Clinical & Experimental Metastasis

, Volume 24, Issue 2, pp 69–78

Expression of the cytoskeleton linker protein ezrin in human cancers

Authors

  • Benjamin Bruce
    • Tumor and Metastasis Biology Section, Center for Cancer ResearchNational Cancer Institute
    • Howard Hughes Medical Institute
  • Gaurav Khanna
    • Tumor and Metastasis Biology Section, Center for Cancer ResearchNational Cancer Institute
    • Howard Hughes Medical Institute
  • Ling Ren
    • Tumor and Metastasis Biology Section, Center for Cancer ResearchNational Cancer Institute
  • Goran Landberg
    • Malmö University HospitalLund University
  • Karin Jirström
    • Malmö University HospitalLund University
  • Charles Powell
    • Department of MedicineColumbia University College of Physicians & Surgeons
  • Alain Borczuk
    • Department of PathologyColumbia University College of Physicians & Surgeons
  • Evan T. Keller
    • Department of Urology, School of MedicineUniversity of Michigan
  • Kirk J. Wojno
    • Department of Urology, School of MedicineUniversity of Michigan
  • Paul Meltzer
    • Cancer Genetics BranchNational Human Genome Research Institute
  • Kristin Baird
    • Cancer Genetics BranchNational Human Genome Research Institute
  • Andrea McClatchey
    • Department of Pathology, Harvard Medical SchoolMass General Hospital Cancer Center
  • Anthony Bretscher
    • Department of Molecular Biology and GeneticsCornell University
  • Stephen M. Hewitt
    • Tissue Array Research Program, Laboratory of Pathology, Center for Cancer ResearchNational Cancer Institute
    • Tumor and Metastasis Biology Section, Center for Cancer ResearchNational Cancer Institute
Original Paper

DOI: 10.1007/s10585-006-9050-x

Cite this article as:
Bruce, B., Khanna, G., Ren, L. et al. Clin Exp Metastasis (2007) 24: 69. doi:10.1007/s10585-006-9050-x

Abstract

Expression of the metastasis-associated protein, ezrin, in over 5,000 human cancers and normal tissues was analyzed using tissue microarray immunohistochemistry. Ezrin staining was compared between cancers and their corresponding normal tissues, between cancers of epithelial and mesenchymal origin, in the context of the putative inhibitor protein, merlin, and against clinicopathological data available for breast, lung, prostate cancers and sarcomas. Ezrin was found in most cancers and normal tissues at varying levels of intensity. In general ezrin was expressed at higher levels in sarcomas than in carcinomas. By normalizing the expression of ezrin in each cancer using ezrin expression found in the corresponding normal tissue, significant associations between ezrin were found in advancing histological grade in sarcomas (P = 0.02) and poor outcome in breast cancer (P = 0.025). Clinicopathologic associations were not changed by simultaneous assessment of ezrin and merlin in each patient sample for the cancer types examined. These data support a role for ezrin in the biology of human cancers and the need for additional studies in breast cancer and sarcoma patients that may validate ezrin as a marker of cancer progression and as a potential target for cancer therapy.

Keywords

EzrinMerlinImmunohistochemistryTissue microarrayBiomarkerPrognosis

Introduction

Ezrin is a member of the band 4.1 superfamily of proteins, including ezrin, radixin, and moesin (collectively known as ERM proteins). Ezrin and the other ERM proteins act as linkers between the cell membrane and the actin cytoskeleton. A model describing ezrin’s role as a linker protein suggests that ezrin can reside in a closed inactive state that results from association of its N and C terminal domains. This closed state may result from intramolecular self-association or through formation oligomers that include ezrin and other ERM proteins [1]. Following phosphorylation, the N–C terminal domain association is released, resulting in a conformational “opening” of ezrin. In its open form ezrin then provides its linker function by allowing its N-terminus to bind directly or indirectly to the cell membrane and its C-terminus to bind f-actin [2]. The ezrin linker function has both physical and functional consequences. The physical connection is thought to allow cellular processes including membrane extension, motility, invasion and adherence [3, 4]. The functional linkage allows for optimal signaling from membrane-associated receptors, including hepatocyte growth factor receptor, epidermal growth factor receptor, CD44 (hyaluronic acid receptor), and ICAM-1 to their respective downstream effector pathways, e.g., Cdc42, MAP kinase, PI3kinase/pAKT and mTOR [59]. Redundancy in ERM protein function has been demonstrated both in vitro and in vivo [2]. Mice that are homozygous null for ezrin have a phenotype associated with intestinal villous malformation [10]. Interestingly intestinal epithelial cells mostly, if not exclusively, express ezrin rather than the other members of the ERM family of proteins (radixin and moesin). The lack of a defining phenotype in tissues that express radixin and moesin lends support to the functional redundancy of ERM proteins during normal tissue development and maintenance.

