrin promotes carcinoma invasion. Cell in integrin-mediated carcinoma invasion. Nat Cell Biol forms. BMC Genomics 2006;7:96. 1997;91:949^60. 2002;4:540 ^ 4. 21. Korkola JE, Blaveri E, DeVries S, et al. Identification 6. Rabinovitz I, Mercurio AM. The integrin a6h4 13. Lipscomb EA, Mercurio AM. Mobilization and acti- of a robust gene signature that predicts breast cancer functions in carcinoma cell migration on laminin-1by vation of a signaling competent a6h4 integrin under- outcome in independent data sets. BMC Cancer 2007; mediating the formation and stabilization of actin- lies its contribution to carcinoma progression. Cancer 7:61. containing motility structures. J Cell Biol 1997;139: Metastasis Rev 2005;24:413^ 23. 22. Ma XJ,Wang Z, Ryan PD, et al. A two-gene expres- 1873 ^ 84. 14. Rabinovitz I, Mercurio AM. Dynamic functions of sion ratio predicts clinical outcome in breast cancer 7. Bachelder RE, Ribick MJ, Marchetti A, et al. p53 the a6h4 integrin in carcinoma. In: A.Wells, editor.Cell patients treated with tamoxifen. Cancer Cell 2004;5: inhibits a6h4 integrin survival signaling by promoting motility in cancer invasion and metastasis. New York 607^16. the caspase 3-dependent cleavage of AKT/PKB. JCell (NY): Springer; 2006. pp. 159 ^ 87. 23. Eisen MB, Spellman PT, Brown PO, Botstein D. Biol 1999;147:1063 ^ 72. 15. Dajee M, Lazarov M, Zhang JY, et al. NF-nBblock- Cluster analysis and display of genome-wide expres- 8. O’Connor KL, Shaw LM, Mercurio AM. Release of ade and oncogenic Ras trigger invasive human epider- sion patterns. Proc Natl Acad Sci U S A 1998;95: cAMP gating by the a6h4 integrin stimulates la- mal neoplasia. Nature 2003;421:639 ^ 43. 14 8 6 3 ^ 8.

www.aacrjournals.org 1057 Clin Cancer Res 2008;14(4) February 15, 2008 Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Imaging, Diagnosis, Prognosis

