Differentiation (1990) 43 : 146-1 56
Different ialion Ontogeny and Neoplasia 0 Springer-Verlag 1990
Regulation of vimentin expression in cultured human mammary epithelial cells Christina Msrk', Bo van Deursl, and Ole William PetersenL.2.* Structural Cell Biology Unit, Department of Anatomy, Panum Institute, University of Copenhagen, Blegdamsvej 3, DK-2200 Copenhagen N, Denmark Laboratory of Tumor Endocrinology, Fibiger Institute, DK-2100 Copenhagen 0, Denmark
'
Accepted in revised form March 3, 1990
Abstract. Using five different monoclonal antibodies to vimentin, we have examined the expression of vimentin in cryostat sections and serum-free cultures of normal human breast tissue. In cryostat sections, myoepithelial cells as well as stromal cells showed immunoreactivity to vimentin, irrespective of the antibody used. In contrast, lumiiial epithelial cells were negative for vimentin, but positive for keratin K18. In culture, myoepithelial cells showed immunoreactivity to vimentin from their first appearance in monolayer. Moreover, a fraction of luminal epithelial cells expressed vimentin in addition to keratin K18. We found a clear, reversible correlation between proliferation, determined by incorporation of [3H]-TdR, and induction of vimentin in the luminal epithelial cells. Thus, in growth-stimulated cultures on a medium containing cholera toxin (CT), epidermal growth factor (EGF), transferrin (TO, hydrocortisone (H) and insulin (I), the fraction of vimentin-positive luminal epithelial cells increased, while it decreased within 14 days from approximately 36% to 3% on a medium containing CT and EGF, only. We therefore conclude: (1) vimentin is constantly expressed in myoepithelial cells in situ and in vitro, and (2) expression of vimentin in luminal epithelial cells in vitro is not a result of monolayer cultivation as such, but rather associated with the increased growth rate seen in culture.
Introduction Vimentin is a 57-kDa intermediate filament protein expressed in stromal cells, whereas epithelial cells are characterized by expression of keratins [38]. Only a few examples are known in which epithelial cells express both keratins and vimentin in situ. These include mesothelial cells, nucleus pulposus cells, epithelial cells of the human ovary, of the fallopian tubes, in the synovial tissue, in the thyroid gland, some epithelial cells during human
* To whom offprint requests should be sent
fetal development and a number of epithelial tumors [2, 7, 10, 12, 24, 29, 33, 34, 36, 37, 58, 60, 641. The human mammary gland comprises two major epithelial cell types: luminal epithelial cells and myoepithelial cells. Luminal epithelial cells of the resting mammary gland are characterized by expression of keratins K7, K8, K16, K18 and K19, while myoepithelial cells express keratins K5, K7, K14 and K17 [4, 23, 28, 38, 39, 40,44, 61, 671. Numerous contradictory observations have been reported with respect to vimentin expression in human breast tissue. Thus, while there has been general agreement on the expression of vimentin in the mammary stromal cells and on the absence of vimentin in the luminal epithelial cells, conflicting results have emerged with respect to vimentin expression in myoepithelial cells in mammary glands of different species [3, 11, 13, 15, 21, 23, 27,28,47, 691. Regulation of vimentin expression has been analysed mostly in culture model systems. It seems clear from these studies that vimentin may appear in many types of cultured cells not expressing this filament in situ, as in various types of epithelial cells. Experiments to prevent this induced expression in culture have suggested that it is primarily related to the structural changes associated with plating the cells into a monolayer and the exponential growth not normally seen in situ [9, 13, 21, 51, 52, 59, 65, 681. Another possibility with particular reference to the human mammary gland and other tissues containing a mixed population of vimentin-positive and vimentin-negative cells is that negative cells may give rise to positive cells through intermediate cell types and that this process is accelerated in rapidly proliferating cultures. It has been reported that vimentin expression is induced in cultured mammary epithelial cells as early as 3-4 days after plating, compared to an absence of vimentin expression among the cells in situ [ 1 1, 131. Recently, progress has been made in the study of mammary epithelial cell differentiation by using serumfree culture conditions, with which it is possible to manipulate the cell cycle status as well as the growth rate simply by omission of potent mitogens [47]. With the
147
present level of culture technology it should be possible to determine which of the above-mentioned possibilities cause the induction of vimentin filaments in cultured mammary epithelial cells. The aim of the present study has therefore been : (a) to clarify whether or not myoepithelial cells in situ express vimentin, using a broad panel of monoclonal antibodies to vimentin and immunoperoxidase cytochemistry on cryostat sections of human breast tissue; and (b) to analyse the kinetics of vimentin expression in cultured human mammary epithelial cells using a growth-factor-defined culture model system.
Methods Cell cultures. Primary and extended cultures were established from human breast tissue obtained during reduction mammoplasty as previously described [45, 46, 491. The tissue specimen was transferred directly after surgery to a chemically defined basal medium (DME-F12) at room temperature, and within 1 h disaggregated mechanically in phosphate-buffered saline (PBS) with opposing scalpels into 1- to 2-mm fragments. The tissue fragments were then incubated overnight on a rotary shaker (40-60 rpm) at 37" C in 6 ml DME-F12 supplemented by 900 IU/ml Collagenase C 0130 (Sigma Chemical Co., St. Louis, Missouri, USA) in the absence of any growth factors or hormones. This procedure yielded a suspension of epithelial organoids, ix., intact acini and ductules lacking the surrounding stroma and basement membranes [45, 46, 491. The experimental medium for initial characterization of cells consisted of DME-F12 (1 : 1) supplemented by growth factors and hormones. These additives included: 250 ng/ml (3 pg/ml for stimulation of growth-arrested cells) insulin (Nordisk Insulin Laboratory, Copenhagen, Denmark), 100 ng/ml epidermal growth factor, 0.5 pg/ml (1.38 phf) hydrocortisone (both from Collaborative Research, Bedford, Mass, USA), 25 pg/ml transferrin, 10 ng/ml cholera toxin, 2 m M glutamine (all from Sigma). Cells were cultured in T-25 flasks (Nunc, Roskilde, Denmark) coated with 200 pg Vitrogen (Flow Lab., Herts, U K , 1461) per flask. All media were supplemented with 250 IU/ml penicillin (kindly provided by Leo Pharmaceuticals, Ballerup, Denmark) and 25 pg/ml streptomycin (Bie and Berntsen, Rudovre, Denmark). Culture media were always prepared on the day