Altered intracellular region of MUC1 and disrupted correlation of polarity-related molecules in breast cancer subtypes.

MUC1 glycoprotein is overexpressed and its intracellular localization altered during breast carcinoma tumorigenesis. The present study aimed to clarify the relationship of cytoplasmic localization of MUC1 with the breast cancer subtype and the correlation of 10 molecules associated with cell polarity in breast cancer subtypes. We immunostained 131 formalin-fixed and paraffin-embedded breast cancer specimens with an anti-MUC1 antibody (MUC1/CORE). For 48 of the 131 tumor specimens, laser-assisted microdissection and real-time quantitative RT-PCR were performed to analyze mRNA levels of MUC1 and 10 molecules, β-catenin, E-cadherin, claudin 3, claudin 4, claudin 7, RhoA, cdc42, Rac1, Par3 and Par6. Localization of MUC1 protein varied among breast cancer subtypes, that is, both the apical domain and cytoplasm in luminal A-like tumors (P < 0.01) and both the cytoplasm and cell membrane in luminal B-like (growth factor receptor 2 [HER2]+) tumors (P < 0.05), and no expression was found in triple negative tumors (P < 0.001). Estrogen receptor (ER)+ breast cancers showed higher MUC1 mRNA levels than ER- breast cancers (P < 0.01). The incidence of mutual correlations of expression levels between two of the 10 molecules (55 combinations) was 54.5% in normal breast tissue and 38.2% in luminal A-like specimens, 16.4% in luminal B-like (HER2+), 3.6% in HER2 and 18.2% in triple negative specimens. In conclusion, each breast cancer subtype has characteristic cytoplasmic localization patterns of MUC1 and different degrees of disrupted correlation of the expression levels between the 10 examined molecules in comparison with normal breast tissue.

Breast cancer is a common cause of death among women worldwide,1 and its incidence is increasing among Japanese women. Although breast cancer is a heterogeneous disease, gene expression profiling by DNA microarray analysis is furthering our understanding of its biology.2–4 Intrinsic breast cancer subtypes include luminal A, luminal B, human epidermal growth factor receptor 2 (HER2) overexpressing, basal-like and normal breast-like, which exhibit different prognoses, therapeutic responses and biological behaviors. The St. Gallen International Conference 2013 proposed practical subtypes based on the intrinsic subtypes, and recommended specific treatments.5 The practical breast cancer subtypes are based on immunohistochemistry results for estrogen receptor (ER), progesterone receptor (PgR), HER2 and Ki-67.5 Clinicopathologic surrogate definitions exist for luminal A-like, luminal B-like (HER2−), luminal B-like (HER2+), HER2 and triple negative (TN) subtypes. Because treatments are recommended according to subtype, their identification is important.

MUC1 is a highly O-glycosylated, membrane-bound member of the mucin family that is expressed at the apical domain of epithelial cells, including normal breast cells.6,7 MUC1 has a core protein of 20 amino acid tandem repeats covered by sugar chains.7–9 The MUC1 core protein includes an extracellular domain, a transmembrane domain and a cytoplasmic tail. During tumorigenic transformation, MUC1 is overexpressed and localized in various intracellular regions.10,11 Overexpression of MUC1 enables cancer cells to avoid apoptosis,12,13 and the MUC1 cytoplasmic tail binds to ER-α, β-catenin and p53, which influence tumor growth.6 From a clinical perspective, a peptide vaccine that targets MUC1 has been developed and is being tested in clinical studies.14–16 Whether intracellular localization of MUC1 varies according to breast cancer subtype is unclear. Therefore, in this study, we examined the intracellular localization of MUC1 protein in breast cancer subtypes.