We have previously demonstrated the importance of ezrin in the biology of metastasis in two highly metastatic pediatric cancers, osteosarcoma and rhabdomyosarcoma [4]. The suppression of metastasis through disruption of ezrin expression or function in murine models of these pediatric cancers was seen despite the sustained expression of radixin and moesin, suggesting a unique role for ezrin amongst ERM proteins in cancer and cancer progression. Recent histology specific studies have demonstrated the expression and relevance of ezrin in a number of cancers including carcinomas of endometrium, breast, colon, ovary, uveal and cutaenous melanoma, brain tumors, and most recently soft tissue sarcomas [1113, 14].

The role of ezrin in cancers is likely to be complex and variable and may be defined by its relative expression in corresponding normal tissues, the balance of kinases and phosphatases expressed in a given cancer, or by the expression of proteins that bind and interact with ezrin function. Merlin, also a member of the band 4.1 superfamily of proteins, is co-expressed with ezrin in many cell types and shares binding partners and interacts directly with ezrin [1517]. In vitro studies support the role of merlin in controlling cell growth, motility and adherence [18]. Interestingly, mice heterozygous for NF2, the gene encoding merlin, spontaneously develop osteosarcoma [18, 19]. Together, these data suggest a model wherein ezrin and merlin may act both independently and together to regulate the actin-cytoskeleton linkage necessary for cancer progression and metastasis in some cancers.

The purpose of this study was to define the expression of ezrin in a broad spectrum of human cancers, including breast, lung and prostate. In addition we have applied a consistent detection technique to determine if the expression of ezrin in these cancers is associated with available clinical outcome or predictors of outcome, as available. Using tissue microarrays we examined ezrin expression in over 5,000 human cancers and normal tissue samples. We found that ezrin was expressed in most normal tissues and most cancers. For normal tissues, the expression of ezrin was greater in epithelial versus mesenchymal tissues, meanwhile, expression of ezrin was higher in most mesenchymal tumors compared to most epithelial tumors. Normalizing the expression of ezrin in each cancer with their corresponding normal tissue counterparts provided a consistent basis for evaluation of ezrin across cancer histologies. Doing so uncovered significant associations between ezrin expression and available clinicopathological data in soft tissue sarcomas and in breast cancers. This report represents the first wide scale survey ezrin in human cancers and suggests the complexity of ezrin’s role in the biology of cancer. Furthermore, these data suggest potential opportunities to evaluate ezrin as a prognostic maker or therapeutic target specifically in breast cancer and sarcomas.

Materials and methods

Antibodies and cell lines

Monoclonal anti-ezrin antibody was purchased from Sigma 3C12 (aa 362–585). Affinity purified rabbit antiserum to human merlin was derived as previously reported [17].

Western: Ezrin and merlin antibodies were used at concentrations of 1:4,000 and 1:500, respectively. Human cancer cell lines included Rh30 rhabdomyosarcoma, RD rhabdomyosarcoma, HOS osteosarcoma, MDA-MB-435 and MDA-MB-231T breast cancer (kindly provided by Dr. P. Steeg), A549 and HEK157 lung cancer (kindly provided by Dr. P. Dennis), and PC3M prostate cancer (kindly provided by Dr. D. Bottaro). Murine cancer cell lines included K7M2 and K12 osteosarcoma, previously shown to be high and low ezrin expressors, respectively. Mouse embryonal fibroblasts, derived from merlin knockout and wild-type mice were used as negative and positive controls, respectively, for merlin expression [16].

