The Expression Pattern of Epidermal Differentiation Marker Keratin 10 in the Normal Human Breast and Breast Cancer Cells

Abstract

Cells of the human breast gland express an array of keratins, of which some are used for characterizing both normal and neoplastic breast tissue. However, the expression pattern of certain keratins has yet to be detailed. Here, the expression of a differentiation marker of epidermal epithelium, keratin 10 (K10), was investigated in the human breast gland. While in normal breast tissue generally less than 1% of luminal epithelial cells expressed K10, in women >30 years of age glandular structures with K10-positive (K10^pos) cells were found at higher frequency than in younger women. K10^pos cells belong to a mature luminal compartment as they were negative for cKIT, positive for Ks20.8, and mostly non-cycling. In breast cancer, around 16% of primary breast carcinomas tested were positive for K10 by immunohistochemistry. Interestingly, K10^pos tumor cells generally exhibit features of differentiation similar to their normal counterparts. Although this suggests that K10 is a marker of tumor differentiation, data based on gene expression analysis imply that high levels of K10 dictate a worse outcome for breast cancer patients. These findings can form the basis of future studies that should unravel which role K10 may play as a marker of breast cancer:

Introduction

The female human breast gland consists of an epithelium organized in a differentiation hierarchy that has yet to be fully understood. Unraveling the organization of this hierarchy is complicated by the fact that several events can lead to changes in the composition of the gland with increased age, including repeated cycles of menstruation, pregnancy and lactation, as well as postmenopausal regression. Acquiring a detailed insight into the various cellular subsets that populate the epithelial compartment in the normal gland may be key to understand the mechanisms that underlie the development of different breast cancer subtypes.

The distinct expression patterns of keratin intermediary filaments have widely been used to characterize the epithelia of various organs. In a number of epithelia, a subcellular change of keratin expression marks a transit between differentiation states. For example, in the epidermis there is a switch from expression of K5 and K14 in cells of the basal layer that provide the stem cell compartment to K1 and K10 in the differentiating suprabasal keratinocytes.¹ Thus, a distinct distribution of keratins within an epithelium may be involved in compartmentalizing various differentiation states.

The luminal epithelial cells that line the ducts and lobular ductules in the human breast gland express a selection of keratins that characterize simple epithelia, including K7,² K8,³ K18,³ and K19.⁴ However, the expression pattern is more complex as subsets of cells also express several keratins characteristic of stratified epithelia, including K5,⁵ K6a,⁶ K14,⁷ K15,⁶ and K17.⁷ Documentation of the presence of other stratified epithelial keratin markers in the breast gland is limited.⁸ For K10, its expression has not been described in the normal gland, and in breast cancer it has been reported to be present in rare cases.^9–11

This study reveals that in fact rare K10^pos cells are found in the normal human breast gland and that structures that contain these cells are more frequently present in women >30 years of age. Marker expression analysis demonstrated that K10^pos cells are part of a differentiated compartment in both normal breast and breast cancer tissue, and that a significant subset of tumors are in fact K10^pos.

Materials and Methods

Human Breast Tissue Samples

Normal human breast tissue was donated by consenting women undergoing reduction mammoplasty for cosmetic reasons performed at Capio CFR Hospitaler at Lyngby or Hellerup, Denmark. All personal information on donors was confidential except for the date and year of birth. In some cases, only the year of birth was available. Material from some of the biopsies has been used in previous studies.^12–15 Samples of breast cancer were obtained completely anonymously from patients at the State University Hospital, Copenhagen. The storage and use of human material have been reviewed and approved by the Regional Scientific Ethical Committees (H-2-2011-052 and H-3-2010-095) and the Danish Data Protection Agency (2011-41-6722) and have been handled according to established guidelines in subsequent experiments.

Immunostaining

Cryopreserved tissue specimens were cut at 6 μm on a cryostat (Microm HM560; Axlab, Vedbæk, Denmark) and were subsequently fixed either in ice cold methanol for 5 min and air-dried or in 3.7% formaldehyde (Merck; Darmstadt, Germany), followed by permeabilization with Triton X-100 detergent (Sigma-Aldrich, Denmark). For peroxidase staining, specimens were incubated with primary antibody for 1 hr, and Ultravision ONE Detection System was used according to the manufacturer’s instructions (Labvision; Thermo Fisher Scientific, Denmark). Nuclei were stained with hematoxylin. For immunofluorescence, the specimens were incubated for 2 hr with primary antibody followed by incubation with secondary isotype-specific Alexa Fluor antibodies for 30 min. Nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole). Cultured cells were fixed and stained as described above. For staining of smears, cells were first dissociated to single cells from breast epithelial organoids isolated as described previously.^12,16 Apart from classifying them by the presence of estrogen receptor α (ERα), breast carcinomas were also subdivided into molecular phenotypes by staining for progesterone receptor and HER2: Luminal A (ERα^pos and/or PR^pos, no HER2 overexpression), Luminal B (ERα^pos and/or PR^pos, HER2 overexpression), HER2-enriched (ERα^neg, PR^neg, HER2 overexpression), and Triple-Negative (ERα^neg, PR^neg, no HER2 overexpression). Antibodies and dilutions are listed in Table 1.