The intrinsic subtype is ambiguously related to morphological features such as tubule formation and mucin production. For example, tubular carcinoma and mucinous carcinoma are categorized predominantly into the luminal subtype, while medullary carcinoma belongs mainly to the basal-like subtype.17 Tubules are an integrated structure based on cell polarity.18 The tight junction functions as a boundary to distinguish apical and lateral domains, and as a barrier to protect against unselective passage in the intercellular space.19 Domains of the plasma membrane are regulated by Par and Crumbs for the apical domain and Scribble for the basolateral domain.20 In the process to form 3-D tissue from individual cells, the cells alter their actin filament dynamics through activation of the Rho family of GTPases.18 Thus, cell polarity is organized by molecules regulating junctional structures, domain identity and Rho family GTPases in a coordinated manner. Therefore, we hypothesized that if intracellular localization of MUC1 is altered in breast cancer subtypes, it would be associated with maintenance of cell polarity. To this end, we examined the relationships among representative molecules of junctional structures (β-catenin, E-cadherin, claudin 3, claudin 4 and claudin 7), domain identity (Par3 and Par6), and Rho family members (Cdc42, Rac1 and RhoA) to further understand the altered intracellular localization of MUC1 in breast cancer subtypes.

Materials and Methods

Patients

We examined breast cancer specimens from 131 patients who underwent surgery at Nihon University Hospital Division of Breast and Endocrine Surgery between 2005 and 2007 (Table1). The patients were 40–70 years of age at the time of their surgeries. No patients received pre-surgical treatment. Breast cancer tissues were fixed in formalin, embedded in paraffin and sectioned. All samples were pathologically examined according to the General Rules for Clinical and Pathological Recording of Breast Cancer21 and the World Health Organization classification system.22 The specimens included 125 invasive carcinomas and six noninvasive carcinomas. The study protocol was approved by the institutional ethics committee and conformed to the provisions of the Declaration of Helsinki.

Table 1

Patients' clinicopathological characteristics

Category	Number of cases
Subtype
Luminal A	81
Luminal B (HER2−)	11
Luminal B (HER2+)	12
HER2 (non-luminal)	9
Triple negative (ductal)	18
Histology
Papillotubular	59
Scirrhous	35
Solid tubular	9
Mucinous	6
Noninvasive	6
Mixed type	3
Apocrine	3
Invasive lobular	1
Others	9
Stage
0	5
I	54
II	77
III	4
IV	2
Tumor size (cm)
≤2	71
2–≤5	54
>5	6
Lymph node status
Negative	100
Positive	31
Total cases	131

Immunohistochemistry

All breast cancers specimens were examined immunohistochemically for ER (clone 1D5, Dako Cytomation, Carpinteria, CA, USA), PgR (clone PgR686, Dako), HER2 (clone SV2-61γ, Nichirei Bioscience, Tokyo, Japan) and Ki-67 (clone MIB-1, Dako) according to the manufacturers' instructions. Thresholds for receptor positivity were ER ≥ 1%, PgR ≥ 20% and Ki-67 ≥ 14%. HER2 staining was evaluated according to the American Society of Clinical Oncology/College of American Pathologists guidelines.23 Immunohistochemical intrinsic type definitions were luminal A-like: ER+ /PgR+/HER2−/Ki67low; luminal B-like (HER2−): ER+/HER2−/Ki67high and/or PgR− or low; luminal B-like (HER2+): ER+/HER2+/Ki-67any/PgRany; HER2: ER−/PgR−/HER2+; and TN: ER−/PgR–/HER2−.

All samples were immunostained with an anti-MUC1 antibody (MUC1/CORE, clone Ma552, mouse IgG, Leica Biosystems, Newcastle, UK). Immunohistochemical staining was performed as follows. Paraffin-embedded sections (4-μm thick) of breast cancer specimens were dewaxed, subject to heat-induced antigen retrieval at 121°C for 15 min, cooled and then incubated for 5 min in PBS containing 0.1% Triton-X. The sections were stained with the MUC1/CORE primary antibody (1:200) for 30 min and then the simple Stain MAX PO (MULTI) secondary antibody (Nichirei Bioscience, Tokyo, Japan) for 30 min. Colorization was performed with PBS containing 3,3′-diaminobenzidine and 0.5% H2O2.