Tissue arrays

Tissue microarrays (TMA) were generally constructed as previously reported [4]. A list of specific TMAs used in this report are provided in Table 1. The normal tissue arrays were from the Cooperative Human Tissue Network (CHTN) (65 normal human tissues). The Tissue Array Project (TARP5) array consisted of 324 cancer tissue samples of multiple tumor types and histologies [20] (www.cancer.gov/tarp). The cancer-specific tissue arrays were selected to evaluate ezrin in the most common cancers seen in patients including cancers of the breast, prostate and lung. Interestingly relatively little data on the role of ezrin in these cancers was previously available. To compliment assessment in these epithelial cancers we included a sarcoma specific array that included a wide number of mesenchymal cancers, an area of active interest in our laboratory. These cancer-specific arrays included the Columbia Lung Cancer array (58 different lung cancer patients with associated clinicopathologic and outcome data [21, 22], the Consecutive Breast Cancer Array—Lund [23] (574 patients associated with outcome data) and the Michigan prostate tissue microarrays TMA07B and TMA45 [24] (250 normal prostate, prostatic intraepithelial neoplasia, and prostate cancer including Gleason score; Gleason score of each core was confirmed by H&E section adjacent to the ezrin stained array). Patient and clinical correlative data for each of these arrays has been previously described and reported as above. Lastly we utilized the National Human Genome Research Institute (NHGRI) sarcoma array (327 patients encompassing 11 sarcoma types).
Table 1

Tissue microarrays used in survey of ezrin expression in human cancers

Tissue microarray name

Type/ Histology

Patient Numbers

Clinicopathological data

Endpoints assesseda

Cooperative Human Tissue Network Array

Normal Tissue

65

None

Normal tissue expression

Tissue Array Project 5 Array

Multi-tumor and Normal Tissue

324

None

Normal and tumor expression

NHGRI Sarcoma Array

Sarcoma specific

327

Histological grade

Expression versus histological grade

Columbia Lung Cancer Array

Lung cancer specific

58

Survival

Expression versus survival

Consecutive Breast Cancer Array

Breast cancer specific

574

Disease free survival, stage, histologic grade, age, size, ki67, estrogen receptor expression

Expression versus disease free survival, stage, histologic grade, age, size, ki67, estrogen receptor expression

Michigan Prostate Tissue Microarray (TMA07B and TMA45)

Prostate specific (normal prostate, PIN, prostate cancer)

250

Gleason score

Expression versus Gleason Score

aExpression includes naïve and normalized ezrin and merlin, and ezrin/merlin

Immunohistochemistry

Immunohistochemical staining of tissue arrays was as previously described [25]. Slides were antigen retrieved with buffer (DAKO) for 15 min. Endogenous peroxidase was quenched and block was applied for 30 min. The slides were incubated overnight with primary antibody (Ezrin, Sigma 3C12 at 1:500 or merlin at 1:125, in 0.2% non fat drymilk) at 4C. Slides were developed with avidin–biotin complex kit (DAKO). Overlapping fragments of the N (1–293) and C (290–580) termini of ezrin were used to demonstrate specificity of the commercially available ezrin antibody in our hands. Ezrin and merlin staining were conducted on sequential slides cut from each TMA. Each core was scored based on intensity of staining. A scale from 0 (no staining) to 4 (high intensity staining) was used for assessment of intensity. The score assigned to a core was based on the highest staining intensity found within at least 5% of the total core. Cellular localization was included in the assessment of ezrin staining as a quality filter and noted as membranous, cytoplasmic alone, and diffuse. Cores that contained less than 15% of the original tissue were excluded from the analysis. The lack of non-specific binding of the secondary antibody was shown by performing a parallel experiment without primary antibody. All tissue cores were co-scored by BB and at least one other co-author to reach a consensus score.

Assessment of the relationship between ezrin expression and clinical outcome was made by normalizing the “native” tumor expression intensity of ezrin, independently, against expression in corresponding normal tissue for each tumor core:

“High Expressor” = ezrin expression greater than one-half standard deviation above the mean expression for that cancer’s normal tissue

“Low Expressors” = ezrin expression less than one-half standard deviation below the mean expression for that cancer’s normal tissue.

All other samples were categorized as “Normal Expressors”.

For merlin “High” and “Low Expressors” were similarly defined, using expression intensities one standard deviation above or below of the mean merlin expression intensity in corresponding normal tissue. All other samples were defined as “Normal Expressors”.