of use, and the media were changed every 2-3 days, and kept in a humidified atmosphere of 75% N,, 20% o,, 5% c o , . The cultured cells were passaged using 0.25% trypsin-1 m M EDTA in PBS. The cells were briefly exposed to trypsin, then the trypsin was almost removed, and the cells incubated for 10 rnin at 37" C, only covered by a thin layer of trypsin. Finally 30 pl 1% soyabean trypsin inhibitor (Sigma) was added and the cells were resuspended in medium and passaged. To obtain sections of organoids with preserved three-dimensional structural organization, cultures were rinsed twice with 2.5 ml Hepes at room temperature (rt), flasks were cut open and cells removed from the bottom with a policeman. The cells were then resuspended in H ep es t 10% normal goat serum (NGS), centrifuged for 5 min, 1000 rpm, and the pellet was stored at -80" C until use. Zmmunocytochernistry. All the primary antibodies used were monoclonal antibodies (mAbs), and most of them have been extensively characterized. These included: anti-keratin K8 (1 :10, R P N . f f 6 6 , Amersham, Buckinghamshire, UK, [ 5 , 321, anti-keratin K5 + K8 (1 : 10, clone SB2, 3008, Sanbio, AM UDEN, Holland, according to supply from Monosan), anti-keratin K8 f K 1 8 (undiluted, C A M 5.2, Becton Dickinson, Albertslund, Denmark, [6, 181, anti-keratin K14 (1 :200, clone CKBI, Sigma [S]), anti-keratin K18 (1 :50, clone M9, F3006, Sanbio, according to Moll classification [38]), anti-
keratin K18 (2:50, CK5, Sigma, 18, 62]), anti-keratin K18 ( I :loo, KS-B17.2, Sigma, [35]), anti-keratin K19 (1 : 50, M772, Dakopatts, Glostrup, Denmark, [4]), anti-vimentin (1 :400, M725, clone V9, [41], Dakopatts), anti-vimentin (1 : 10, Labsystems OY, Helsinki, SF, [43]), anti-vimentin (1 :200, MA907, Enzo Biochem Inc., NY, USA, [13]), anti-vimentin (1 :loo, clone 13.2, Sigma, [61]), antivimentin (1 : 50, VZMS, Medac, mbH, FRG). The second antibody used was a rabbit antibody to mouse immunoglobulin (1 :20, 2259, Dakopatts), and the third reagent was a monoclonal PAP-complex 1: 100, P850, Dakopatts). Cryostat sections (4 pin) of breast tissue and three-dimensional organoids were fixed in methanol for 5 rnin at -20" and incubated for another 5 rnin with 0.1% Triton-X 100 in PBS. Cultured cells were rinsed twice in 2 ml PBS at room temperature (rt), and then fixed for 5 rnin at -20" C in methanol. Then the cells were rinsed again in 2 ml PBS at rt, the flasks opened, incubation fields framed, and the cells dried for I5 rnin at rt. The fixed, dried cells or the sections were then incubated for 5 min in 50-200 p1 PBS+IO% NGS followed by 30 rnin in PBS-NGS supplemented with primary antibody, 30 min with secondary antibody, and 30 min with the tertiary antibody, all at rt. Between the antibody incubations, cells were rinsed twice in PBS at the same volume (50-200 pl) as used for the antibodies, except for the last rinse, in which the whole culture plate was rinsed in a PBS bath. Finally reaction-product formation was achieved by incubating for 10 min at rt using 0.5 mg/ml3,3-diaminobenzidine(DAB, D 5637, Sigma) in PBS supplemented with 0.5 pl/ml 30% H 2 0 2 . Double-incubation experiments were performed according to the "unlabeled antibody method" 1571. The first sequence of antibodies was incubated as described above using DAB-H,O, to yield a brown reaction product that masks antigens and immunoreagents of this sequence and with that prevents cross-binding of antibodies in the second sequence. The second sequence was carried out after exposing the cells twice for 5 min each to 0.1% Triton X-100 in PBS to remove any membrane-bound endogenous alkaline phosphatase. Next the cells were incubated with primary antibody (30 min), rabbit antibodies against mouse immunoglobulins (1 :20, 30 min, 2259, Dakopatts), followed by monoclonal APAAP (1 :20, 30 min, D651, Dakopatts). The reaction product was developed using a mixture of 1 mg naphtol AS-BI phosphate (Sigma) diluted in 50 pl n,n-dimethylformamide and 6 mg Fast Red Violet (Difco Laboratories, Surrey, UK) in 5 ml in 0.2 M Tris buffer (pH 8.3). Cultures incubated as described above, but with omission of the primary antibody, served as controls. To determine the percentage of cells stained with a particular mAb probe, the cells were scored using a bright-field microscope. Random fields were counted using an ocular grid and a 25 x objective. At least 400 cells in each category were counted. Nuclei were briefly counterstained with haematoxylin. Thymidine incorporation. [3H]-thymidine ([3HJ-TdR, Dupont, NEN-Products, Mass, USA) with a specific activity of 20 Ci/mmol was diluted in saline 1 :40. Cell cultures were rinsed once in medium without thymidine (GI-MEM), incubated for 1 h with 5 ml GIMEM+ CT-EGF and 0.5 ml [3H]-TdR, and then the immunocytochemical procedure was performed as described above. All cultures for autoradiography were coated with Ilford K.2 emulsion (Ilford, UK) and exposed for 1 week. As control served a new cell line from this laboratory. This cell line, HMT-3909S8, which is known to express vimentin [50], was incubated with ['HITdR followed by peroxidase or [ 'HI-TdR followed by alkaline phosphatase to ensure that the immunocytochemistry procedure did not influence the [3H]-TdR incorporation. After being developed the cultures were counterstained with haematoxylin, and the number of [3H]-TdR-labeled cells was estimated using bright-field microscopy. Random fields were counted using an ocular grid and a 25 x objective. Gel electrophoresis and immunohlotting. Preparations of intermediate filaments were made according to the method of Achtstatter [11. Then SDS polyacrylamide gel electrophoresis was performed
148 using 10%-17% polyacrylamide gradient gel [31]. Next, the proteins were transferred to Immobilon Transfer Membrane (Millipore), essentially by the method of Towbin [63]. Vimentin bands were identified immunochemically. Briefly, lanes cut from the membrane were blocked with PBS -t 2.0% Tween 80 (Sigma) for 5 min, incubated for 19-20 h with the specific monoclonal antibodies ( V I M S 1 : 50, MA907 1 : 100, M725 1 :400, Labsystems 1 : 10, F3006 1 : 50, all diluted in incubation buffer (50 m M Tris. 150 m M NaCl+0.05% Tween SO), rinsed five times for 5 min each with washing buffer (50 mM Tris, 150 m M NaCI), incubated for 2 h with peroxidase-conjugated rabbit anti-mouse Ig (P260 1 : 850 in incubation buffer, Dakopatts), rinsed five times for 5 rnin each in washing buffer, incubated for 5 niin in 50 m M Na-acetate (pH 5.2) and then stained. Substrate and colour reagents were 20 mg 3 amino-9 ethylcarbazole (Sigma), 2.5 ml N,N-dimethylformamide (Merck, FRG) in 50 ml 50 m M Na-acetate plus 25 p1 30% H A . Lanes with molecular-weight markers (Bio-Rad Laboratories, Richmond, Calif, USA) were rinsed four times for 5 min each with PBS+0.3% Tween 20 (Merck, FRG) and stained with India Ink (1 : 1000 in PBS-TW, Staedtler marsmatic 745, FRG) for 2 h.