Laser assisted microdissection and extraction of total RNA from breast tumor tissues

As previously described, tissues were mounted as 8-μm-thick paraffin sections on film-coated glass slides, dewaxed with xylene and stained with toluidine blue.24–26 The breast tumor areas were microdissected using a laser-assisted microdissection system (PALM MBIII, Carl Zeiss Microscopy, Munich, Germany). The microdissected tissues were treated with 200 μL denaturing buffer (2% SDS, 0.1 mM EDTA and 10 mM Tris-HCl). The tissues were then incubated at 55°C with proteinase K until digested completely. Total RNA was purified by adding 20 μL of 2 M sodium acetate (pH 4.0), 220 μL citrate-saturated phenol (pH 4.3) and 60 μL chloroform-isoamyl alcohol, followed by centrifugation for 15 min at 25000g. The upper aqueous layer was transferred into a new tube, mixed with 200 μL isopropanol and 2 μL glycogen, and then incubated at −80°C for at least 30 min. The samples were centrifuged for 30 min at 20000g, washed with 70% ethanol, air dried on ice, dissolved in 5–10 μL RNase-free water and then quantified at an optical density of 260 nm using a Nanodrop 1000 (Thermo Fisher Scientific, Waltham, MA, USA). Total RNA samples were stored at −80°C until use. Both genomic DNA elimination and cDNA synthesis were performed with a QuantiTect Reverse Transcription Kit (QIAGEN, Tokyo, Japan) according to the manufacturer's instructions.

Of the 131 specimens examined immunohistochemically, those with a sufficient quantity and quality of extracted mRNA were studied further, which included 48 tumor and 15 normal breast samples.

Real-time quantitative RT-PCR analyses of MUC1 and cell–cell junction and cell polarity-related genes

The mRNA expression levels of MUC1, β-catenin, E-cadherin, claudins 3,4 and 7, RhoA, Cdc42, Rac1, Par3, Par6 and the internal control GAPDH were measured by semi-nested quantitative (snq) RT-PCR.24 The first RT-PCR was carried out for each target and control cDNA using AmpliTaq Gold 360 Master Mix (Life Technologies Japan, Tokyo, Japan) and the primer sets listed in Table2. Samples were incubated at 95°C for 10 min and then subjected to 25 cycles of PCR at 94°C for 30 s, 60°C for 30 s and 72°C for 60 s. The first reaction was performed on a conventional PCR machine (PC808, ASTEC, Fukuoka, Japan). One microliter of each resulting product was used as the template in the second snqPCR amplification in an ABI Prism 7000 Sequence Detection System (Life Technologies) using SYBR Green detection chemistry. Briefly, snqPCR amplification was performed in a 20-μL reaction volume containing 900 nmol/L of each primer used in the first RT-PCR and 1× SYBR Green PCR Master Mix (Life Technologies). Reaction mixtures were preheated at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Each relative mRNA value was calculated using the ΔΔCt method27 with threshold cycle times for each target mRNA and GAPDH.

Table 2

Sequences of primer sets used for real time RT-PCR

Gene	Forward primers	Reverse primers
MUC-1	5′-CTAGCAGTACCGATCGTAGCC-3′	5′-CCACTGCTGGGTTTGTGTAA-3′
β-catenin	5′-TGGATACCTCCCAAGTCCTG-3′	5′-CAGGGAACATAGCAGCTCGT-3′
E-cadherin	5′-GACTCGTAACGACGTTGCAC-3′	5′ -GGTCAGTATCAGCCGCTTTC-3′
Claudin 3	5′ -TGCTCTGCTGCTCGTGTC-3′	5′-GGTGTTGGTGGTGGTGGT-3′
Claudin 4	5′-AGATGGGTGCCTCGCTCTAC-3′	5′-CCAGGGAAGAACAAAGCAGA-3′
Claudin 7	5′-GGGTGGAGGCATAATTTTCA-3′	5′-AGGACAGGAACAGGAGAGCA-3′
RhoA	5′-CGCTTTTGGGTACATGGAGT-3′	5′-CAAGACAAGGCACCCAGATT-3′
Cdc42	5′-CGAGTTCAAGCGATTCTCCT-3′	5′-CGCCCACAACAACACACTTA-3′
Rac1	5′-AACCAATGCATTTCCTGGAG-3′	5′-GGATAGGATAGGGGGCGTAA-3′
Par 3	5′-TGTTCGACGCAGTAGTGACC-3′	5′-CAGCAGTGTTCTGCTTGAGG-3′
Par 6	5′-TCGCTGGAAAGATCAAAACC-3′	5′-TCCATGGATGTCTGCATAGC-3′
GADPH	5′-GGAAGGTGAAAGGTCGGAGTCA-3′	5′-GTCATTGATGGCAACAATATCCACT-3′

Analysis of the relationships among molecules related to cell polarity

To examine relationships among molecules related to cell polarity, we estimated correlations between their expression in all 55 combinations. When the numbers of correlated combinations decreased significantly, they were assumed to be deviations from those in normal breast tissue and designated as mild (>50% of normal tissue correlation), moderate (33–50%) or severe (<33%). These deviations were considered to reflect disruptions of cell polarity.