Statistical analysis

Descriptors of ezrin and merlin expression intensity, including mean, median, and confidence intervals, were determined using Excel and Prism software. Fischer’s exact t-test and tests of repeated measures (ANOVA) were used to compare mean expression intensities (ezrin, merlin). Univariate survival analysis for lung cancer and breast cancer samples was undertaken using the LogRank test and Cox’s proportional hazzards analysis. For the breast cancer samples, correlations between variables were tested according to Spearman’s correlation test. A P-value of less than 0.05 was considered statistically significant. Statistical analysis was performed using Prism 3.0 (for MacIntosh) and SPSS 11.0 software (SPSS Inc., Illinois, USA).

Results

The specificity of antibodies for ezrin and merlin were confirmed by western analysis of human cells (Fig. 1). Controls for ezrin expression included murine osteosarcoma lines (K7M2 and K12) previously shown to express high and low ezrin, respectively. Negative control lysates included murine cells knocked out for merlin. Ezrin and merlin staining of normal tissue by immunohistochemistry were consistent with previously published data demonstrating expression of both proteins in most normal tissues [26] (Table 2 and Supplementary Table 1; CHTN and TARP TMA). Blocking experiments using overlapping fragments of the N (1–293) and C (290–580) termini of ezrin confirmed the specificity of the commercially available ezrin antibody (data not shown). Ezrin expression was highest in breast, colon, endometrium, renal cortex, lymph nodes, placenta, prostate and spleen. Cerebellum, liver, muscle and thyroid all demonstrated low ezrin expression. Positive, strong merlin expression was seen in the endothelial cells of all tissues, as well as kidney, salivary gland and testis. Low expression for merlin was seen in bone, spleen, lymph nodes, and placenta. The scoring system for ezrin and merlin was primarily based on staining intensity with some consideration for the percent of cells expressing ezrin or merlin in a core. Ezrin stained in an even and diffuse pattern in all TMA cores stained, and is consistent with the pattern observed in staining of whole sections of tumors (data not shown). As a result we defaulted to strongest intensity of staining as a metric of ezrin expression. A similar pattern was observed with merlin, however an internal positive control was routinely used—positive staining of endothelial cells. We found ezrin localization to be both diffuse cytoplasmic and membranous in nearly all samples evaluated, as such localization was not used as an independent endpoint for ezrin assessment.
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-006-9050-x/MediaObjects/10585_2006_9050_Fig1_HTML.gif
Fig. 1

Ezrin and Merlin are expressed in human cancer cell lines. A. Western blot for ezrin revealed an appropriately sized 80 kd band, consistent with ezrin in the positive control high ezrin expressing K7M2 osteosarcoma cell line and no band in negative control metastatic low ezrin expressing K12 cell line. No cross reactivity with other 4.1 superfamily proteins such as radixin or moesin was seen using extended gels (data not shown). B. Western blot for merlin yielded a single band at approximately 69kb in positive control embryonal fibroblasts from mice wild type ice, but not in negative control embryonal fibroblasts from mice homozygous null for merlin.C. Western blot on a panel of human cancer cell lines for ezrin and merlin. Cell lines include HOS-MNNG and U2 OS (osteosarcoma), RH30 and RD (rhabdomyosarcoma), MDA-MB-231T and MDA-MB-435 (breast carcinoma), A549 and K157 (lung carcinoma), and PC3M (prostate carcinoma)

Table 2

Immunohistochemical expression of ezrin in normal human tissues

Samplea

Number samplesb

Mean ezrin intensityc

Median ezrin intensity

SDd

Min

Max

Bone Marrow

3

1.7

1

1.7

0

4

Breast

7

2.7

3

0.8

1

4

Cerebellum

12

1.3

1

1.3

0

4

Colon

13

2.8

3

1.3

2

4

Endometrium

6

2.7

2

0.9

2

4

Kidney

11

3.0

3

0.7

2

4

Liver

11

0.4

0

0.5

0

1

Lung

144

2.9

3

0.9

0

1

Lymph Node

7

3.9

4

0.3

3

4

Muscle

8

0.9

1

0.0

0

1

Pancreas

16

2.9

3

0.8

2

4

Placenta

6

2.8

3.5

1.5

0

4

Prostate

73

3.2

3

0.9

0

4

Salivary

11

2.4

2

0.9

1

4

Skin

1

3.0

3

0.0

3

3

Spleen

9

3.2

4

1.3

2

4

Testis

10

2.8

3

0.4

2

4

Thyroid

3

1.3

0

1.8

0

4

aHuman normal tissues present on the TARP4, TARP5 and CHTN tissue arrays were stained for ezrin by immunohistochemistry