Results Immunoperoxidase cytochemistry on methanol-fixed cryostat sections of human breast tissue revealed a rather consistent pattern of reactivity using five different monoclonal antibodies to vimentin (Table 1). Thus, immunoreactivity was seen in almost all myoepithelial cells in both interlobular ducts and terminal duct lobular units (TDLU); (Fig. 1). However, the intensity of reactivity varied somewhat, the weakest being seen in sections incubated with the IgM class of monoclonal antibodies (Table 1). Whereas no immunoreaction to vimentin was observed in luminal epithelial cells, either in interlobular ducts or in TDLU, these cells showed 100% immunoreactivity when incubated with antibody to keratin K18 (Fig. 1). All antibodies to vimentin showed consistent reactivity with all stromal cell types, while no reactivity was seen with antibodies to keratin K18 (Fig. 1). Similar patterns of reactivity were seen no matter whether the tissue was unfixed, fixed in methanol, in methanol-acetone, or in acetone alone (not shown), but fixation in formaldehyde or acetic acid-containing fixatives strongly reduced the immunoreactivity. It is therefore concluded that vimentin is expressed by myoepithelial cells in situ
as contrasted by absence of reactivity in luminal epithelial cells. Next, we analysed the kinetics of vimentin expression in mammary epithelial cells from freshly explanted tissue in primary culture. All of the monoclonal antibodies used for the in vitro analyses showed only one band in the 50-60 kDa range after gel electrophoresis and immunoblotting, indicating monospecificity for vimentin (Fig. 2). The collagenase digest was plated as epithelial cell clumps (organoids) in serum-free medium (Fig. 3). The first cell type known to appear in monolayer from the organoids is the myoepithelial cell [47] (Fig. 3). As expected from the in situ analysis, all myoepithelial cells (100%) showed immunoreactivity for vimentin from their first appearance in monolayer, i.e. on day 3 (Fig. 3). The myoepithelial cells were easily distinguished from stromal cells, since the myoepithelial cells showed strong immunoreactivity for keratins K5 and K14. Also, these cells were immunoreactive with antibodies to a-smooth muscle iso-actin (not shown) [47,48]. In theory, the first vimentin-positive cell type to appear in culture could be luminal epithelial cells, in which vimentin was induced due to the transition into a twodimensional monolayer. To exclude this possibility, we incubated sections of organoids, in which the cells showed preserved three-dimensional structural organization. These experiments showed that myoepithelial cells, also inside the organoids, were immunoreactive for vimentin, whereas the luminal epithelial cells remained vimentin-negative (not shown). This indicates that the myoepithelial cells continued to express vimentin from the situation in situ through explantation and monolayer cultivation. In addition, there were no signs of downregulation of vimentin expression in myoepithelial cells, even after prolonged cultivation or passaging of the cells. After 3 days of cultivation, the central zone of small organoids was essentially free of migrating myoepithelial cells, leaving behind an almost pure population of luminal epithelial cells (Fig. 3). We now analysed the kinetics of vimentin expression in this population of luminal epithelial cells. Luminal epithelial cells were identified using three different antibodies to keratin K18, and by additional antibodies to keratin K8 and K19. No reactivity was observed in the myoepithelial cells (not shown). To
Table 1. Antibodies: In situ myoepithelial cells (MEP) showed immunoreactivity with all the vimentin antibodies used, while luminal epithelial cells (LEP) were vimentin-negative, except for a weak nonfilamentous reaction obtained mAb M725. In vitro all antibodies gave a positive reaction in MEP, and in addition a fraction of LEP were positive for vimentin as well; cc, anti Anti-
Code
Clone
Company
Specics
Ig-subtypc
Dilution
In situ
In vitro -~
~~
Vimentin
M725 MA907 V5255 VIMS
-
V9 -
VIM 13.2 -
Weak nonfilamentous reaction Weak reaction ' Not used in vitro
Dakopatts, DK Labsystems, SF Enzo, USA Sigma, USA Medac, F R G
Mouse cc-swine Mouse cc-human Mouse cc-human Mouse cc-human Mousc w o w
IgG, IgG, IgM IgM k G1
1:400 1:lO 1 :200 1:lOO 1 :50
LEP
MEP
LEP
(+)" -
+
(+)"
~
+ (+y + ?' + + +
(+)h
MEP
+
+ + ?' +
149
Fig. 1a-d. Two consecutive cryostat sections (a and c ; details shown in b and d) of human breast tissue fixed in methanol. Sections were incubated immunocytochemically under conditions identical to those used for cultured cells. When incubated with antibody to vimentin (a, c). immunoreactivity was obtained in all rnyoepithelial cells ( M E P ) and in stromal cells ( S o ,while luniiiial epithelial cells (LEP) were negative. On the other hand luminal a
b
c
d
e
f
k
91 66-
4331-
2 2-
1li-
Fig. 2. Immunoblots showing that the various antibodies used are specific for vimentin. Lane a, molecular-weight markers; lane b, VIMS 1:50; lane c, MA907 3:lOO; lane d, A4725 1:400; lane e, LabSystems 1 :SO; lunef, control
epithelial cells showed immunoreactivity with antibodies to keratin K18 (b, d). As controls served sections incubated as described in Methods, but with omision of the primary antibody. In thcse sections no staining occurred. a, c Vimentin monoclonal antibody (mAb) VIMS. b, d MAb to kcratin K18, F3006. Bur, 100 ym (a, b), 50 pm (c, d)
prevent any confusion between myoepithelial cells and luminal epithelial cells during this analysis, we performed a double incubation of the cultures to demonstrate both keratin K18 - specific for luminal epithelial cells - and vimentin. Also, we measured the growth rate by [3H]-thymidine ([3H]-TdR) incorporation. To ensure that the immunocytochemical incubation did not perturb the subsequent autoradiography procedure, we incubated cultures from a cell line known to express vimentin with [3H]-TdR followed by peroxidase or [3H]TdR followed by alkaline phosphatase, and compared it with a control incubated with [3H]-TdR alone. No significant difference was found. The experiment revealed that the number of luminal epithelial cells that were positive for vimentin increased significantly within 7 days of cultivation (Fig. 4a). Thereafter a slight decrease in the frequency was observed corresponding to the time when cells reached confluency (Fig. 4a). A time-sequence analysis of the [3H]TdR incorporation showed that the luminal epithelial cell population exhibited a growth rate pattern very similar to the pattern of vimentin immunoreactivity (Fig. 4a). However, using primary cultures, the standard
150
Fig. 3a, b. Light micrographs of primary cultures stained with immunoperoxidase to demonstrate vimentin (mAb V/Ms). a After 3 days of cultivation the remnants of organoids ( O R ) are surrounded by a central zone of small, compact luminal epithelial cells (LEP) essentially free of migrating myoepithelial cells. Immun-
ocytochemistry with mAb VZMS revealed that all the myoepithelial cells (MEP) contained vimentin, while the luminal epithelial cells were all negative. b Higher magnification of myoepithelial cells stained with mAb VIMS. Distinct filament bundles are present in all the myoepithelial cells. Bar, 50 pm (a), 25 pm (b)
