Short-term primary culture of epithelial cells derived from human breast tumours.

As experimental models for breast cancer, most studies rely on established human breast cancer cell lines. However, many of these lines were established over 20 years ago, many from pleural effusions rather than the primary tumour, so the validity of using them as representative models is questionable. This paper describes our experiences, over a 3-year period, in establishing short-term epithelial-cell-enriched preparations from primary breast tumours based on differential centrifugation followed by culture in selective media. Epithelial cells were successfully cultured from 55% of samples, but culture success did not appear to be correlated with tumour histology, stage, grade or node status. Epithelial cell-enriched cultures were immunopositive for broad-spectrum cytokeratin and epithelial membrane antigen (EMA). Positivity for keratin 19 confirmed that the cultures contained tumour-derived cells, which additionally showed significantly higher activity of the reductive pathway of the steroid-converting enzyme 17beta-hydroxysteroid dehydrogenase type I. That the cultures contained tumour and not normal epithelial cells was further substantiated by the complete absence of the calmodulin-like gene NB-1 in tumour-derived cultures; this is only associated with normal breast epithelia. Eighty-five per cent of cultures established from oestrogen receptor (ER)-positive tumours expressed ER in vitro; this was functional in 66% of cultures, although ER-positive phenotype was gradually lost over time. In conclusion, epithelial cells can be isolated and maintained as short-term cultures from primary breast tumours irrespective of histopathological or clinical details, providing a model system with a greater biological and clinical relevance than breast cancer cell lines.

The majority of scientists and clinicians involved with breast cancer research use established breast cancer cell lines for in vitro studies, and a significant amount of useful information has been accumulated from these over the years. However, little consideration is given to the fact that many of these cell lines were originally described over 20 years ago and, with genetic drift, may bear very little resemblance to the primary tumour from which they were derived. This point is reinforced by the observation that conflicting results have often been reported from different laboratories in identical studies using the 'same' target cell (Osborne et al, 1987). Further, the most used in vitro model for breast cancer, the MCF-7 cell line, first described in 1973 (Soule et al, 1973), was actually derived from an aggressive metastatic pleural effusion, as is the case with most of the breast cancer cell lines currently in use today (Cailleau et al, 1974). Despite this, a database search of the literature over the past three decades using Medline (Win SPIRS version 2.0 on Silverplatter) and Bath Information Database Service (BIDS), revealed that the MCF-7 cell line still remains the model of choice as a research tool, with 97% of the published literature using this as a model for breast cancer and less than 3% of published studies using primary cultures derived from breast tumours.
The reasons for the continued use of cell lines as in vitro models are obvious, the prime one being their ease of use. Cell lines are well characterized homogeneous populations that (usually) give consistent and reproducible results from experiment to experiment.
Correspondence to: V Speirs They are also easily replaced if lost through contamination. Many of these lines express functional oestrogen and/or progesterone receptors and some have been characterized at the genetic level. Nevertheless, it has been a long-standing target of many laboratories to establish and maintain primary cultures, particularly of luminal epithelial cells, from which most breast cancers arise (Taylor-Papadimitrou et al, 1989), as a more representative in vitro model. However, these experiments have met with limited success, mainly because breast cancer epithelial cells can be difficult to establish, with slow doubling times. Furthermore, such cultures are often poorly characterized, with rapid overgrowth by fibroblasts frequently representing a considerable technical challenge.
The aim of this study was to establish a reproducible technique suitable for the short-term culture of human breast epithelial cells derived from breast tumours to provide sufficient quantities of cells for molecular and cell biology studies. Using a modification of a method based on differential centrifugation followed by culture in selective media (Emerson and Wilkinson, 1990;Speirs et al, 1996a), this report describes our experiences in establishing such cultures over a 3-year period and discusses their benefits over cell lines. The cultures have been characterized by a number of criteria including immunocytochemistry and flow cytometry with a range of primary antibodies, RT-PCR to detect the NB-] gene, a calmodulin-like protein of unknown function that is expressed in cultured normal, but not malignant, mammary epithelial cells (Yaswen et al, 1990;Stampfer and Yaswen, 1993) and the presence of increased reductive activity of the steroid-converting enzyme 17p-hydroxysteroid dehydrogenase type I (17-HSD), which is significantly up-regulated in breast tumours (Vermeulen et al, 1986;Peltoketo et al, 1996).  Ti   25  T2  73  T3  5  T4  1  Not known  5 aThis group comprised one apocrine, one leiomyosarcoma, two medullary, two mixed lobular-ductal carcinoma and one tumour of ductal-mucoid origin.