bThe number of individual samples from each tissue type available for staining

cEzrin staining intensity was assessed as on a semiquantitative scale from 0 to 4

dStandard deviation (SD) of the mean ezrin staining intensity

Using the optimized and standardized staining conditions described above, the expression of ezrin in cancers of epithelial (TARP5 array; 238 intact samples) and mesenchymal (NHGRI Sarcoma array; 215 intact samples) origin was analyzed. Examples of ezrin staining are shown in Fig. 2 and a summary of the staining analysis is presented in Table 3. Ezrin expression intensity was higher in sarcomas (mean = 3.3) compared to epithelial cancers (mean = 2.7; unpaired t-test with Welch’s correction; P < 0.0001). Among sarcomas, mean ezrin expression was highest in Ewing’s sarcoma, gastrointestinal stromal tumor (GIST), rhabdomyosarcoma and synovial sarcomas. Prostate cancers were found to have the highest ezrin expression of epithelial cancers. Ezrin expression was lowest in ovarian cancers.
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-006-9050-x/MediaObjects/10585_2006_9050_Fig2_HTML.jpg
Fig. 2

Immunohistochemical staining of ezrin in human cancer tissue array cores. Representative photomicrographs of different tumors stained for ezrin,A adenocarcinoma of the lung,B adenocarcinoma of the breast,C adenocarcinoma of the prostate, and D Ewings sarcoma. Tissue array cores were scored for ezrin staining intensity on a scale from 0 to 4. Staining intensities of ezrin shown are 3 or 4. Total magnification of all images 150X. The intensity defined for each core was determined by the highest level of staining expressed in at least 5% of the total stained core.

Table 3

Ezrin expression in epithelial and mesenchymal cancers

https://static-content.springer.com/image/art%3A10.1007%2Fs10585-006-9050-x/MediaObjects/10585_2006_9050_Tab3_HTML.gif

The expression of ezrin in normal tissue and the corresponding cancer tissue was compared where available (Fig. 3; TARP5, NHGRI, and CHTN TMAs). The relative expression of ezrin was highest in sarcomas (rhabdomyosarcoma and leiomyosarcoma) compared to their normal tissue counterparts. Ezrin expression in osteosarcoma was also high relative to normal bone, which expresses low ezrin protein (unpublished results, C. Khanna). In breast, colon, lung and prostate tumors, the mean expression intensity was similar to their normal tissue counterparts. Given the diversity of ezrin expression seen between cancers and normal tissues we hypothesized that a comparison of ezrin biology across cancer histologies would be improved by consideration of tumor versus corresponding normal tissue expression of ezrin. As such, “native” tumor ezrin expression was normalized by comparison with the mean ezrin expression for that cancer’s normal tissue of origin, and defined as High, Normal or Low expression. Using this normalization, the expression of ezrin was assessed in comparison to available clinicopathological data (Table 1) on the sarcoma, lung, and breast, and prostate cancer arrays. Significant associations were found between ezrin expression and histologic grade and clinical outcome in sarcoma and breast cancer, respectively (Fig. 4). For sarcoma samples “native” and “normalized” ezrin expression were nearly identical since mesenchymal tissues expressed low levels of ezrin. Histological grade was available on 105 cores for which ezrin staining intensity was available. The intensity of ezrin expression was positively and significantly associated with tumor grade (ANOVA, P = 0.0005; Fig. 4a). No significant differences were seen in the expression of ezrin in primary tumor and a small number of un-matched metastases (data not shown). Patient outcome was not available for sarcoma array samples. In the breast cancer specific array (n = 443), no significant association between “native” ezrin intensity score and clinical outcome was identified in the entire population of patients or within subgroups of patients defined by clinical stage or histological sub-type (data not shown). Using the “normalized” ezrin intensity, at 200 months, there was a significant correlation between “High Expression” and recurrence (Fig. 4b). Patients in the “High Expressors” group had 45% recurrence, Normal Expressors had 37% recurrences and Low Expressors had 28% recurrence (P = 0.048). A significant disease free survival advantage was associated with “Low” ezrin expression in breast cancer patients (n = 92) whereas, “High” ezrin expression was associated with the most aggressive disease course (n = 130) and patients with “Normal” ezrin expression had intermediate outcomes (n = 210; Hazard ratio = 1.33, 95% CI = 1.04–1.72; P = 0.025). In breast cancer patient samples, modest but significant correlations were found between “normalized” ezrin expression and advancing histological grade (Spearman’s rho = 0.181, P < 0.001), increasing age (Spearman’s rho = 0.110, P = 0.021), increasing primary tumor size (Spearman’s rho = 0.100, P < 0.035), KI67 expression (Spearman’s rho = 0.130, P = 0.007). No association was found between normalized ezrin expression and increasing estrogen receptor expression (Spearman’s rho = 0.015, P = 752). No significant associations were found between “native” or “normalized” ezrin expression and clinical outcome or Gleason scoring in lung cancer and prostate cancer patients, respectively (data not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-006-9050-x/MediaObjects/10585_2006_9050_Fig3_HTML.gif
Fig. 3