deviations on our vimentin immunoreactivity data were too high, making a firm interpretation of the results difficult (Fig. 4a). Therefore, another experiment was performed, in which the cells were passaged on day 5, allowing a clear separation of the luminal epithelial and myoepithelial cells in monolayer. This experiment essentially confirmed our initial observations (Fig. 4b). On the basis of these experiments, we conclude that vimentin is actually induced in luminal epithelial cells in monolayer culture. This induction appears to correlate with the growth rate of the cells. Next, we questioned to what extent the transition into monolayer and the growth rate, respectively, contributed to the induction of vimentin in luminal epithelial cells. One approach to answering this question was to arrest the cells by mitogen depletion of the culture medium, thereby simulating the growth situation in vivo. Exponentially growing cells in primary culture were passaged into a culture medium containing only CT and EGF, and growth was measured using [3H]-TdR incorporation along with double-labeling immunocytochemistry for vimentin and keratin K18 (Fig. 5). As seen in Fig. 6a, vimentin immunoreactivity in luminal epithelial cells - defined by the presence of keratin K1S - decreased gradually from approximately 37% on day 2 to 3% on day 16 in medium containing CT-EGF only. No further
reduction was obtained by extending the culture period for up to 4 weeks. By comparing the kinetics of vimentin immunoreactivity in luminal epithelial cells during the first 16 days of cultivation in secondary cultures with the frequency of cells incorporating [3H]-TdR, it is evident that growth declines with the same kinetics as the vimentin expression (Fig. 6a). This association between growth and vimentin expression among lurninal epithelial cells is not universal among cultured mammary epithelial cells, since the myoepithelial cells under identical growth conditions remained vimentin-positive (Fig. 6 b). Even though vimentin was induced in proliferating luminal epithelial cells, the immunoreactivity mostly did not resemble the typical filamentous patterns seen in myoepithelial cells. In the majority of vimentin-positive lumjnal epithelial cells, the reaction appeared as punctate or fragmented filaments (Fig. 7). Based on this experiment, we conclude that it is possible to revert the vimentin expression induced in cultured luminal epithelial cells to almost zero. Therefore, the transition of cells into monolayer cultivation is not strictly associated with the acquisition of vimentin filaments. Rather, this phenomenon seems to be tightly coupled to the rate of growth in these cells. We then questioned whether luminal epithelial cells were able to enter the cell cycle and proliferate in the
151 30 15
I
2
a 70
6 TIME (days)
0
I
0
b
4
0 10
I 30
5
A
10
15
pression may have been present earlier at the transcriptional level. Finally we measured the frequency of cells that had incorporated [3H]-TdR cells among exponentially growing vimentin-positive and vimentin-negative luminal epithelial cells. This experiment was performed on secondary cultures, 2 days after they had been growth-arrested, and it revealed that 56.5% of the vimentin-positive luminal epithelial cells were labeled with [3H]-TdR, while 11.0% of the vimentin-negative cells were [3H]-TdR labeled. Thus, even though the vimentin-positive luminal epithelial cells were growing faster, there was a distinct fraction of vimentin-negative luminal cells growing as well.
20
TlME(days)
Fig. 4a, b. Relationship between growth rate (- - - -0- - - -) and vimentin-positive luminal epithelial cells (---D----) grown in DME-F12 containing cholera toxin (CT), epidermal growth factor (EGF), transferrin (Tf), hydrocortisone (H), and insulin (I). Each point represents the mean frequency of at least 4 times 100 cells. a In primary cultures the number of luminal epithelial cells expressing both keratin K18 and vimentin showed a significant increase within 7 days of cultivation. Thereafter, a minor decrease of double-labeled luminal epithelial cells was observed corresponding to the time when cultures reached confluency. From [3H]-TdR incorporation it was clear that the growth rate pattern correlate very well with the pattern of double-labeled luminal epithelial cells. However, standard deviations were very high using primary cultures. Therefore secondary cultures were used in b. Primary cultures were grown in DME-F12 containing CT, EGF, Tf, H and I and passaged at day 5 (arrowhead). As seen in a the frequency of double-labeled cells increased within 7 days after passaging, followed by a decrease, when cultures reached confluency. The growth rate curve ([3H]-TdR incorporation) showed a similar pattern
absence of vimentin expression. We therefore stimulated luminal epithelial cells that had been growth-arrested in DME-F12 containing only CT-EGF. As seen in Fig. 8, we found a remarkable increase in the frequency of [3H]-TdR-labeled cells within the first 6 days. During the same period, almost no increase in the frequency of vimentin-immunoreactive cells was noticed (Fig. 8), indicating that vimentin expression is not essential for the cells to enter the cell cycle. However, vimentin ex-
Discussion In the present study we have shown that in human breast tissue vimentin is expressed in myoepithelial cells and stromal cells, while it is absent in luminal epithelial cells. Primary cultivation of breast tissue did not alter the expression of vimentin in myoepithelial cells and stromal cells, whereas luminal epithelial cells acquired this phenotypic trait. The frequency of cultured luminal epithelial cells expressing vimentin in culture correlated with the growth rate. The observation that myoepithelial cells in interlobular ducts as well as in TDLU express vimentin is in accordance with results obtained on rat mammary gland [15, 271, and on the human mammary gland [47], but differs from most other studies on breast and salivary gland in which vimentin was found in stromal cells only [3, 11, 13, 20, 22, 28, 691. Also, it has been reported that vimentin is present primarily in myoepithelial cells of TDLU [23]. The variations in the reports on vimentin expression in mammary epithelial cells in situ may be due to the fact that tissue from different mammals was examined, different antibodies were used, and lactating glands were used in some of the studies [21, 27, 421. Other reasons for the discrepancies include the use of either frozen material or fixed and embedded material, the effects of different fixatives on the immunoreactivity of intermediate filaments, and the choice of immunocytochemical method [64]. In our study it is not surprising to find vimentin immunoreactivity in all of the cultured myoepithelial cells, since we find it in myoepithelial cells in situ as well. From studies using the antibody MA907, it has been reported that vimentin is induced upon cultivation in myoepithelial cells not expressing this phenotype in situ [13]. Using the same antibody we have found vimentin in myoepithelial cells in situ, indicating that the immunoreactivity seen in cultured cells does not reflect actual induction. We do not find it likely that the coexpression of keratin and vimentin in myoepithelial cells reflects an aberrant induction of keratins in stromal cells [30, 471. This is mainly based on our own analysis of characterized stromal cells in such cultures [47]. However the reactivity seen with mAb MA907 was weak in sections and in cultured cells. A few reports based on
152
Fig. Sa, b. Double incubation of cultured cells by the “Unlabeled Antibody Method” showing vimentin (mAb VfMS) with the brown DAB-H,O, reaction product in the first sequence and cytokeratin K18 with Fast Red Violet-alkaline phosphatase in the second sequence. The b1ur.k nuclei represent [3H]-TdR incorpora-
tion of proliferating cells. We found that myoepithelial cells ( M E P ) were positive with anti-vimentin, and luminal epithelial cells ( L E P ) were positive with anti-cytokeratin K18, but in addition there was a fraction of the luminal epithelial cells that showed immunoreactivity with both antibodies (LEP-D).Bar, 25 pm
153
a
-
TIME (days)
100
-
1
I,. ,
J
.
