A stable explant culture of HER2/neu invasive carcinoma supported by alpha-Smooth Muscle Actin expressing stromal cells to evaluate therapeutic agents

Two stock BALB-NeuT transgenic males were obtained through collaboration with Dr. Guido Forni. The BALB-NeuT strain originated from a transgenic CD1 random-bred breeder male mouse (no. 1330) carrying the mutated rat HER2/neu oncogene driven by the MMTV promoter [16]. The mutated gene encodes a single point mutation that replaces the valine residue at position 664 in the transmembrane (TM) domain of p185/neu with glutamic acid. This mutation promotes p185/neu homo- and heterodimerization and transforms the HER2/neu protooncogene into a dominant transforming oncogene. In these studies animals were used for tissue harvest only. Nonetheless, all animal research was conducted following an approved protocol filed with the Animal Investigation Committee at Wayne State University that oversees the Division of Laboratory Animal Resources (DLAR) at this institution in strict accordance with NIH guidelines.

"Iressa™" (Zeneca Pharmaceuticals Macclesfield, Cheshire) ZD1839, 4-(3-chloro-4-fluoroanilino)-7-methoxy-6-(3-morpholinopropoxy) quinazoline was suspended in DMSO at 10 mM and stored at -20°C until use. Dilutions were prepared in culture medium to make final concentrations ranging from 0.25–10 μM. The Bam1a cell line has been previously described and characterized [6].

The MAM-1 co-culture model was established from a BALB-NeuT transgenic mouse with an advanced mammary gland lobular carcinoma involved with hemorrhage. The tissue was removed asceptically and rinsed extensively in sterile PBS. The tissue was minced and dissociated briefly with collagenase. Cells were washed extensively with complete medium and explanted into tissue culture media. Cultures grew in an organization of tumor cell nests surrounded by myofibroblasts and maintained a consistent 1:1 ratio of tumor cells to stromal cells. These cultures were maintained in vitroin Dulbecco's modified Eagle's medium (DMEM) with high glucose and supplemented with 10% heat inactivated Serum Supreme (BioWhittaker, St. Louis, MO), 0.5 mM sodium pyruvate, 2 mM L-glutamate, 0.1 mM MEM nonessential amino acids, 100 units/ml penicillin and 100 μg/ml streptomycin and 10 μM Dexamethasone. Inclusion of dexamethasone, impairs the growth of contaminating fibroblasts and endothelial cells with limited proliferative and immortalization potential. Cells were cultured in 10% CO₂, media was changed every 3–4 days until confluence and split at a 1:5 ratio for routine passaging. -Co-cultures can be expected to maintain a 1:1 ratio during active growth. This ratio may be altered if cultures are kept beyond confluence when stromal cells continue to grow. At time of passage, the majority of stromal cells are collected separately by a brief trypsinization. The remaining co-culture is further trypsinized and monodispersed and collected separately. Both collections are counted by hemacytometer which readily allows for distinguishing between the tumor (small) and stromal (large) cells. If need be, the cell ratios can be readjusted to accommodate the desired 1:1 ratio. All in vitro data presented in this manuscript used MAM-1 between passages 5–20.

At time of necropsy, tumor tissues were fixed in 10% neutral buffered formalin and paraffin embedded using standard histochemical techniques. Blocks were sectioned 4-micron. Tissue sections were stained with Hematoxylin and Eosin for basic histological evaluation. For immunodetection of tissue antigens, histological grade primary antibodies were applied to the samples and incubated according to manufacturer's recommendation: HER2, PAD: Z4881, cat# 08-1204 2^nd Gen; PCNA, cat# 08-1110 (Zymed, South San Francisco, CA), PCNA Ab-1 clone (PC10) cat# MS-106-R7; Actin, Smooth Ab-1 (1A4) cat# MS-113-R7 (Neomarkers, Fremont, CA). Samples were washed and labeled using the SuperPicTure™ Polymer Detection Kit, cat#87-9263; Zymed), developed with DAB Substrate and counterstained with hematoxylin. Samples were evaluated using a Zeiss microscope and images were collected through a Sony 970 CCD camera interfaced with the MCID5+ imaging software package. Alternatively images were collected using a Nikon inverted microscope equipped with a SPOT digital cooled camera and imaging software. Stained sections were evaluated by a board certified pathologist to generate descriptions.