Statistical analysis

Statistical analysis was performed using Stat View for Windows version 5.0. The uniqueness of MUC1 immunohistochemical localization in each subtype was evaluated using the χ2-test. Correlations between MUC1 mRNA expression and the breast cancer subtype were analyzed using the Mann–Whitney U-test. Deviations in coordinated expression of cell polarity-related molecules from those in normal tissue were estimated using the χ2-test. The correlation of the expression level of cell polarity-related molecules was analyzed using the Pearson correlation coefficient. Pearson's r-value, the strength of the linear relationship between two variables, is shown simply as the “r” value, and its statistical significance is shown as P-values. P < 0.05 was considered to be significant.

Results

Intracellular localization of MUC1 protein and breast cancer subtypes

MUC1/CORE immunoreactivity was positive in 93.9% of samples (123/131) and localized at the apical domain, in the cytoplasm and/or at the cell membrane. The localization of MUC1 protein was classified into six patterns (Fig.1a–f): (1) apical domain only; (2) at both the apical domain and in the cytoplasm; (3) at the apical domain, in the cytoplasm and at the cell membrane; (4) in the cytoplasm and at the cell membrane; (5) cytoplasm only; and (6) no MUC1 expression detected.

Figure 1

Intracellular localization pattern of MUC1. MUC1 expression was observed in three locations: the apical domain, cytoplasm and cell membrane. Accordingly, MUC1 localization patterns were classified as: (1) apical only (a); (2) apical + cytoplasm (b); (3) apical + cytoplasm + cell membrane (c); (4) cytoplasm + cell membrane (d); (5) cytoplasm only (e); and (6) negative (f).

We analyzed the correlations between the cytoplasmic localization patterns of MUC1 protein and each breast cancer subtype (Table3). In luminal A-like tumors, MUC1 immunoreactivity was localized significantly at the apical domain and in the cytoplasm more frequently than the total of the other five subtypes (P = 0.008). In luminal B-like (HER2+) tumors, MUC1 immunoreactivity was localized significantly in the cytoplasm and at the cell membrane more frequently than the total of the other five subtypes (P = 0.030). TN tumors were negative for MUC1 more frequently than the total of the other five subtypes (P = 0.006). Significant correlations between MUC1 localization and breast cancer subtypes are summarized in Table4.

Table 3

Intracellular localization patterns for MUC1 in breast cancer subtypes

Pattern	LA		LB (HER2−)		LB (HER2+)		HER2		TN		Total N
	n	P	n	P	n	P	n	P	n	P
(1)	11	0.172	0	0.231	0	0.209	1	0.966	2	0.950	14
(2)	40	0.008*	4	0.773	4	0.598	3	0.652	2	0.006**	53
(3)	1	0.430	0	0.761	0	0.750	0	0.785	0	0.689	1
(4)	5	0.243	2	0.221	3	0.030*	0	0.347	1	0.640	11
(5)	23	0.109	4	0.839	4	0.984	5	0.148	8	0.294	44
(6)	1	0.003**	1	0.666	1	0.735	0	0.428	5	0.001>*	8
Total	81		11		12		9		18		131

(1) Apical-only; (2) apical + cytoplasm; (3) apical + cytoplasm + cell membrane; (4) cytoplasm + cell membrane; (5) cytoplasm-only; and (6) negative.

positive correlation

negative correlation. LA, luminal A; LB, luminal B; TN, triple negative.

Table 4

Summary of significant patterns of MUC1 intracellular localization by breast cancer subtype as shown by immunohistochemistry

Subtype	MUC1 localization
Luminal A	Apical + cytoplasm**
Luminal B (HER2−)	NS
Luminal B (HER2+)	Cytoplasm + membrane*
HER2	NS
Triple negative	Negative***

P < 0.05

P < 0.01

P < 0.001. NS, not significant.