- 30
. F 2
Different ialion Ontogeny and Neoplasia 0 Springer-Verlag 1990
Regulation of vimentin expression in cultured human mammary epithelial cells Christina Msrk', Bo van Deursl, and Ole William PetersenL.2.* Structural Cell Biology Unit, Department of Anatomy, Panum Institute, University of Copenhagen, Blegdamsvej 3, DK-2200 Copenhagen N, Denmark Laboratory of Tumor Endocrinology, Fibiger Institute, DK-2100 Copenhagen 0, Denmark
'
Accepted in revised form March 3, 1990
Abstract. Using five different monoclonal antibodies to vimentin, we have examined the expression of vimentin in cryostat sections and serum-free cultures of normal human breast tissue. In cryostat sections, myoepithelial cells as well as stromal cells showed immunoreactivity to vimentin, irrespective of the antibody used. In contrast, lumiiial epithelial cells were negative for vimentin, but positive for keratin K18. In culture, myoepithelial cells showed immunoreactivity to vimentin from their first appearance in monolayer. Moreover, a fraction of luminal epithelial cells expressed vimentin in addition to keratin K18. We found a clear, reversible correlation between proliferation, determined by incorporation of [3H]-TdR, and induction of vimentin in the luminal epithelial cells. Thus, in growth-stimulated cultures on a medium containing cholera toxin (CT), epidermal growth factor (EGF), transferrin (TO, hydrocortisone (H) and insulin (I), the fraction of vimentin-positive luminal epithelial cells increased, while it decreased within 14 days from approximately 36% to 3% on a medium containing CT and EGF, only. We therefore conclude: (1) vimentin is constantly expressed in myoepithelial cells in situ and in vitro, and (2) expression of vimentin in luminal epithelial cells in vitro is not a result of monolayer cultivation as such, but rather associated with the increased growth rate seen in culture.
Introduction Vimentin is a 57-kDa intermediate filament protein expressed in stromal cells, whereas epithelial cells are characterized by expression of keratins [38]. Only a few examples are known in which epithelial cells express both keratins and vimentin in situ. These include mesothelial cells, nucleus pulposus cells, epithelial cells of the human ovary, of the fallopian tubes, in the synovial tissue, in the thyroid gland, some epithelial cells during human
* To whom offprint requests should be sent
fetal development and a number of epithelial tumors [2, 7, 10, 12, 24, 29, 33, 34, 36, 37, 58, 60, 641. The human mammary gland comprises two major epithelial cell types: luminal epithelial cells and myoepithelial cells. Luminal epithelial cells of the resting mammary gland are characterized by expression of keratins K7, K8, K16, K18 and K19, while myoepithelial cells express keratins K5, K7, K14 and K17 [4, 23, 28, 38, 39, 40,44, 61, 671. Numerous contradictory observations have been reported with respect to vimentin expression in human breast tissue. Thus, while there has been general agreement on the expression of vimentin in the mammary stromal cells and on the absence of vimentin in the luminal epithelial cells, conflicting results have emerged with respect to vimentin expression in myoepithelial cells in mammary glands of different species [3, 11, 13, 15, 21, 23, 27,28,47, 691. Regulation of vimentin expression has been analysed mostly in culture model systems. It seems clear from these studies that vimentin may appear in many types of cultured cells not expressing this filament in situ, as in various types of epithelial cells. Experiments to prevent this induced expression in culture have suggested that it is primarily related to the structural changes associated with plating the cells into a monolayer and the exponential growth not normally seen in situ [9, 13, 21, 51, 52, 59, 65, 681. Another possibility with particular reference to the human mammary gland and other tissues containing a mixed population of vimentin-positive and vimentin-negative cells is that negative cells may give rise to positive cells through intermediate cell types and that this process is accelerated in rapidly proliferating cultures. It has been reported that vimentin expression is induced in cultured mammary epithelial cells as early as 3-4 days after plating, compared to an absence of vimentin expression among the cells in situ [ 1 1, 131. Recently, progress has been made in the study of mammary epithelial cell differentiation by using serumfree culture conditions, with which it is possible to manipulate the cell cycle status as well as the growth rate simply by omission of potent mitogens [47]. With the
147
present level of culture technology it should be possible to determine which of the above-mentioned possibilities cause the induction of vimentin filaments in cultured mammary epithelial cells. The aim of the present study has therefore been : (a) to clarify whether or not myoepithelial cells in situ express vimentin, using a broad panel of monoclonal antibodies to vimentin and immunoperoxidase cytochemistry on cryostat sections of human breast tissue; and (b) to analyse the kinetics of vimentin expression in cultured human mammary epithelial cells using a growth-factor-defined culture model system.
Methods Cell cultures. Primary and extended cultures were established from human breast tissue obtained during reduction mammoplasty as previously described [45, 46, 491. The tissue specimen was transferred directly after surgery to a chemically defined basal medium (DME-F12) at room temperature, and within 1 h disaggregated mechanically in phosphate-buffered saline (PBS) with opposing scalpels into 1- to 2-mm fragments. The tissue fragments were then incubated overnight on a rotary shaker (40-60 rpm) at 37" C in 6 ml DME-F12 supplemented by 900 IU/ml Collagenase C 0130 (Sigma Chemical Co., St. Louis, Missouri, USA) in the absence of any growth factors or hormones. This procedure yielded a suspension of epithelial organoids, ix., intact acini and ductules lacking the surrounding stroma and basement membranes [45, 46, 491. The experimental medium for initial characterization of cells consisted of DME-F12 (1 : 1) supplemented by growth factors and hormones. These additives included: 250 ng/ml (3 pg/ml for stimulation of growth-arrested cells) insulin (Nordisk Insulin Laboratory, Copenhagen, Denmark), 100 ng/ml epidermal growth factor, 0.5 pg/ml (1.38 phf) hydrocortisone (both from Collaborative Research, Bedford, Mass, USA), 25 pg/ml transferrin, 10 ng/ml cholera toxin, 2 m M glutamine (all from Sigma). Cells were cultured in T-25 flasks (Nunc, Roskilde, Denmark) coated with 200 pg Vitrogen (Flow Lab., Herts, U K , 1461) per flask. All media were supplemented with 250 IU/ml penicillin (kindly provided by Leo Pharmaceuticals, Ballerup, Denmark) and 25 pg/ml streptomycin (Bie and Berntsen, Rudovre, Denmark). Culture media were always prepared on the day of use, and