24. Saldanha AJ. Java Treeviewextensible visualiza- 34. Turner NC, Reis-Filho JS. Basal-like breast cancer nosis in breast cancer patients. Clin Cancer Res 2002; tion of microarray data. Bioinformatics 2004;20: and the BRCA1 phenotype. Oncogene 2006;25: 8:1167^71. 3246 ^ 8. 5846^53. 44. Funahashi H, Koshikawa T, Ichihara S, Ohike E, 25. Yehiely F, Moyano JV, Evans JR, Nielsen TO, Cryns 35. McCarthy A, Savage K, Gabriel A, Naceur C, Reis- Katoh K. Different distributions of immunoreactive VL. Deconstructing the molecular portrait of basal-like Filho JS, Ashworth A. A mouse model of basal-like S10 0-a and S100-h protein expression in human breast cancer.Trends Mol Med 2006;12:537 ^ 44. breast carcinoma with metaplastic elements. J Pathol breast cancer. J Surg Oncol 1998;68:25 ^9. 26. Falcioni R, Antonini A, Nistico P, et al. a6h4and 2007;211:389 ^ 98. 45.Van Waes C, Kozarsky KF,Warren AB, et al. The A9 a6h1integrins associate with ErbB-2 in human carci- 36. Mainiero F, Pepe A, Yeon M, Ren Y, Giancotti FG. antigen associated with aggressive human squamous noma cell lines. Exp Cell Res 1997;236:76 ^ 85. The intracellular functions of a6h4 integrin are regu- carcinoma is structurally and functionally similar to the 27. Gambaletta D, Marchetti A, Benedetti L, Mercurio lated by EGF. J Cell Biol 1996;134:241^ 53. newly defined integrin a6h4. Cancer Res 1991;51: AM, Sacchi A, Falcioni R. Cooperative signaling be- 37. Pullar CE, Baier BS, Kariya Y, et al. h4 Integrin and 2395 ^ 402. tweena(6)h(4)integrinandErbB-2receptorisrequired epidermal growth factor coordinately regulate electric 46.Van Waes C, Surh DM, Chen Z, et al. Increase in to promote phosphatidylinositol 3-kinase-dependent field-mediated directional migration via Rac1. Mol Biol suprabasilar integrin adhesion molecule expression invasion. JBiol Chem 2000;275:10604^10. Cell 2006;17:4925 ^ 35. in human epidermal neoplasms accompanies in- 28. Reis-FilhoJS, Milanezi F, Steele D, et al. Metaplastic 38. Abd El-Rehim DM, Pinder SE, Paish CE, et al. Ex- creased proliferation occurring with immortalization breast carcinomas are basal-like tumours.Histopathol- pression of luminal and basal cytokeratins in human and tumor progression. Cancer Res 1995;55: ogy 2006;49:10^ 21. breast carcinoma. J Pathol 2004;203:661 ^ 71. 5434^44. 29. Carey LA, Perou CM, Livasy CA, et al. Race, breast 39. Arnes JB, Brunet JS, Stefansson I, et al. Placental 47. Adams JC. Roles of fascin in cell adhesion and mo- cancer subtypes, and survival in the Carolina Breast cadherin and the basal epithelial phenotype of tility.CurrOpinCellBiol2004;16:590^6. Cancer Study. JAMA 2006;295:2492 ^ 502. BRCA1-related breast cancer. Clin Cancer Res 2005; 48. Popowicz GM, Schleicher M, Noegel AA, HolakTA. 30. Chang HY, Nuyten DS, Sneddon JB, et al. Ro- 11: 4 0 0 3 ^ 11. : promiscuous organizers of the . bustness, scalability, and integration of a wound- 40. Rodriguez-Pinilla SM, Sarrio D, Honrado E, et al. Trends Biochem Sci 2006;31:411^ 9. response gene expression signature in predicting Prognostic significance of basal-like phenotype and 49. Yoder BJ,Tso E, Skacel M, et al. The expression of breast cancer survival. Proc Natl Acad Sci U S A fascin expression in node-negative invasive breast fascin, an actin-bundling motility protein, correlates 2005;102:3738 ^ 43. carcinomas. Clin Cancer Res 2006;12:1533 ^ 9. with hormone receptor-negative breast cancer and a 31. Sotiriou C, Neo SY, McShane LM, et al. Breast can- 41. Nielsen TO, Hsu FD, Jensen K, et al. Immunohisto- more aggressive clinical course. Clin Cancer Res cer classification and prognosis based on gene ex- chemical and clinical characterization of the basal-like 2005;11:186 ^ 92. pression profiles from a population-based study. Proc subtype of invasive breast carcinoma. Clin Cancer Res 50. Seymour PA, Freude KK, Tran MN, et al. SOX9 is Natl Acad Sci U S A 2003;100:10393^ 8. 2004;10:5367 ^ 74. required for maintenance of the pancreatic progeni- 32. Mercurio AM, Rabinovitz I. Towards a mechanistic 42. SteinT, Price KN, Morris JS, et al. Annexin A8 is up- tor cell pool. Proc Natl Acad Sci U S A 2007;104: understanding of tumor invasion-lessons from the regulated during mouse mammary gland involution 1865 ^ 70. a6h4 integrin. Semin Cancer Biol 2001;11:129^ 41. and predicts poor survival in breast cancer. Clin Can- 51. Wang H, McKnight NC, Zhang T, Lu ML, Balk SP, 33. Banerjee S, Reis-Filho JS, Ashley S, et al. Basal-like cer Res 2005;11:6872 ^ 9. Yuan X. SOX9 is expressed in normal prostate basal breast carcinomas: clinical outcome and response to 43. Miyoshi Y, Ando A, Egawa C, et al. High expression cells and regulates androgen receptor expression in chemotherapy. J Clin Pathol 2006;59:729 ^ 35. of Wilms’ tumor suppressor gene predicts poor prog- prostate cancer cells. Cancer Res 2007;67:528^ 36.

Clin Cancer Res 2008;14(4) February 15, 2008 1058 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research. Analysis of Integrin β4 Expression in Human Breast Cancer: Association with Basal-like Tumors and Prognostic Significance

Shaolei Lu, Karl Simin, Ashraf Khan, et al.

Clin Cancer Res 2008;14:1050-1058.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/14/4/1050

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2021/03/11/14.4.1050.DC1

Cited articles This article cites 50 articles, 25 of which you can access for free at: http://clincancerres.aacrjournals.org/content/14/4/1050.full#ref-list-1

Citing articles This article has been cited by 24 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/14/4/1050.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/14/4/1050. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from clincancerres.aacrjournals.org on September 29, 2021. © 2008 American Association for Cancer Research.