Ezrin expression in mesenchymal cancers is higher than epithelial cancers, while expression in normal mesenchymal tissues is lower than most epithelial tissues. Arrays of epithelial cancers, sarcomas and normal tissues (TARP5, NHGRI, and CHTN TMAs) were stained with anti-ezrin antibody and scored for ezrin intensity as described above. Mean ezrin expression for each cancer was compared with ezrin expression for their respective normal tissues of origin. Error bars represent the standard deviation. While epithelial cancers do not express significantly different ezrin levels compared with their normal tissue of origin, sarcomas have significantly higher ezrin expression compared with their normal tissue of origin. Astericks represents significant differences (P < 0.05; unpaired t-test with Welch’s correction). Intestinal stroma was assessed in 2 samples and for this reason could not be assessed statistically (no error bar included). Ezrin expression was highest in normal breast, colon, endometrium, renal cortex, lymph nodes, placenta, prostate and spleen. Cerebellum, liver, muscle and thyroid all demonstrated low ezrin expression

https://static-content.springer.com/image/art%3A10.1007%2Fs10585-006-9050-x/MediaObjects/10585_2006_9050_Fig4_HTML.gif
Fig. 4

Assessment of ezrin expression in sarcomas, lung, breast and prostate cancer. A. Mean ezrin expression increases with histological grade in sarcomas. The intensity of scoring for ezrin varied with increasing histological grade (ANOVA, P = 0.0005; non-parametric t-test of grade I versus grade III tumors, P = 0.02). Mean ezrin staining intensity for sarcoma tissue microarray cores was evaluated with histological grade for each sample. Error bars represent the standard deviation. B. Ezrin expression in breast cancer tissue array samples (n = 432). High normalized ezrin staining intensity was associated with a poor outcome in breast cancer patients by Kaplan–Meyer survival analysis (Log Rank Test P = 0.04). Patients were identified as “Low” or “High” based on a “normalization” for ezrin expression in normal human breast tissue (as described). All other patients were defined as “Normal” expressors. A significant disease free survival advantage was associated with “Low” ezrin expression in breast cancer patients (n = 92). “High” ezrin expression was associated with the most aggressive disease course (n = 130); patients with “Normal” ezrin expression had intermediate outcomes (n = 210; Hazard ratio = 1.33, 95% CI = 1.04–1.72; P = 0.025)

Merlin expression in cancers and normal tissues was assessed on sequential sections cut from the same tissue arrays as described for ezrin (Supplementary Fig. 1 and Supplementary Table 1 and 2). This data provided a cancer-wide survey of merlin expression in cancer and the basis to assess the role of ezrin on the biology of lung, breast, and prostate cancers and sarcomas, in the context of this putative ezrin inhibitory protein. Merlin expression was significantly greater in mesenchymal tumors (mean = 2.72) compared to epithelial tumors (mean = 1.22; unpaired t-test with Welch’s correction; P < 0.0001). To determine whether the concurrent assessment of ezrin and merlin (from sequential sections) was more informative than ezrin alone, samples were classified into one of four groups, based on expression intensity: high ezrin-high merlin, low ezrin-high merlin, high ezrin-low merlin, or low ezrin-low merlin. Our initial hypothesis for these studies was that tumors with relatively high ezrin and low merlin would be more likely to follow an aggressive clinical course. The concurrent assessment of ezrin and merlin on a sample-by-sample basis did not improve or change the previously demonstrated associations found for ezrin expression alone in any of the cancers studied (sarcoma, lung, breast and prostate cancers; data not shown). No associations between “native” or “normalized” merlin were found in any of the cancers studied (data not shown).