the media were changed every 2-3 days, and kept in a humidified atmosphere of 75% N,, 20% o,, 5% c o , . The cultured cells were passaged using 0.25% trypsin-1 m M EDTA in PBS. The cells were briefly exposed to trypsin, then the trypsin was almost removed, and the cells incubated for 10 rnin at 37" C, only covered by a thin layer of trypsin. Finally 30 pl 1% soyabean trypsin inhibitor (Sigma) was added and the cells were resuspended in medium and passaged. To obtain sections of organoids with preserved three-dimensional structural organization, cultures were rinsed twice with 2.5 ml Hepes at room temperature (rt), flasks were cut open and cells removed from the bottom with a policeman. The cells were then resuspended in H ep es t 10% normal goat serum (NGS), centrifuged for 5 min, 1000 rpm, and the pellet was stored at -80" C until use. Zmmunocytochernistry. All the primary antibodies used were monoclonal antibodies (mAbs), and most of them have been extensively characterized. These included: anti-keratin K8 (1 :10, R P N . f f 6 6 , Amersham, Buckinghamshire, UK, [ 5 , 321, anti-keratin K5 + K8 (1 : 10, clone SB2, 3008, Sanbio, AM UDEN, Holland, according to supply from Monosan), anti-keratin K8 f K 1 8 (undiluted, C A M 5.2, Becton Dickinson, Albertslund, Denmark, [6, 181, anti-keratin K14 (1 :200, clone CKBI, Sigma [S]), anti-keratin K18 (1 :50, clone M9, F3006, Sanbio, according to Moll classification [38]), anti-
keratin K18 (2:50, CK5, Sigma, 18, 62]), anti-keratin K18 ( I :loo, KS-B17.2, Sigma, [35]), anti-keratin K19 (1 : 50, M772, Dakopatts, Glostrup, Denmark, [4]), anti-vimentin (1 :400, M725, clone V9, [41], Dakopatts), anti-vimentin (1 : 10, Labsystems OY, Helsinki, SF, [43]), anti-vimentin (1 :200, MA907, Enzo Biochem Inc., NY, USA, [13]), anti-vimentin (1 :loo, clone 13.2, Sigma, [61]), antivimentin (1 : 50, VZMS, Medac, mbH, FRG). The second antibody used was a rabbit antibody to mouse immunoglobulin (1 :20, 2259, Dakopatts), and the third reagent was a monoclonal PAP-complex 1: 100, P850, Dakopatts). Cryostat sections (4 pin) of breast tissue and three-dimensional organoids were fixed in methanol for 5 rnin at -20" and incubated for another 5 rnin with 0.1% Triton-X 100 in PBS. Cultured cells were rinsed twice in 2 ml PBS at room temperature (rt), and then fixed for 5 rnin at -20" C in methanol. Then the cells were rinsed again in 2 ml PBS at rt, the flasks opened, incubation fields framed, and the cells dried for I5 rnin at rt. The fixed, dried cells or the sections were then incubated for 5 min in 50-200 p1 PBS+IO% NGS followed by 30 rnin in PBS-NGS supplemented with primary antibody, 30 min with secondary antibody, and 30 min with the tertiary antibody, all at rt. Between the antibody incubations, cells were rinsed twice in PBS at the same volume (50-200 pl) as used for the antibodies, except for the last rinse, in which the whole culture plate was rinsed in a PBS bath. Finally reaction-product formation was achieved by incubating for 10 min at rt using 0.5 mg/ml3,3-diaminobenzidine(DAB, D 5637, Sigma) in PBS supplemented with 0.5 pl/ml 30% H 2 0 2 . Double-incubation experiments were performed according to the "unlabeled antibody method" 1571. The first sequence of antibodies was incubated as described above using DAB-H,O, to yield a brown reaction product that masks antigens and immunoreagents of this sequence and with that prevents cross-binding of antibodies in the second sequence. The second sequence was carried out after exposing the cells twice for 5 min each to 0.1% Triton X-100 in PBS to remove any membrane-bound endogenous alkaline phosphatase. Next the cells were incubated with primary antibody (30 min), rabbit antibodies against mouse immunoglobulins (1 :20, 30 min, 2259, Dakopatts), followed by monoclonal APAAP (1 :20, 30 min, D651, Dakopatts). The reaction product was developed using a mixture of 1 mg naphtol AS-BI phosphate (Sigma) diluted in 50 pl n,n-dimethylformamide and 6 mg Fast Red Violet (Difco Laboratories, Surrey, UK) in 5 ml in 0.2 M Tris buffer (pH 8.3). Cultures incubated as described above, but with omission of the primary antibody, served as controls. To determine the percentage of cells stained with a particular mAb probe, the cells were scored using a bright-field microscope. Random fields were counted using an ocular grid and a 25 x objective. At least 400 cells in each category were counted. Nuclei were briefly counterstained with haematoxylin. Thymidine incorporation. [3H]-thymidine ([3HJ-TdR, Dupont, NEN-Products, Mass, USA) with a specific activity of 20 Ci/mmol was diluted in saline 1 :40. Cell cultures were rinsed once in medium without thymidine (GI-MEM), incubated for 1 h with 5 ml GIMEM+ CT-EGF and 0.5 ml [3H]-TdR, and then the immunocytochemical procedure was performed as described above. All cultures for autoradiography were coated with Ilford K.2 emulsion (Ilford, UK) and exposed for 1 week. As control served a new cell line from this laboratory. This cell line, HMT-3909S8, which is known to express vimentin [50], was incubated with ['HITdR followed by peroxidase or [ 'HI-TdR followed by alkaline phosphatase to ensure that the immunocytochemistry procedure did not influence the [3H]-TdR incorporation. After being developed the cultures were counterstained with haematoxylin, and the number of [3H]-TdR-labeled cells was estimated using bright-field microscopy. Random fields were counted using an ocular grid and a 25 x objective. Gel electrophoresis and immunohlotting. Preparations of intermediate filaments were made according to the method of Achtstatter [11. Then SDS polyacrylamide gel electrophoresis was performed
148 using 10%-17% polyacrylamide gradient gel [31]. Next, the proteins were transferred to Immobilon Transfer Membrane (Millipore), essentially by the method of Towbin [63]. Vimentin bands were identified immunochemically. Briefly, lanes cut from the membrane were blocked with PBS -t 2.0% Tween 80 (Sigma) for 5 min, incubated for 19-20 h with the specific monoclonal antibodies ( V I M S 1 : 50, MA907 1 : 100, M725 1 :400, Labsystems 1 : 10, F3006 1 : 50, all diluted in incubation buffer (50 m M Tris. 150 m M NaCl+0.05% Tween SO), rinsed five times for 5 min each with washing buffer (50 mM Tris, 150 m M NaCI), incubated for 2 h with peroxidase-conjugated rabbit anti-mouse Ig (P260 1 : 850 in incubation buffer, Dakopatts), rinsed five times for 5 rnin each in washing buffer, incubated for 5 niin in 50 m M Na-acetate (pH 5.2) and then stained. Substrate and colour reagents were 20 mg 3 amino-9 ethylcarbazole (Sigma), 2.5 ml N,N-dimethylformamide (Merck, FRG) in 50 ml 50 m M Na-acetate plus 25 p1 30% H A . Lanes with molecular-weight markers (Bio-Rad Laboratories, Richmond, Calif, USA) were rinsed four times for 5 min each with PBS+0.3% Tween 20 (Merck, FRG) and stained with India Ink (1 : 1000 in PBS-TW, Staedtler marsmatic 745, FRG) for 2 h.