Correlation between MUC1 mRNA expression and breast cancer subtypes

MUC1 mRNA expression levels normalized to GAPDH expression were significantly higher in breast carcinoma (relative MUC1 mRNA value: 2.01 ± 3.62 [mean ± SD]) than in normal breast tissue (relative MUC1 mRNA value: 0.42 ± 0.45; P = 0.024; Fig.2). In ER+ breast cancers (i.e. luminal A-like and luminal B-like), MUC1 mRNA expression was significantly higher (relative MUC1 mRNA value: 3.07 ± 4.46) than in ER− cancers (i.e. HER2 and TN) (relative MUC1 mRNA value: 1.14 ± 2.38; P < 0.01).

Figure 2

MUC1 mRNA levels normalized to GAPDH expression in breast cancer subtypes. The number of cases for each breast cancer subtype and the relative MUC1 mRNA value (mean ± SD) are shown. Expression levels were significantly lower in normal breast tissue than in breast carcinoma (P = 0.024). ER+ subtypes showed significantly higher MUC1 expression than ER− subtypes (P < 0.01). ER, estrogen receptor; LA, luminal A-like; LB, luminal B-like; TN, triple negative.

Correlation of expression levels of molecules related to cell polarity in breast cancer subtypes

Based on the hypothesis that maintenance of cell polarity affects cytoplasmic localization of MUC1 protein, we analyzed the mRNA expression of 10 molecules related to tight junction (TJ)(claudin 3, 4 and 7), adherens junction (AJ) (β-catenin and E-cadherin), Rho family members (cdc42, Rac1 and RhoA) and domain identity (Par3 and Par6) (Table S1).

We examined correlations between molecules in all 55 combinations, of which 30 of the 55 pairs (54.5%) were correlated significantly and well maintained in normal breast tissues (Table5), but decreased to 21/55 pairs (38.2%) in luminal A-like (no significant difference compared with 54.5% in normal breast tissue), 9/55 (16.4%) pairs in luminal B-like (HER2+) (P < 0.001), 2/55 (3.6%) pairs in HER2 (P < 0.001) and 10/55 (18.2%) pairs in TN (P < 0.001) breast tumors (Table5; Fig.3). Because we only had two luminal B-like (HER2−) cases, this subtype was not examined in the correlation analysis. All subtypes except for luminal A-like had disrupted cell polarity with moderate (TN) or severe (luminal B-like [HER2+], HER2) degrees according to the criteria described in the Materials and Methods. Representative correlations between the molecule pairs are shown in Fig.4, and their significant correlations are schematically presented in Fig.5. Expression of AJ-TJ-Rho-Par was positively correlated to normal breast tissue, TJ-Rho-Par were partially correlated to luminal A-like, and only incomplete correlations were found in the other types (Fig.5).

Table 5

Significant correlations among mRNA levels of genes related to cell–cell junction and cell polarity, by breast cancer subtype