Tissue samples
Breast tumours were obtained from 109 patients undergoing surgery for the removal of a clinically confirmed breast lesion who presented sequentially. The mean age was 63 years (range 32-90). Clinicopathological details of the tumours are presented in Table 1.
Cell isolation and culture Upon receipt, tissue was washed extensively in phosphate-buffered saline (PBS) supplemented with 200 U of penicillin, 200 gg ml-1 streptomycin and 5 ,ug ml-' fungizone (all Life Technologies, Paisley, UK), then minced finely and disaggregated for 18-20 in 0.1% collagenase type III (Life Technologies) as previously described (Speirs et al, 1996b). Digested tissue was removed from the incubator and shaken vigorously by hand to disaggregate any remaining large clumps. Three cell populations were then isolated using differential centrifugation (Emerson and Wilkinson, 1990). This is described in greater detail in the results section and illustrated schematically in Figure 1. For the first 24 h, cells from the organoid and epithelial fractions were plated in 75% organoid medium (OM) and 25% Dulbecco's modified Eagle medium (DMEM) containing 10% heat-inactivated fetal bovine serum (HIFBS) to promote cell attachment. OM consisted of DMEM supplemented with 100 U ml-' penicillin, 100 ,ug ml-' streptomycin, 2 mm glutamine, 10 mm Hepes, 0.075% bovine serum albumin (BSA), 10 ng ml-' cholera toxin (ICN Biomedicals, Oxon, UK), 0.5 gg ml-' hydrocortisone, 5 gg ml-' insulin and 5 ng ml-1 epidermal growth factor (EGF) (all Sigma, Poole  (Speirs et al, 1995). This was amplified with oligonucleotide primers designed to detect NB-I, a calmodulinrelated gene expressed by most normal but not by malignant mammary epithelial cells (Yaswen et al, 1990;Stampfer and Yaswen, 1993). Primer sequences were as follows: 5'-TGT TTG ACA AGG ATG GGG AC-3' water substituted for cDNA acted as a negative control. Product identity was confirmed by restriction mapping. An aliquot of 5 ,ul of PCR product was incubated with BanI restriction enzyme (New England Biolabs) for 2 h at 37°C, which yielded two discrete frag-, ments of 252 and 128 bp. Cleaved and uncleaved products were separated on a 2% agarose gel and visualized by ethidium bromide

Statistical analysis
The chi-squared test was used to determine statistical significances for cell culture criteria and the unpaired Student's t-test was used for 17-HSD experiments.

RESULTS
Criteria used to ascertain culture success Epithelial cultures were considered to be successful if they displayed the following features: (1) attachment and growth of phenotypically epithelial-like cells to the culture vessel after 3-5 days in vitro in defined medium; (2) continued growth, under these conditions, for up to 6-8 weeks with a doubling time of approximately 3-6 days, with no evidence of fibroblastic infiltration; (3) immunopositivity for broad-spectrum cytokeratin, clonal reactivity for K19 and EMA, with no expression of vimentin or CALLA; (4) lack of expression of the NB-I gene by RT-PCR; (5) preferential reductive 17-HSD activity in cell monolayers. These characteristics are described in greater detail below.

Phenotypic identity
The differential centrifugation technique gave rise to cell populations from all three centrifugal fractions (Figure 1). The organoid fraction consisted predominantly of small fragments of partially digested tissue from which epithelial-like cells began to grow after 1-2 days in organoid medium. Epithelial cell-enriched cell pellets were obtained by recentrifuging the supernatant at higher speed, yielding mainly single cells, which, when cultured in organoid medium, displayed a characteristic cobblestone morphology indicative of epithelial cells. The epithelial supernatant was recentrifuged to yield stromal cells, which had a typical spindle-shaped morphology and formed parallel arrays upon reaching confluence, a feature typical of fibroblasts. Confirmation of the phenotypic identity of the three populations was validated by immunostaining with specific primary antibodies and has previously been published (Speirs et al, 1996a). Focal positivity for K19, which is associated with luminal epithelial cells, and EMA, an epithelial cell-surface marker, was observed in some organoid and the majority of epithelial fractions, but absent from the stromal fraction ( Figure 2).

FACS analysis
To complement the immunostaining, FACS analysis was carried out using specific cell-surface antigens. As predicted, approximately 50% of epithelial cells expressed EMA with no expression of the myoepithelial marker CALLA (Figure 3).