Results Immunoperoxidase cytochemistry on methanol-fixed cryostat sections of human breast tissue revealed a rather consistent pattern of reactivity using five different monoclonal antibodies to vimentin (Table 1). Thus, immunoreactivity was seen in almost all myoepithelial cells in both interlobular ducts and terminal duct lobular units (TDLU); (Fig. 1). However, the intensity of reactivity varied somewhat, the weakest being seen in sections incubated with the IgM class of monoclonal antibodies (Table 1). Whereas no immunoreaction to vimentin was observed in luminal epithelial cells, either in interlobular ducts or in TDLU, these cells showed 100% immunoreactivity when incubated with antibody to keratin K18 (Fig. 1). All antibodies to vimentin showed consistent reactivity with all stromal cell types, while no reactivity was seen with antibodies to keratin K18 (Fig. 1). Similar patterns of reactivity were seen no matter whether the tissue was unfixed, fixed in methanol, in methanol-acetone, or in acetone alone (not shown), but fixation in formaldehyde or acetic acid-containing fixatives strongly reduced the immunoreactivity. It is therefore concluded that vimentin is expressed by myoepithelial cells in situ
as contrasted by absence of reactivity in luminal epithelial cells. Next, we analysed the kinetics of vimentin expression in mammary epithelial cells from freshly explanted tissue in primary culture. All of the monoclonal antibodies used for the in vitro analyses showed only one band in the 50-60 kDa range after gel electrophoresis and immunoblotting, indicating monospecificity for vimentin (Fig. 2). The collagenase digest was plated as epithelial cell clumps (organoids) in serum-free medium (Fig. 3). The first cell type known to appear in monolayer from the organoids is the myoepithelial cell [47] (Fig. 3). As expected from the in situ analysis, all myoepithelial cells (100%) showed immunoreactivity for vimentin from their first appearance in monolayer, i.e. on day 3 (Fig. 3). The myoepithelial cells were easily distinguished from stromal cells, since the myoepithelial cells showed strong immunoreactivity for keratins K5 and K14. Also, these cells were immunoreactive with antibodies to a-smooth muscle iso-actin (not shown) [47,48]. In theory, the first vimentin-positive cell type to appear in culture could be luminal epithelial cells, in which vimentin was induced due to the transition into a twodimensional monolayer. To exclude this possibility, we incubated sections of organoids, in which the cells showed preserved three-dimensional structural organization. These experiments showed that myoepithelial cells, also inside the organoids, were immunoreactive for vimentin, whereas the luminal epithelial cells remained vimentin-negative (not shown). This indicates that the myoepithelial cells continued to express vimentin from the situation in situ through explantation and monolayer cultivation. In addition, there were no signs of downregulation of vimentin expression in myoepithelial cells, even after prolonged cultivation or passaging of the cells. After 3 days of cultivation, the central zone of small organoids was essentially free of migrating myoepithelial cells, leaving behind an almost pure population of luminal epithelial cells (Fig. 3). We now analysed the kinetics of vimentin expression in this population of luminal epithelial cells. Luminal epithelial cells were identified using three different antibodies to keratin K18, and by additional antibodies to keratin K8 and K19. No reactivity was observed in the myoepithelial cells (not shown). To
Table 1. Antibodies: In situ myoepithelial cells (MEP) showed immunoreactivity with all the vimentin antibodies used, while luminal epithelial cells (LEP) were vimentin-negative, except for a weak nonfilamentous reaction obtained mAb M725. In vitro all antibodies gave a positive reaction in MEP, and in addition a fraction of LEP were positive for vimentin as well; cc, anti Anti-
Code
Clone
Company
Specics
Ig-subtypc
Dilution
In situ
In vitro -~
~~
Vimentin
M725 MA907 V5255 VIMS
-
V9 -
VIM 13.2 -
Weak nonfilamentous reaction Weak reaction ' Not used in vitro
Dakopatts, DK Labsystems, SF Enzo, USA Sigma, USA Medac, F R G
Mouse cc-swine Mouse cc-human Mouse cc-human Mouse cc-human Mousc w o w
IgG, IgG, IgM IgM k G1
1:400 1:lO 1 :200 1:lOO 1 :50
LEP
MEP
LEP
(+)" -
+
(+)"
~
+ (+y + ?' + + +
(+)h
MEP
+
+ + ?' +
149
Fig. 1a-d. Two consecutive cryostat sections (a and c ; details shown in b and d) of human breast tissue fixed in methanol. Sections were incubated immunocytochemically under conditions identical to those used for cultured cells. When incubated with antibody to vimentin (a, c). immunoreactivity was obtained in all rnyoepithelial cells ( M E P ) and in stromal cells ( S o ,while luniiiial epithelial cells (LEP) were negative. On the other hand luminal a
b
c
d
e
f
k
91 66-
4331-
2 2-
1li-
Fig. 2. Immunoblots showing that the various antibodies used are specific for vimentin. Lane a, molecular-weight markers; lane b, VIMS 1:50; lane c, MA907 3:lOO; lane d, A4725 1:400; lane e, LabSystems 1 :SO; lunef, control
epithelial cells showed immunoreactivity with antibodies to keratin K18 (b, d). As controls served sections incubated as described in Methods, but with omision of the primary antibody. In thcse sections no staining occurred. a, c Vimentin monoclonal antibody (mAb) VIMS. b, d MAb to kcratin K18, F3006. Bur, 100 ym (a, b), 50 pm (c, d)
prevent any confusion between myoepithelial cells and luminal epithelial cells during this analysis, we performed a double incubation of the cultures to demonstrate both keratin K18 - specific for luminal epithelial cells - and vimentin. Also, we measured the growth rate by [3H]-thymidine ([3H]-TdR) incorporation. To ensure that the immunocytochemical incubation did not perturb the subsequent autoradiography procedure, we incubated cultures from a cell line known to express vimentin with [3H]-TdR followed by peroxidase or [3H]TdR followed by alkaline phosphatase, and compared it with a control incubated with [3H]-TdR alone. No significant difference was found. The experiment revealed that the number of luminal epithelial cells that were positive for vimentin increased significantly within 7 days of cultivation (Fig. 4a). Thereafter a slight decrease in the frequency was observed corresponding to the time when cells reached confluency (Fig. 4a). A time-sequence analysis of the [3H]TdR incorporation showed that the luminal epithelial cell population exhibited a growth rate pattern very similar to the pattern of vimentin immunoreactivity (Fig. 4a). However, using primary cultures, the standard
150
Fig. 3a, b. Light micrographs of primary cultures stained with immunoperoxidase to demonstrate vimentin (mAb V/Ms). a After 3 days of cultivation the remnants of organoids ( O R ) are surrounded by a central zone of small, compact luminal epithelial cells (LEP) essentially free of migrating myoepithelial cells. Immun-
ocytochemistry with mAb VZMS revealed that all the myoepithelial cells (MEP) contained vimentin, while the luminal epithelial cells were all negative. b Higher magnification of myoepithelial cells stained with mAb VIMS. Distinct filament bundles are present in all the myoepithelial cells. Bar, 50 pm (a), 25 pm (b)