A variety of antibodies were used for flow cytometric analyses. Selection of antibody combinations was based on fixation, host species, avidity/affinity for specific epitopes and antigen density. Antibodies used in these studies included: antibody to the rat Her2/neu (Ab-9, clone B10, cat# MS-326); erbB-2 (Ab-15, clone 3B5, cat# MS-599, NeoMarkers, Fremont, CA). CD24 (M1/69) sc-19651-PE; CD29 (Integrin β1 (HMβ1-1) sc-19656-PE; p-c-Jun (Ser63)-PE (KM-1) sc-822-PE all from Santa Cruz Biotechnology, Santa Cruz, CA. Ready to use histological grade antibodies that were also used on formaldehyde fixed samples prepped for flow cytometry or immunofluorescence included: (HER2, PAD: Z4881, cat# 08-1204 2^nd Gen; PCNA, cat# 08-1110 Zymed, South San Francisco, CA). or PCNA Ab-1 clone (PC10) cat# MS-106-R7; Actin, Smooth Ab-1 (1A4) cat# MS-113-R7 or cat# RB-9010-R7, Neomarkers, Fremont, CA. Phospho-specific antibodies from Cell Signaling Technologies, (Beverly, MA) used in FACS and Immunofluorescence included: p-p44/42 MAP kinase (Thr202/Tyr204, #9101), p-MEK ½ (Ser217/221. #9121)

Routine cell surface staining of fresh cell cultures was as previously described [5]. For evaluation of intracellular antigens, MAM-1 were plated in 6 well tissue cultures plates to produce a confluent, organized co-culture (1:1 tumor to stroma cell ratio) within 2–3 days of seeding. Cells were harvested at specific time points after treatment with trypsin/EDTA, quenched with complete medium and collected by centrifugation. Pellets were resuspended in 1 mL of PBS and monodispersed by passing through a small bore pipet. Cells were fixed by titrating in 5% Formaldehyde to a final concentration of 1%, incubated at 37°C for 10 minutes, then chilled on ice for 2 minutes. Absolute, ice-cold (-20°C) methanol was added while vortexing to a final concentration of 90%. Samples were vortexed rigorously and incubated on ice for 30 minutes. Cells were collected by centrifugation and blocked for 30 minutes on ice with chilled FACscan buffer (PBS supplemented with 2% serum and 0.1% sodium azide). Cells were washed once with fresh FACscan buffer prior to the addition of primary antibodies or isotype controls and stained overnight at 4°C in accordance to the manufacture's guidelines. Cells were washed twice and stained with the appropriate -PE or -FITC conjugated secondary antibodies: goat anti-mouse IgG-PE, cat# 111-116-144; donkey anti-rabbit IgG-FITC, cat#711-096-152; donkey anti-mouse IgG-PE, cat# 715-116-150 (Jackson ImmunoResearch, West Grove, PA) at 1:50 to 1:200 on ice for 45 min. After staining, samples were washed twice with FACscan buffer, placed on ice and evaluated by flow cytometry using the dual-color laser option (FL1 v. FL2) in the FACsCaliber. At least 20,000 events were collected for every sample. Data were analyzed using WinMDI version 2.8 software. Importantly, all experiments were reproduced at least three times using MAM-1 of different passage numbers and overlapping experimental treatments and time points. The specificities of all stains were validated by using various combinations of monoclonal and polyclonal antibodies for each antigen and different secondary labels. All phospho-specific markers and PCNA stains were evaluated by direct reciprocal dual-staining for the test antigen with in combination with antibodies to ErbB2 and α-SMA to verify that all ErbB-2 negative responses were identical to α-SMA positive responses and vice versa.

MAM-1 suspensions were plated on glass cover slips in 6 well plates in complete media and grown to 90–95% of confluence and treated as described in the text. Following treatment, cells were fixed and permeabilized by immersion in ice-cold 100% methanol and incubated at -20°C for 10–20 min. Methanol was aspirated and cover slips were air-dried, washed and blocked prior to administration of primary antibodies. Primary antibodies were diluted and incubated on cover slips according to the recommendations for each specific antibody. Cover slips were washed 4 times before adding the secondary FITC- or TRITC-conjugated goat anti-mouse or rabbit IgG (cat# 115-096-071, 111-096-046, 111-116-144 or 115-116-071; Jackson ImmunoResearch, West Grove, PA) antibodies diluted 1:50-1:200 in FACscan buffer. Samples were incubated with secondary for 45 min. at room temperature in the dark. Stained cover slips were washed and mounted in Fluorescent mounting media (cat# HT08, Oncogene Science) containing DAPI for nuclear counterstaining. Slides were visualized at 100× under oil with a Zeiss microscope equipped with a Sony 970 digital cooled camera. Fluorescence photomicrographs were imaged using MCID5+ software. Live cell images and phase contrast of cells fixed in 6-well culture dishes were evaluated with a Nikon inverted microscope and imaged using the attached SPOT digital cooled camera and imaging software.