Table 1.

List of Antibodies Used for Immunostaining.

Antibody Target	Clone	Company/Cat. No.	Peroxidase	Fluorescence
cKIT	K45	NeoMarkers/MS-289	1:100	1:25
Estrogen receptor α	1D5	Dako/M7047	1:100	—
HER2	TAB250	Invitrogen/28-0003Z	1:200	—
Keratin 10	DE-K10	Invitrogen/MA5-13705	1:50	1:25
Keratin 10	RKSE60	Millipore/MAB3230	1:10	—
Keratin Ks20.8	Ks20.8	Dako/M7019	—	1:10
Keratin	CAM5.2	BD Biosciences/345779	—	1:25
Ki67	SP6	Thermo Scientific/RM-9106	—	1:25
Progesterone receptor	Pgr636	Dako/M3569	1:100	—
Alexa Fluor 488 goat anti-mouse IgG1	—	Invitrogen/A21151	—	1:500
Alexa Fluor 568 goat anti-mouse IgG2a	—	Invitrogen/A21134	—	1:500
Alexa Fluor 568 goat anti-rabbit IgG	—	Invitrogen/A11036	—	1:500

Microscopy and Image Acquisition

A lobular or ductal segment with the presence of one or more K10-expressing cells after performing immunohistochemistry (IHC) was assigned as K10^pos. For breast carcinomas, a tumor was considered positive for K10, cKIT, Ks20.8, ERα, or PR when >1% of the neoplastic cells expressed the given marker. Quantification of smear stainings and sections stained for Ki67 was performed on images that were acquired and digitally stored. Images were acquired on a Leica DM550B microscope equipped with a DFC550 digital camera or a Zeiss LSM 700 confocal microscope located at the Core Facility for Integrated Microscopy, Faculty of Health and Medical Sciences, University of Copenhagen. In smears, 1000 CAM5.2-positive (CAM5.2^pos) luminal cells were quantified for each biopsy. For quantification of Ki67-positive cells in cultures, 200 cells were counted (either within the K10^pos population or on the overall population) in each of three different culture flasks. On tissue sections, 200 K10^pos cells or CAM5.2^pos cells were evaluated for Ki67 expression for each of three tumors.

Cell Culture

MCF10A is a spontaneously immortalized breast epithelial–derived non-tumorigenic cell line.¹⁷ MCF10A cells were cultured in Dulbecco’s Modified Eagle’s Medium/Nutrient mixture F-12 (DMEM/F12, 1:1; Life Technologies, Denmark) supplemented with 2 mmol/l glutamine, 20 ng/ml epidermal growth factor (PeproTech, Germany), 10 µg/ml insulin, 0.5 µg/ml hydrocortisone, 10 ng/ml cholera toxin (Sigma-Aldrich), and 5% fetal bovine serum (FBS). MCF7 breast cancer cells were cultured in DMEM 1965 supplemented with 2 mmol/l glutamine, 7 non-essential amino acids (NEA), 6 ng/ml insulin (Sigma-Aldrich), and 5% FBS. All cells were cultured at 37C in a humidified incubator with an atmosphere of 5% CO₂.

RNA Isolation and Real-time Quantitative PCR

For each biopsy, 10 tissue slices with a thickness of 6 to 8 µm were sectioned on a cryostat and dissolved in Trizol (Life technologies). Total RNA was isolated by a spin column method (Zymo Research, CA). The quality and integrity of the extracted RNA was evaluated by a spectrophotometer (Nanodrop 1000). RNAs were reverse transcribed using the High Capacity RNA-to-cDNA Kit (Applied Biosystems, Denmark). Quantified real-time PCR was performed using Taqman gene expression assays (Applied Biosystems) on Bio-Rad CFX manager 3.0 and thermocycler (Bio-Rad, Denmark). Expression of KRT10 (Hs00166289_m1) was normalized to GAPDH (Hs02758991_g1) and EPCAM (Hs00901885_m1).