Cell growth
As illustrated in Figure 4, growth rates of epithelial cells derived from different tumours varied, with population doubling times between 3 and 6 days. After 21 days, cells began to reach plateau. Cell viability was assessed by trypan blue exclusion and ranged from 99% when the cultures were first established to 85% after 21 days in vitro. Cultures described in this paper were not routinely continued beyond 21 days, although in a previous study we have shown similar cultures to be both phenotypically and genotypically stable in vitro for up to 6-8 weeks (Speirs et al, 1996a).

Genetic analysis
Breast cDNA was amplified with primers designed to detect fragments corresponding to the NB-I gene. In 13/17 breast epithelial cultures derived from reduction mammoplasties, RT-PCR revealed a single band of 380 bp that corresponded to the predicted size of the NB-] cDNA transcript, the identity of which was confirmed by restriction mapping (Figure 5). This transcript was not detected in 23 tumour-derived primary epithelial cultures or in tumour-derived    Activity of 17p-hydroxysteroid dehydrogenase type I in normal and malignant breast epithelial cultures Reductive (El -*E2) and oxidative (E2-4E1) activities for epithelial culture derived from normal breast and breast tumours are illustrated in Figure 6. In tumour-derived cultures, the reductive pathway was preferred; this was significantly greater than reductive activity associated with epithelial cells derived from normal breast. For the oxidative pathway, the reverse was true with this pathway superior in cultures derived from normal rather than malignant breast.
British Journal of Cancer (1998) 78 (11) To ascertain if tumours that yielded successful cultures were asso-:ciated with a particular clinical feature, culture statistics were ...~~~~~~~~correlated with clinical details. Tumour histology did not appear to be an important factor in establishing viable epithelial cultures, with cultures equally likely to be established from tumours of } X _ ... ductal or lobular origin (Table 2). Similarly, tumour grade had no bearing on culture success (Table 3). Correlating this with patient age, epithelial growth appeared to be more likely from tumours taken from patients in the fourth to sixth decades (Table 4).
Tumour stage or lymph node status were both unimportant in terms of successful epithelial cell growth (data not shown). A notable feature was that the number of successful cultures we established increased over the 3 years since the initiation of the study, giving a mean success rate of 55%, suggesting that culture success is due to technical expertise rather than a specific clinical feature.

C
Correlating tissue ER with cell culture ER status ER data was available for 36 breast tumours. Of these, 21 expressed ER by immunohistochemistry. This is summarized in  Figure 7. Addition of exogenous 17p-oestradiol to these cultures resulted in a growth stimulation of up to twofold by approximately two-thirds of all cultures, confirming the functionality of the receptor (Figure 8). However, ER expression was lost after two or three passages, indicating that the receptor is not phenotypically stable in these cultures (data not shown).