Gene 1	Gene 2	Normal breast	LA	LB (HER+)	HER2	TN
		n = 15	n = 11	n = 10	n = 8	n = 17
β-cat	E-cad	r = 0.89**	NS	NS	NS	NS
	Cl3	r = 0.90**	r = 0.80**	NS	NS	r = 0.81***
	Cl4	r = 0.97**	r = 0.77**	NS	NS	NS
	Cdc42	r = 0.95**	r = 0.62*	NS	NS	NS
	Rac1	r = 0.95**	r = 0.72*	NS	NS	r = 0.57*
	Par3	r = 0.64*	r = 0.66*	r = 0.79**	r = 0.76*	r = 0.94**
	Par6	r = 0.90**	NS	r = 0.92***	NS	r = 0.83**
E-cad	Cl3	r = 0.82**	NS	NS	NS	NS
	Cl4	r = 0.82**	NS	r = 0.86**	NS	NS
	Cdc42	r = 0.87**	NS	NS	NS	NS
	Rac1	r = 0.89**	NS	r = 0.79*	NS	NS
	Par6	r = 0.88**	NS	NS	NS	NS
Cl3	Cl4	r = 0.86**	r = 0.10**	NS	NS	NS
	Cl7	r = 0.86**	r = 0.90***	NS	NS	NS
	Cdc42	r = 0.94**	r = 0.75**	NS	NS	NS
	Rac1	r = 0.95**	r = 0.95**	NS	NS	r = 0.50*
	Par3	NA	r = 0.90***	NS	NS	r = 0.64**
	Par6	r = 0.87**	NS	NS	NS	r = 0.59*
Cl4	Cl7	r = 0.87**	r = 0.87***	NS	NS	NS
	Cdc42	r = 0.92**	r = 0.77**	NS	NS	NS
	Rac1	r = 0.91**	r = 0.96**	r = 0.75*	NS	NS
	Par3	r = 0.63*	r = 0.92***	NS	NS	NS
	Par6	r = 0.82**	NS	NS	NS	NS
Cl7	Cdc42	r = 0.94**	r = 0.64*	NS	NS	NS
	Rac1	r = 0.93**	r = 0.83**	NS	NS	NS
	Par3	r = 0.88**	r = 0.75**	r = 0.91***	NS	NS
	Par6	NA	r = 0.74**	r = 0.85**	NS	NS
Rho A	Par6	r = 0.50*	NS	NS	NS	NS
Cdc42	Rac1	r = 0.97**	r = 0.89***	NS	r = 0.79*	r = 0.94**
	Par3	r = 0.63*	r = 0.82**	r = 0.67*	NS	NS
	Par6	r = 0.94**	NS	NS	NS	NS
Rac1	Par3	NA	r = 0.93**	NS	NS	r = 0.53*
	Par6	r = 0.96**	NS	NS	NS	NS
Par3	Par6	NA	NS	r = 0.87**	NS	r = 0.86**

P < 0.05

P < 0.01

P < 0.001. β-cat, β-catenin; E-cad, E-cadherin; Cl3, claudin 3; Cl4, claudin 4; Cl7, claudin 7. LA, luminal A; LB, luminal B; NA, not available; NS, not significant.

Figure 3

Degree of correlation between molecules related to cell polarity in breast cancer subtypes. The y-axis shows the proportion of correlated pairs (number of correlated combinations among 55 combinations). The deviation from normal breast tissue (54.5%) was analyzed statistically. *P < 0.05 and **P < 0.0001. LA, luminal A-like; LB, luminal B-like; TN, triple negative.

Figure 4

Correlations between mRNA expression of claudin3-cdc42, claudin3-Rac1, and claudin4-Rac1 in normal breast tissue and breast cancer subtypes. Normal breast (a-c), Luminal A-like (d-f), Luminal B-like (HER2+)(g-i), HER2 (j-l) and Triple negative (m-o) breast cancers were examined for correlations between claudin3-cdc42 (a, d, g, j, m), claudin3-Rac1 (b, e, h, k, n) and claudin4-Rac1 (c, f, i, l, o). Trend lines, correlation coefficients (r-values) and statistical significance (P-values) are shown. NS; not significant Correlations between cell polarity-related molecules were highly preserved in normal tissue, but less in luminal A-like tumor tissue, and disrupted in other subtypes.

Figure 5

Schematic summary of correlations between molecules associated with adherens junction (AJ), tight junction (TJ), domain identity (Par) and Rho family members (Rho) in normal breast tissue and breast cancer subtypes. Significant correlations listed in Table5 are shown as colored solid lines. (a) Normal breast tissue shows significant correlations between AJ-TJ-Par (*1), AJ-Par-Rho (*2), AJ-TJ-Rho (*3), TJ-Par-Rho (*4), and claudins 3, 4 and 7 (*5). (b) The luminal A-like group showed limited correlations among TJ-Par-Rho (*4) and claudins 3, 4 and 7 (*5). Luminal B-like (HER2+) (c), HER2 (d) and triple negative (e) groups showed incomplete correlations among AJ, TJ, Par and Rho.