Kaplan–Meier Plots

A web tool, KM-plotter (kmplot.com), was used to generate Kaplan–Meier plots based on data sets from more than 6000 breast cancer patients.¹⁸ Analysis of expression of KRT10 in the data sets was performed using the JetSet best probe set (Affymetrix ID: 213287_s_at). Expression data were trichotomized to compare the patients with the highest levels of KRT10 tumor expression (upper tertile) vs the lowest (lower tertile) for relapse-free survival (RFS) and overall survival (OS). Patients with ERα-positive and ERα-negative tumors were analyzed separately, as well as in total. Furthermore, more detailed subcluster analysis was performed by including parameters such as molecular subtype, tumor grade, lymph node status, and treatment modality. The number of patients analyzed can be found in Supplemental Table S4.

Results

Rare K10^pos Luminal Cells Are Most Frequently Present in Breast Gland Structures of Women >30 Years of Age

Here, we analyzed the expression of K10 in human normal breast glands by IHC on cryostat sections using a well-characterized antibody, clone DE-K10.^19,20 A minor subset of glandular structures demonstrated the presence of K10^pos cells. Typically, when present, K10^pos cells were located as single cells or in small foci as part of the luminal compartment (Fig. 1A). The staining pattern was validated by comparison with another often-used monoclonal antibody against K10, RKSE60,^21–23 in serial sections (Supplemental Fig. S1). The actual proportion of luminal cells expressing K10 was quantified on trypsinized cellular smears derived from breast gland tissue from five women (age range: 19–46 years). In all cases, less than 1% of luminal cells were positive (Supplemental Fig. S2).

Figure 1.

Rare K10^pos cells are most frequently present in breast glands of women >30 years of age. (A) Immunoperoxidase staining demonstrating expression of K10 in a small subset of luminal cells. Scale bar, 50 µm. (B) Proportion of biopsies with K10^pos cells in breast gland structures from women divided into two age groups. Note that the number of K10^pos biopsies are significantly higher in women aged >30 years (10 of 23) compared with those aged <30 years (3 of 23) (Pearson’s Chi-square test, p<0.05). Abbreviation: K10^pos, K10-positive.

To obtain a more comprehensive expression pattern of K10 in the normal breast gland, tissue sections from 45 women aged 13 to 74 years were immunostained. A total of 956 glandular structures were evaluated, of which 54 (5.6%) contained K10^pos cells (Supplemental Table S1). Overall, K10^pos structures showed a fairly even distribution between lobules (38 of 604 lobules, 6.3%) and ducts (16 of 352 ducts, 4.5%) (Table 2; Supplemental Table S1).

Table 2.

Expression of K10 in the Normal Breast Gland by Immunohistochemistry.

Fraction of Structures With K10pos Cells
	n	%
Lobules	38/604	6.3
Ducts	16/352	4.5
Total	54/956	5.6

Abbreviation: K10^pos, K10-positive.

Dividing the biopsies into two age groups revealed that for women <30 years, 3 of 23 biopsies (13.0%) had K10^pos structures, whereas they were present in 10 of 23 biopsies (43.5%) from women >30 years (Fig. 1B, p<0.05 by Pearson’s Chi-square test).

Thus, rare K10^pos luminal cells are more frequently found in normal breast gland structures in women >30 years of age.

K10^pos Cells Belong to a Mature Luminal Compartment

We have previously used markers that define various differentiation states within the luminal lineage by IHC.^6,13,24 Whereas cKIT defines luminal progenitor cells,^14,24–26 Ks20.8 is a marker that encompasses a mature luminal compartment expressing ERα, GATA3, and CD166.¹³

In tissue sections double-stained for K10 and cKIT by immunofluorescence, the two markers generally segregated into different cells (Fig. 2A). Moreover, when staining for K10 and Ks20.8, the two markers co-stained, albeit with Ks20.8 more widely present in luminal cells (Fig. 2B). Furthermore, no K10^pos cells were found to express the cell cycle marker Ki67 (Fig. 2C). Due to the low frequency of K10^pos cells in the tissue sections, it was not possible to estimate the actual frequency of cells in cycle in comparison with luminal cells in general. Instead, a normal-derived breast epithelial cell line MCF10A containing a minor population of K10^pos cells was tested. Although around 88% of MCF10A cells in general were cycling by evaluating Ki67, only around 9% of K10^pos cells expressed Ki67 (Supplemental Fig. S3).