DISCUSSION
This study has described the successful short-term culture and phenotypic characterization of breast cancer epithelial cells derived ire 7 Immunostaining for ER (A) 7-day primary breast epithelial culture, from primary human tumours. Routine primary culture of epithelial MCF-7 positive control, (C) MCF-7 negative control. In A and B strong ear staining is evident (arrows). No staining is evident in C. Scale bars cells derived from such tumours has been a goal of many laboratories ,ml (A) and 20 gm (B and C)  variety of cell and molecular biology studies (Speirs et al, 1996a;Green et al, 1997). These cultures retain many of the characteristics associated with breast tumours in vivo, particularly the presence of functional ER and a high level of reductive 17-HSD, an enzyme involved in steroid biosynthesis whose activity is up-regulated in breast tumours (Vermeulen et al, 1986;Peltoketo et al, 1996). Epithelial cell-enriched preparations were isolated using a differential centrifugation method. This technique gave rise to three individual fractions, termed organoid, epithelial and stromal, all of which have previously been characterized (Speirs et al, 1996a). The organoid fraction was heterogeneous, consisting of small pieces of partially digested tissue and single cells. For this reason and as it had been reported that tumour organoids give rise to rapidly proliferating epithelial cells, the majority of which are genetically normal (Wolman et al, 1985), we chose not to use this fraction as a source of epithelial cells. The epithelial fraction was characterized using two different immunological methods. By immunohistochemistry, the epithelial nature of the cultures was confirmed; this revealed positivity for broad-spectrum cytokeratin and focal positivity for EMA. Further characterization by FACS analysis of specific surface antigens showed EMA positivity on about 50% of epithelial cells. As predicted, there was no evidence for CALLA, a myoepithelial cell marker (O'Hare et al, 1991), in these cells.
ER was detected in 85% of cultured breast epithelial cells isolated from ER-positive tumours. Its functionality was confirmed by a growth response to exogenous 17p-oestradiol in approximately two-thirds of these cultures. However, the receptor was not phenotypically stable with a gradual loss of expression over time. Loss of ER in vitro has previously been reported in breast (Pink et al, 1996) and osteosarcoma (Nasir and Speirs, 1997) cell lines and has been attributed to oestrogen deprivation in culture. Although we do not routinely add E2 to the organoid medium used to maintain our epithelial cultures, in a small number of cases we supplemented the medium added to the longer term cultures (i.e. those that had undergone at least two passages) with E2 but were still unable to detect ER (unpublished observations). This suggests that loss of ER by these cultures is an irreversible process. Using normal breast as a model, it has recently been suggested that ER-positive cells do not generally proliferate, but rather act as sensor cells for circulating E2, whereas the proliferating cells act as effecter cells in terms of oestrogen-stimulated mitogenesis (Clarke et al, 1997). As such, proliferation of ERnegative cells may be controlled by paracrine factors released from ER-positive cells under the influence of oestradiol; this may be disrupted in breast tumours (Clarke et al, 1997).
Basic immunohistochemistry and FACS analysis confirmed the phenotypic nature of the epithelial fraction; however, one of the major stumbling blocks in this type of work is knowing whether or not the isolated cells are of neoplastic origin, because breast tumours contain cells of normal as well as malignant origin (Petersen et al, 1990). Therefore, epithelial cultures were additionally stained with K19, which is universally expressed in breast tumours (Bartek et al, 1985). Although this antigen is expressed in tissue sections by luminal epithelial cells within the terminal ductal lobular units of both normal and malignant origin, in vitro it is generally only associated with tumour epithelial cells (Taylor-Papadimitrou et al, 1989) and as such has been used as a marker to detect lymph node micrometastases in breast cancer patients (Schoenfield et al, 1994;. In accordance with other studies, isolated epithelial cells were not universally positive for K19; instead, K19 positivity was restricted to single cells or islands of cells adjacent to colonies that were K19 negative (Shearer et al, 1992;Bergstraesser and Weitzman, 1993;Ethier et al, 1993). It has been reported that fetal bovine serum (FBS) is required to maintain a K 19-positive phenotype (Ethier et al, 1993). In the present study, cells were exposed to FBS (final concentration 2.5%) for the first 24 h before transferring into organoid medium, which contained BSA rather than FBS (we substituted BSA for FBS to inhibit overgrowth with fibroblasts). This appeared to be sufficient to maintain the K19 phenotype, at least for the duration of our experiments, although it would be interesting to establish the long-term effects of BSA on K19 expression in vitro.
To further confirm the phenotype of our epithelial cultures we carried out RT-PCR studies on the NB-] gene. This encodes a calmodulin-like protein of unknown function (Yaswen et al, 1990). In normal breast epithelial cultures, the gene was detectable in 13/17 instances, but not found in either tumour-derived epithelial cultures or in tumour cell lines derived from breast or bone. This concurs with other studies that have shown its expression in cultured normal, but not malignant, mammary epithelial cells (Yaswen et al, 1990;Stampfer and Yaswen, 1993). It has also been suggested that in normal breast, the gene is down-regulated in vitro (Yaswen et al, 1990), which may explain the lack of expression of NB-1 in four normal epithelial cultures. The sensitivity of the PCR, in theory capable of amplifying message from a single cell, unequivocally demonstrates that our cultures contained tumour and not normal epithelial cells. If contaminating normal epithelial cells were present in our tumour cell preparations, these would have certainly been detected by RT-PCR for the NB-I gene, particularly after 35 amplification cycles.