Discussion

We have examined ezrin expression in over 5,000 human tissues using an identical immunohistochemical staining method on TMAs. These data provide a broad perspective of ezrin expression in a diverse set of human cancers as well as normal tissues, providing a context to understand the role of this cytoskeleton-linker protein in the biology of cancer. Ezrin was found to be expressed in most human cancer and normal tissues; however, it was highest in cancers of mesenchymal origin and in prostate cancers. Whether ezrin expression was associated with the biology of a given cancer was not necessarily predicted by the tissue of origin, by the basal expression level of ezrin in the corresponding normal tissue, nor by the concurrent expression of merlin. The assessment of ezrin expression with established covariates of outcome (i.e., histological grade in sarcomas, Gleason score in prostate cancers) or directly with clinical outcome for breast and lung cancer revealed distinct relationships between ezrin and each cancer type. These data suggest that cancers of mesenchymal origin and cancers of breast as the most promising initial targets for studies of ezrin biology and for studies of ezrin-based prognostication and therapeutic targeting.

One limitation of past reports of ezrin expression in cancer has been small studies that included single cancer types with limited consideration of normal tissue expression. Tissue microarrays offered many advantages, not the least of which was the opportunity to survey ezrin and then merlin expression over a large number of specimens. Antibodies for ezrin and merlin staining have been used and reported previously and were validated by Western analysis with knock-out mouse control tissue lysates, the use of murine and human cell lines previously shown to express ezrin, and through comparison with immunohistochemical staining patterns previously reported for normal tissues [26]. The analysis of ezrin expression in this study included a broad representation of normal tissues. Ours and previous studies are also potentially limited by the assessment of total ezrin and not the conformationally open or phosphorylated form (defined in part by phosphorylation at the c-terminal threonine 567). Detection of phosphorylated ezrin T567 is not possible using immunohistochemical techniques since antibodies directed against this phosphorylated form of ezrin cannot distinguish the other phosphorylated ERM family members. It is believed that following phosphorylation of ezrin T567 that secondary phosphorylation events occur through the ezrin protein. Specific ezrin phosphorylation sites have been associated with specific and distinct ezrin functions. The use of phosphospecific antibodies directed against other sites including Y353, S66, and Y477 may be the basis of follow-up studies of ezrin in specific cancers defined by the data presented herein. Whether phosphorylated ezrin expression is in fact a more valuable predictor of clinical biology of cancers is not known and may be challenged by recent data suggesting the importance of both phosphorylated and non-phosphorylated forms of ERMs [27].

It may be argued that ubiquitous expression of ezrin in normal tissues may limit its future utility as a maker of prognosis or as a therapeutic target in cancer. From the standpoint of prognostication, normal tissue expression must be considered in light of its relative expression in tumor. This was the basis of our assessment of “normalized” ezrin within each cancer histology. In terms of considering ezrin as a target for therapy, it is important to consider ezrin’s functional necessity during cancer progression versus normal tissues, rather than to expression alone. Differences in normal tissue versus tumor do not necessarily suggest a role for ezrin in cancer transformation, rather they suggest the potential that ezrin may influence that cancer’s biology. The specific role for ezrin in a given cancer is likely to be multifold. As discussed earlier there is considerable functional redundancy between ERM proteins. This is evidenced most clearly in the ezrin homozygote null mouse that is viable at birth and has a phenotype limited to the intestine, an organ expressing none of the other ERM proteins. It is interesting to speculate that the redundancy provided by ERM proteins in most normal tissues may be protective against toxicities associated with future anti-ezrin therapeutics. This normal tissue functional redundancy may provide a therapeutic window between the effects of ezrin inhibition in normal tissues and in cancers undergoing metastasis where the ERM proteins are not able to compensate for the loss of ezrin expression [3, 4].