Figure 2.

K10-positive (K10^pos) cells belong to a mature luminal compartment. Immunofluorescence on normal breast tissue sections stained for K10 (green) and (A) cKIT, (B) Ks20.8, and (C) Ki67 (red). Although staining of K10 and cKIT or Ki67 segregated into different cells, K10^pos cells were confined to a subset of Ks20.8^pos cells. Arrowheads indicate examples of cells that are single-positive (green or red) or double-positive (yellow). Scale bar, 50 µm.

The lack of cKIT, the presence of Ks20.8, and the low number of cycling K10^pos cells suggest that these cells belong to a subset of mature luminal cells in the normal breast gland.

K10^pos Breast Carcinomas Mimic a Normal Luminal Differentiation Pattern

As carcinomas often reflect the epithelium of origin with regard to keratin expression,¹ we stained a sample of breast carcinomas for K10 (Fig. 3B). In total, 22 of 135 tumors (16.3%) were positive (Fig. 3B; Supplemental Table S2). When subdividing the stained carcinomas into ERα-positive and ERα-negative, respectively, 17 of 84 (20.2%) and 5 of 51 (9.8%) were K10^pos (Fig. 3B; Supplemental Table S2). Further classification of tumors into molecular phenotypes revealed that 13 of 74 (17.6%) Luminal A, 4 of 10 (40%) Luminal B, 1 of 16 (6.3%) HER2-enriched, and 4 of 35 (11.4%) Triple-Negative were K10^pos. Detection of K10 by IHC was further corroborated by its gene expression in breast tumors verified by real-time quantitative reverse transcriptase PCR (Fig. 4).

Figure 3.

K10 is expressed in a subset of breast carciniomas. (A) Immunoperoxidase staining demonstrating heterogeneous expression of K10 in a breast carcinoma. Scale bar, 50 µm. (B) Proportion of K10^pos breast carcinomas in total (n=135), as well as those divided into ERα^pos (n=84) and ERα^neg (n=51). Abbreviations: ERα^pos, estrogen receptor α–positive; ERα^neg, estrogen receptor α–negative; K10^pos, K10-positive.

Figure 4.

The presence of K10 in breast carcinomas by immunohistochemistry correlates to the expression of KRT10 transcript. A selection of five K10^pos and five K10^neg tumors (top panel) were analyzed for relative KRT10 expression by real-time quantitative reverse transcriptase PCR. Scale bar, 50 µm. Abbreviations: K10^pos, K10-positive; K10^neg, K10-negative.

To analyze whether K10^pos cells in breast tumors retain a phenotype mimicking a mature luminal differentiation program found in normal tissue, we verified the expression of cKIT and Ks20.8 in all K10^pos tumors. Nineteen of 22 tumors were negative for luminal progenitor marker cKIT (Supplemental Table S3). None of the three cKIT-positive tumors coexpressed K10 and cKIT when tested by immunofluorescence (Fig. 5A).

Figure 5.

K10-positive (K10^pos) breast carcinomas mimic a normal luminal differentiation pattern. Immunofluorescence on breast cancer tissue sections stained for K10 (green) and (A) cKIT, (B) Ks20.8, and (C) Ki67 (red). Like in normal tissue, staining of K10 and cKIT segregated into different cells, whereas K10 stained a subset of Ks20.8^pos cells. Ki67 rarely overlapped with K10^pos cells. Arrowheads mark examples of cells that are single-positive (green or red) or double-positive (yellow). Scale bar, 50µm.

Of the 22 K10^pos tumors, 21 (95%) also expressed the mature luminal marker Ks20.8 (Supplemental Table S3). Testing a sample of four Ks20.8-positive tumors revealed that K10^pos cells generally also expressed Ks20.8 (Fig. 5B).

Furthermore, the presence of cell cycle marker Ki67 was less frequent in K10^pos cells compared with the overall fraction of cycling neoplastic cells when performing quantification on three tumor biopsies (Supplemental Fig. S4A–C). Immunostaining of breast cancer cell line MCF7 revealed a similar pattern (Supplemental Fig. S4D).

Overall, K10^pos cells in breast carcinomas retain features that resemble mature luminal cells of the normal breast gland.