deviations on our vimentin immunoreactivity data were too high, making a firm interpretation of the results difficult (Fig. 4a). Therefore, another experiment was performed, in which the cells were passaged on day 5, allowing a clear separation of the luminal epithelial and myoepithelial cells in monolayer. This experiment essentially confirmed our initial observations (Fig. 4b). On the basis of these experiments, we conclude that vimentin is actually induced in luminal epithelial cells in monolayer culture. This induction appears to correlate with the growth rate of the cells. Next, we questioned to what extent the transition into monolayer and the growth rate, respectively, contributed to the induction of vimentin in luminal epithelial cells. One approach to answering this question was to arrest the cells by mitogen depletion of the culture medium, thereby simulating the growth situation in vivo. Exponentially growing cells in primary culture were passaged into a culture medium containing only CT and EGF, and growth was measured using [3H]-TdR incorporation along with double-labeling immunocytochemistry for vimentin and keratin K18 (Fig. 5). As seen in Fig. 6a, vimentin immunoreactivity in luminal epithelial cells - defined by the presence of keratin K1S - decreased gradually from approximately 37% on day 2 to 3% on day 16 in medium containing CT-EGF only. No further
reduction was obtained by extending the culture period for up to 4 weeks. By comparing the kinetics of vimentin immunoreactivity in luminal epithelial cells during the first 16 days of cultivation in secondary cultures with the frequency of cells incorporating [3H]-TdR, it is evident that growth declines with the same kinetics as the vimentin expression (Fig. 6a). This association between growth and vimentin expression among lurninal epithelial cells is not universal among cultured mammary epithelial cells, since the myoepithelial cells under identical growth conditions remained vimentin-positive (Fig. 6 b). Even though vimentin was induced in proliferating luminal epithelial cells, the immunoreactivity mostly did not resemble the typical filamentous patterns seen in myoepithelial cells. In the majority of vimentin-positive lumjnal epithelial cells, the reaction appeared as punctate or fragmented filaments (Fig. 7). Based on this experiment, we conclude that it is possible to revert the vimentin expression induced in cultured luminal epithelial cells to almost zero. Therefore, the transition of cells into monolayer cultivation is not strictly associated with the acquisition of vimentin filaments. Rather, this phenomenon seems to be tightly coupled to the rate of growth in these cells. We then questioned whether luminal epithelial cells were able to enter the cell cycle and proliferate in the
151 30 15
I
2
a 70
6 TIME (days)
0
I
0
b
4
0 10
I 30
5
A
10
15
pression may have been present earlier at the transcriptional level. Finally we measured the frequency of cells that had incorporated [3H]-TdR cells among exponentially growing vimentin-positive and vimentin-negative luminal epithelial cells. This experiment was performed on secondary cultures, 2 days after they had been growth-arrested, and it revealed that 56.5% of the vimentin-positive luminal epithelial cells were labeled with [3H]-TdR, while 11.0% of the vimentin-negative cells were [3H]-TdR labeled. Thus, even though the vimentin-positive luminal epithelial cells were growing faster, there was a distinct fraction of vimentin-negative luminal cells growing as well.
20
TlME(days)
Fig. 4a, b. Relationship between growth rate (- - - -0- - - -) and vimentin-positive luminal epithelial cells (---D----) grown in DME-F12 containing cholera toxin (CT), epidermal growth factor (EGF), transferrin (Tf), hydrocortisone (H), and insulin (I). Each point represents the mean frequency of at least 4 times 100 cells. a In primary cultures the number of luminal epithelial cells expressing both keratin K18 and vimentin showed a significant increase within 7 days of cultivation. Thereafter, a minor decrease of double-labeled luminal epithelial cells was observed corresponding to the time when cultures reached confluency. From [3H]-TdR incorporation it was clear that the growth rate pattern correlate very well with the pattern of double-labeled luminal epithelial cells. However, standard deviations were very high using primary cultures. Therefore secondary cultures were used in b. Primary cultures were grown in DME-F12 containing CT, EGF, Tf, H and I and passaged at day 5 (arrowhead). As seen in a the frequency of double-labeled cells increased within 7 days after passaging, followed by a decrease, when cultures reached confluency. The growth rate curve ([3H]-TdR incorporation) showed a similar pattern
absence of vimentin expression. We therefore stimulated luminal epithelial cells that had been growth-arrested in DME-F12 containing only CT-EGF. As seen in Fig. 8, we found a remarkable increase in the frequency of [3H]-TdR-labeled cells within the first 6 days. During the same period, almost no increase in the frequency of vimentin-immunoreactive cells was noticed (Fig. 8), indicating that vimentin expression is not essential for the cells to enter the cell cycle. However, vimentin ex-
Discussion In the present study we have shown that in human breast tissue vimentin is expressed in myoepithelial cells and stromal cells, while it is absent in luminal epithelial cells. Primary cultivation of breast tissue did not alter the expression of vimentin in myoepithelial cells and stromal cells, whereas luminal epithelial cells acquired this phenotypic trait. The frequency of cultured luminal epithelial cells expressing vimentin in culture correlated with the growth rate. The observation that myoepithelial cells in interlobular ducts as well as in TDLU express vimentin is in accordance with results obtained on rat mammary gland [15, 271, and on the human mammary gland [47], but differs from most other studies on breast and salivary gland in which vimentin was found in stromal cells only [3, 11, 13, 20, 22, 28, 691. Also, it has been reported that vimentin is present primarily in myoepithelial cells of TDLU [23]. The variations in the reports on vimentin expression in mammary epithelial cells in situ may be due to the fact that tissue from different mammals was examined, different antibodies were used, and lactating glands were used in some of the studies [21, 27, 421. Other reasons for the discrepancies include the use of either frozen material or fixed and embedded material, the effects of different fixatives on the immunoreactivity of intermediate filaments, and the choice of immunocytochemical method [64]. In our study it is not surprising to find vimentin immunoreactivity in all of the cultured myoepithelial cells, since we find it in myoepithelial cells in situ as well. From studies using the antibody MA907, it has been reported that vimentin is induced upon cultivation in myoepithelial cells not expressing this phenotype in situ [13]. Using the same antibody we have found vimentin in myoepithelial cells in situ, indicating that the immunoreactivity seen in cultured cells does not reflect actual induction. We do not find it likely that the coexpression of keratin and vimentin in myoepithelial cells reflects an aberrant induction of keratins in stromal cells [30, 471. This is mainly based on our own analysis of characterized stromal cells in such cultures [47]. However the reactivity seen with mAb MA907 was weak in sections and in cultured cells. A few reports based on
152
Fig. Sa, b. Double incubation of cultured cells by the “Unlabeled Antibody Method” showing vimentin (mAb VfMS) with the brown DAB-H,O, reaction product in the first sequence and cytokeratin K18 with Fast Red Violet-alkaline phosphatase in the second sequence. The b1ur.k nuclei represent [3H]-TdR incorpora-
tion of proliferating cells. We found that myoepithelial cells ( M E P ) were positive with anti-vimentin, and luminal epithelial cells ( L E P ) were positive with anti-cytokeratin K18, but in addition there was a fraction of the luminal epithelial cells that showed immunoreactivity with both antibodies (LEP-D).Bar, 25 pm
153
a
-
TIME (days)
100
-
1
I,. ,
J
.
- 30
. F 2