Discussion

The present study confirmed that MUC1 mRNA expression was higher in ER+ breast cancers than in ER− breast cancers, as reported previously.28 Our results also demonstrated that cytoplasmic localization of MUC1 protein varies between breast cancer subtypes, that is, at the apical domain and in the cytoplasm of luminal A-like tumors, in the cytoplasm and at the membrane in luminal B-like (HER2+) tumors, and negative in TN tumors. The recruitment of MUC1 protein at the apical domain, a common pattern in normal breast tissue, was preserved in luminal A-like tumors, but not in other subtypes. It is interesting that another ER+ subtype, luminal B-like (HER2+), lacked the restricted localization at the apical domain. It is well known that carcinoembryonic antigen (CEA) is mainly located at the apical domain of intestinal epithelial cells, but its localization is altered in carcinoma cells. Colorectal cancer cells forming a glandular structure express CEA at the apical domain accompanied by occludin, a protein involved in tight junction, whereas cancer cells arranged in solid nests lacking occludin express CEA diffusely in their cytoplasm.29 The present study revealed another example of a human cancer presenting an altered distribution of an apical protein.

We further demonstrated that whereas the expression levels of molecules involved in tight junction, adherens junction and domain identity, as well as Rho family members were well correlated in normal breast tissue (54.5%), these correlations decreased to 38.2% in luminal A-like, 16.4% in luminal B-like (HER2+), 3.6% in HER2 and 18.2% in TN breast cancer subtypes. By analyzing the results in detail, we uncovered several findings. The correlation between β-catenin and E-cadherin, representative proteins of adherens junction, was preserved in normal breast tissue but lacking in all of the breast cancer subtypes. Furthermore, the correlations among tight junction proteins (claudins 3, 4 and 7) and between TJ-Pars-Rho were preserved in normal breast tissue and luminal A-like breast tumors but not in other subtypes. Thus, the present study showed that not only apical localization of MUC1 but also correlations among tight junction proteins (claudins) and TJ-Par-Rho were preserved in luminal A-like breast tumors but not in other subtypes.

Accumulating evidence shows that cells can sense the surrounding environment through receptors (e.g. integrins) and communicate with other cells through junctional structures (adherens and tight junctions). To establish cell polarity, cells have the separated plasma membrane of apical domain, regulated by Par (Par3, Par6, aPKC) and Crumbs (Crumbs, PALS/Stardust, PATJ/Discs) and basolateral domain regulated by Scribble (Scribble, LGL, DLG).30 In the process to form 3-D tissue from individual cells, the cells alter their actin filament dynamics through activation of the Rho family of GTPases (Cdc42, Rac1 and Rho). The junctional structure and domain identity are also coordinated by protein trafficking systems. Membrane proteins are sorted to apical or basolateral domains from the trans-Golgi network (TGN) or recycling endosomes via precise mechanisms. Sorting mechanisms to the basolateral membrane are well investigated. The basolateral membrane arises from the adaptor protein (AP)-enriched domain, and its sorting is stringently regulated by signals at the cytoplasmic tail, AP and scaffold proteins, including clathrin.31 In contrast, the apical domain arises from lipid rafts of the TGN, which are enriched with cholesterol, glycosphingolipids and sphingomyelin.32 Sorting determinants for the apical membrane are diverse and vaguely understood, which include the glycosylphosphatidylinositol anchor,33 hemagglutinin and neuraminidase,34 N-glycans35 and O-glycans,36 although it has not been fully clarified whether MUC1 is a sorting determinant for the apical membrane. The present study showed that representative molecules contributing to establishment of cell polarity (junctional structures, domain identity and Rho family members), as well as MUC1 were well maintained in normal breast tissue, but disrupted by various degrees in human breast cancer subtypes.

Patients with luminal B-like (HER2+) tumors are expected to benefit from anti-HER2 antibody therapy in addition to hormone therapy. Clinical trials of MUC1 vaccines have been performed for non-small cell lung cancers,15,16 and these vaccines might be suitable for patients with ER+ breast cancers with high MUC1 expression. The present study raises a further consideration for vaccine therapy, the cytoplasmic localization of MUC1. Whereas the apical domains of epithelial cells in normal breast tissue and luminal A-like tumors usually lay along the lumens of breast tubules, luminal B-like (HER2+) tumors express MUC1 protein at their circumferential cell membrane and may face the fibrovascular stroma. Thus, non-apical localization of MUC1 in luminal B-like (HER2+) tumors, as shown in the present study, may provide an effective means of selecting patients who may benefit from an MUC1 peptide vaccine.

In conclusion, we demonstrated that the cytoplasmic localization of MUC1 protein varies among intrinsic tumor types, and found different degrees of disrupted correlation of expression levels between the 10 examined molecules in comparison with normal breast tissue.