Expression of KRT10 Correlates to a Worse Outcome for Breast Cancer Patients

Using a web-based tool and publicly available data,²⁷ we compared breast cancer patients with the highest and lowest levels of expression of the K10 gene, KRT10. Surprisingly, the data revealed that when compared with low levels of KRT10 in breast tumors, high levels of KRT10 correlated to a worse RFS as well as OS for breast cancer patients (Fig. 6; Supplemental Table S4A and B). This correlation was evident for both ERα-positive and ERα-negative tumors. Further subsegmentation of the available data revealed similar correlation within the Luminal A subtype, for grade 3 tumors, and for lymph node–positive tumors that were ER-negative (Supplemental Fig. S5 and Table S4C).

Figure 6.

Higher levels of KRT10 in breast cancer correlates to a worse outcome. Kaplan–Meier plots of RFS and OS of breast cancer patients stratified by RNA expression of KRT10 using a web-based survival analysis tool (kmplot.com). Patients were trichotomized into the highest tertile (red) and lowest tertile (black) for KRT10 expression and are shown for all tumors (All), ERα^pos, and ERα^neg. Abbreviations: ERα^pos, estrogen receptor α–positive; ERα^neg, estrogen receptor α–negative; OS, overall survival; RFS, relapse-free survival; HR, hazard ratio.

Thus, transcriptome analysis suggests that K10 is a marker for poor prognosis in breast cancer.

Discussion

In epidermis, K10 is a marker of cells that exit cell cycle to enter terminal differentiation.^28,29 Thus, it is tempting to speculate that the presence of K10 in cells of the breast gland reflects a corresponding placement in a differentiation hierarchy. The features of K10^pos luminal cells of the breast do support a placing in a similar differentiated compartment, as they are negative for cKIT, positive for Ks20.8, and generally non-cycling. However, the low frequency of K10^pos cells can also lead to speculations that formation of these cells is an aberrant reminiscence of epidermal differentiation rather than a functional part of the normal differentiation program of the breast gland. In culture-based assays, myoepithelial cells derived from normal breast glands can undergo squamous cell differentiation, expressing K10.^14,30 Although rare, squamous cell metaplasia can also be found in the breast gland, often with prominent histological features.^31,32 However, when identified in the sample set of normal-derived biopsies that was analyzed in this study, K10^pos cells did not accumulate in characteristic metaplastic structures. Rather, K10^pos cells manifested themselves as infrequent single cells or distinct foci embedded as part of the normal luminal layer of glandular structures. Further investigations are required to determine whether the presence of K10 correlates to underlying metaplastic changes.

As age is the only disclosed information on the donors of the breast tissue in this study, other possible confounding events such as parity cannot fully be taken into account. In Denmark, the average age of primiparous women is 29 years,¹² so the presence of an increased frequency of K10^pos breast gland structures in women >30 years of age may be a reflection of age-related accumulation, of parity, or both. Although the majority of women with K10^pos breast gland structures were in the age range of 43 to 53 years, the present material does not allow us to conclude whether the frequency drops in postmenopausal women.

Although it has been suggested that K10 can be expressed in breast cancer,¹⁰ previously published studies indicate that it is a rare event at most.^9,33,34 A more recent reporting on K10 expression in breast tumors was only concerning rare types of carcinomas with clear squamous differentiation.¹¹ The fact that overall around 16% of breast carcinomas in our sample set were K10^pos clearly demonstrates that the expression of K10 is not confined to metaplastic tumors. The relatively frequent observation of K10^pos carcinomas by immunohistochemical stainings may relate to our use of cryosections rather than paraffin-embedded tissue, which could reduce antigen masking. Interestingly, K10^pos tumors were found in both ERα-positive and ERα-negative subgroups. However, as all but one of the 22 K10^pos carcinomas were also positive for luminal differentiation marker Ks20.8, it could be indicative of a certain stage of differentiation. Although based on a very small sample set of only 10 tumors, it does stand out that 40% of Luminal B carcinomas were K10^pos.

As the expression of K10 correlates to a more differentiated phenotype in normal tissue, it is somewhat surprising that apparently breast cancer with high KRT10 expression generally leads to a worse outcome than tumors with low levels of KRT10. We have previously reported on the presence of increased tumor-initiating capacity in a more differentiated subset of cells within a cancer stem cell hierarchy.^35,36 Also, the fact that most of K10^pos tumors demonstrate considerable intratumoral heterogeneity with respect to this marker may reflect the phenotypic plasticity of the tumor cells rather than the differentiation level of the tumor per se. It should be noted that in hepatocellular carcinoma it has been shown that K10 actually is a marker for poor prognosis.³⁷ Based on this, we suggest that further investigations on the role of K10 in breast cancer are initiated.