Ezrin was consistently found in both epithelial and mesenchymal cancers. Consistent with past studies, our data indicate that most normal epithelial tissues have high ezrin expression. In contrast, normal mesenchymal tissues, such as smooth muscle, skeletal muscle, and endothelium, were found to have relatively low ezrin expression. This finding led to the hypothesis that the relationship between ezrin expression and a specific cancer could be better defined by the relative comparison between expression in cancer tissue and corresponding normal tissue [2830]. To control for differences in normal tissue versus tumor expression of ezrin we chose a statistical basis for defining relative tumor ezrin expression as “High,” “Normal” or “Low.” Normalization of ezrin expression, based on expression in each cancer’s corresponding normal tissue allowed a common basis for the assessment of ezrin across different cancers. As a result of this normalization of ezrin expression, significant associations were found between ezrin expression and histologic grade in sarcomas and with outcome in breast cancer patients. Our finding that ezrin expression was associated with advancing histological grade in sarcomas is consistent with our “normalization” hypothesis and recent reports in the literature demonstrating the potential importance of ezrin in other cancers whose normal tissue counterparts express low ezrin [12, 14, 31, 32]. Unfortunately small sample numbers of metastatic tissues in the sarcoma array precluded the opportunity to assess ezrin’s role in sarcoma progression, through assessment of ezrin expression in primary versus metastatic tissues. Interestingly, in support of our data, a recent report of patients with soft tissue sarcomas identified ezrin as a predictor of the development of metastasis and survival [31]. For breast cancer patients ezrin expression was linked to disease free survival by Kaplan Meirer analysis. Furthermore, modest, although statistically significant correlations were made between normalized ezrin expression and histological grade, age, tumor size, and ki67. Further analysis of ezrin expression in breast cancer samples is warranted based on this data. This analysis should evaluate ezrin expression within histological sub-types and grades and within cancers with distinct estrogen receptor expression profiles. Using this same normalization method in lung cancer, no significant association between ezrin expression and clinical outcome was found. Uniformly high ezrin staining in the prostate cancer samples and the relatively high Gleason score of the samples included in the investigated arrays limited the opportunity to define any associations between ezrin expression and Gleason score in prostate cancers. As such further studies of ezrin in prostate cancer are needed.

While ezrin’s link to cancer biology has been established in the literature, there is a growing body of evidence that suggests that a loss of merlin could play a role in tumor development and progression [18], specifically in tumors of neural origin. That merlin and ezrin are co-expressed in many cell types, share binding partners, and mutually interact suggest a model wherein merlin may inhibit the pro-metastatic function of ezrin [15, 33]. To begin to test the hypothesis that the balance of ezrin and merlin expression in a cancer would be more predictive than assessment of ezrin alone, we concurrently evaluated ezrin and merlin expression in tumor samples. We found merlin to be expressed in most human cancers, following a general pattern of expression seen for ezrin. According to our hypothesis, samples with a relatively high ezrin versus merlin expression would be expected to have a more aggressive clinical outcome than samples with relatively lower ezrin versus merlin. The concurrent evaluation of ezrin and merlin expression in each of the cancers (sarcomas, lung, breast, and prostate cancer) did not yield greater information or stronger clinical associations than the evaluation of ezrin alone. It is reasonable that the action and interaction between ezrin and merlin are dynamically regulated during cancer progression. Accordingly it may not be surprising that the concurrent assessment of these proteins at a fixed point in time, within the primary tumor, was not more informative.

To conclude, this work represents a complete assessment of ezrin expression in over 5,000 tissue samples. Our findings suggest that ezrin is expressed in most human cancers and normal tissues. In general normal epithelial tissues expressed greater levels of ezrin than normal mesenchymal tumors. Meanwhile, epithelial cancers, with the exception of prostate cancer, had relatively lower expression ezrin. The evaluation of ezrin expression against available clinicopathological data revealed a significant association between high ezrin expression and sarcoma histological grade and with risk of recurrence in breast cancer patients. Further studies of ezrin in specific cancers are warranted to define the determinants of the ezrin-associated phenotype.

Acknowledgements

This research was supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. Benjamin Bruce and Gaurav Khanna completed this work while participating in the Howard Hughes Medical Institute Fellowship. Contributions by the University of Michigan were supported by included National Cancer Institute Grants P01 CA093900 and Specialized Program of Research Excellence 1P50 CA69568.

Supplementary material

10585_2006_9050_MOESM1_ESM.ppt (44 kb)
ESM 1 (PPT 44 kb)
10585_2006_9050_MOESM2_ESM.doc (3.1 mb)
ESM 2 (DOC 3,179 kb)

Copyright information

© Springer Science + Business Media B.V. 2007