Our biochemical results provided further evidence that our cultures contained malignant cells because, in cell cultures derived from tumours, the reductive pathway of 17-HSD type I was significantly greater than the oxidative direction. This is in agreement with a number of previous studies that showed this to be the preferred direction in breast tumours (Adams et al, 1988a;Reed et al, 1991;Poutanen et al, 1995;Castgnetta et al, 1996). In contrast, in cultures derived from normal breast, the oxidative pathway was preferred. Differences in enzyme direction between cell cultures from normal and tumorous breast may reflect the ratio of endogenous co-factor activity associated with each tissue type that will drive the reaction in a particular direction. Although co-factors are routinely used when measuring 17-HSD activity in soluble tissue fractions, we deliberately did not add them to our cell cultures as addition of these factors may have influenced the enzyme direction.
Recently, Dairkee et al (1997) have described a method based on a partial enzymatic degradation of tumour stroma that permits enrichment and expansion of breast epithelial cells in vitro. Although their technique differs from the one described herein, both methods give rise to similar proportions of epithelial cells with proliferative capacity (Dairkee, 66%, this study, 55%). However, with Dairkee's method, contamination of the epithelial fraction with fibroblasts was apparent, requiring differential trypinization to remove them. With our method, a combination of differential centrifugation followed by culture in medium designed to encourage epithelial proliferation, fibroblast overgrowth was not a problem. Although we have previously optimized our tumour dispersal technique to give superior cell yields without compromising viability (Speirs et al, 1996b), perhaps a combination of the partial enzymatic digestion (Dairkee et al, 1997), followed by the differential centrifugation method would prove a more appropriate British Journal of Cancer (1998) 78(11), 1421-1429 choice to maximize yields of breast cancer epithelial cells for future studies.
It is important to note that when this study was initiated, it was not our objective to establish a method to support the long-term culture of primary breast cancer epithelial cells or to establish a new cell line(s). Breast tumours, by their nature, contain heterogeneous mixtures of cells and extended culture would tend, by natural selection, to favour cells with a more robust phenotype. Phenotypic changes in response to culture conditions have been reported (Taylor-Papadimitrou et al, 1989) and it is quite likely that genotypic aberrations may be induced with long-term culture. Over the years, new cell lines have been established from only a very small percentage of primary tumours (1-2%; Petersen et al, 1990;Band et al, 1990;Meltzer et al, 1991). However, a more recent report has described the long-term culture of approximately 7% of cultures established from primary breast cancers, nearly all of which went on to form new cell lines (McCallum and Lowther, 1996). Interestingly, the majority of these cultures grew in suspension, in contrast to most other studies, and were more likely to be established from grade III tumours with a steroid receptor-negative phenotype (McCallum and Lowther, 1996). However, in our study, pathoclinical features including tumour histology, grade, stage or node status had no bearing on culture success. Indeed the most notable feature of this study was the increased proportion of viable cultures established over time, which we attribute to technical expertise rather than a particular clinical characteristic. There also appeared to be a trend for cultures to be established from tumours taken from individuals in the fourth to sixth decades. However, as the majority of our samples came from this age group (71/109) this is unlikely to be a significant observation, particularly as cultures were readily established from tumours removed from more elderly patients as well. Thus, age is probably not an important factor in determining culture success.
In accord with many other groups, we were unable to culture breast epithelial cells that were universally positive for the luminal epithelial marker K19. However, breast tumours in vivo do not consist entirely of tumour cells, but contain a heterogeneous mixture comprising normal/benign epithelial cells and stromal cells as well as cells of the immune system. It has been proposed that, in vivo, tumour epithelial cells may have an absolute dependence on paracrine signals from neighbouring cells (Lippmann et al, 1989;Osborne and Arteaga, 1990). This is reinforced by in vitro studies that showed paracrine factors secreted by breast tumour-derived fibroblasts or lymphocytes stimulated proliferation of breast cancer cell lines (Adams et al, 1988b;van Roozendaal et al, 1992) and primary cultures of breast cancer epithelial (Ogmundsdottir et al, 1993;Hofland et al, 1995;Emerman et al, 1996). Thus, paracrine-autocrine interactions are likely to be important factors in determining the biological behaviour of a tumour. Furthermore, a culture system such as ours may give a far more accurate representation of the tumour in situ than studies with isolated, homogeneous, cell types. Additionally, the ability to culture stromal cells independently from epithelial cells offers the possibility of recombination experiments, allowing the recreation, under controlled conditions, of cell interactions that clearly exist in vivo. These interactions should be given consideration in determining the response of a tumour to drug therapy, as they may influence the manner in which the tumour responds.
The value of a reproducible method for primary culture of breast cancer epithelial cells is significant. In the pharmaceutical industry the use of cell cultures derived from human primary material has enormous potential in the hunt for novel therapeutics. Further, there are considerable ethical pressures on scientists to seek alternatives to animals for drug discovery and the use of primary cell cultures could circumvent the need for these types of experiments. Another area where this technique may prove valuable is in predicting patient response to drug therapy. Establishing cell cultures from a tumour and determining the in vitro response to, for example anti-oestrogens or chemotherapeutic drugs may permit specific adjuvant therapies to be tailored to the needs of the individual patient, which, in the long term, may benefit prognosis.