Total RNA was isolated from triplicate wells of Bam1a cells [6] or MAM-1 co-cultures that were treated for 24 h with complete media containing 0 or 1 μM Iressa. Total RNA was isolated using RNeasy kit (Qiagen, Carlsbad, CA) and sent to SuperArray Bioscience Corporation (Frederick, MD) and processed utilizing the GEArray Hybridization and Analysis Service and Agilent Mouse Genome CGH Microarray 44A that has 43,000+ annotated gene sequences with well-characterized genes represented by 1+ probes and cancer-relevant genes by 2+ probes. In some comparisons Universal mouse RNA provided by Agilent was also used in the analyses for normalization. Three independent RNA preparations per condition were analyzed on three separate chips and used to generate the analysis report. To generate lists of highly expressed genes, RNA from MAM-1 was labeled in red color (cy5), universal RNA was labeled in green (cy3). 1) In **treated vs **control comparison, the treated is labeled in red color (cy5), control is labeled in green color (cy3). Fold change (**treated/**control) = Normalized ratio value (cy5/cy3); fold change cutoff >2 fold (up regulate) or < 1/2 (down regulate); confidence cutoff >0.95.

The tumor was obtained from a BALB-NeuT transgenic mouse bearing a lesion that involved a hemorrhage of the #8 thoracic mammary gland. The pathology report describes the tumor as forming a discrete nodule, most of which is contained within a thin rim of collagen (pseudocapsule) with rare isolated acini penetrating it (Fig. 1A). The tumor is made of epithelial cells, arranged in ill-defined nests that are separated by fine fibrous septa. Discrete acini, delimiting central lumens, and filled with lightly eosinophilic secretions, confer to these epithelial nests a distinct overall, sieve-like (cribriform) configuration. Necrotic material and focally hemorrhage, frequently accumulates within the tumor, in a punctate fashion (comedo-type necrosis). Tumor cells show homogenous cytological features and only mild pleomorphism (Fig. 1B). They are polygonal, with a relatively low N/C ratio, eosinophilic cytoplasm, round central nuclei with finely dispersed chromatin and infrequent macronucleoli and undergo frequent mitoses. This description is consistent with the morphology of the lobular carcinomas that arise in BALB-NeuT transgenic mice [16].

The sample of this lesion that was explanted and processed for growth in vitro, demonstrates a significant stromal cell component (Fig. 1). By immunohistochemistry we observed that HER2/neu staining was strong and present in the majority of the tumor cells (Fig. 1C), in a cytoplasmic and membranous distribution and that α-SMA staining was strong and exclusively expressed in the cytoplasm of the stromal cells that were involved in the fibrous sheaths surrounding the tumor cell nests (Fig. 1D).

In mice transgenic for the activated rat HER2/neu under the MMTV promoter, expression of the oncogene in the mammary gland epithelium gives rise to an alveolar type of lobular carcinoma that requires an angiogenic switch for tumor onset and progression to invasive cancer [16]. Coordinated epithelial-stromal interactions that are required for mammary morphogenesis and development are also critical for tumor progression in this model. In the case of human breast cancers, stromal alterations are also integral to the evolution and progression of breast cancer [17, 18].

Preservation of the breast cancer microenvironment is critical for evaluating therapeutic agents especially when designing modalities that target invasive disease. Thus, we sought to establish a homotypic system representing invasive breast cancer by explanting the mammary tumor and associated stromal cells to develop a co-culture model. Upon explantation from the tumor, primary cultures that grew in vitro produced a co-culture that organized as multi-cell layered nests of tumor cells surrounded by concentric rings of stromal cells. These arrangements bear a striking similarity to the morphology of the primary tumor lesion (Fig. 2). This configuration was stable and was able to renew itself under limiting cell dilutions, these co-cultures were designated as MAM-1.

Passage of MAM-1 stock cultures at a 1:5 ratio maintained this organized growth pattern to reproducibly generate co-cultures that consisted of 50% tumor cells and 50% stromal cells. Cultures maintained in this way achieved 95% confluence within 5–7 days reproducibly for out to 20 passages. There was a tendency for stromal cells to proliferate more rapidly. Excess stromal cell growth was controlled by a mild trypsinization that released the superficial layer of stromal cells from the culture. At days 3–5 post-confluence, tumor cell nests formed spheroids that eventually pinched off and were capable of re-establishing the MAM-1 configuration when plated on fresh tissue culture plastic. Thus, spheroid formation is a natural progression in this co-culture model and does not require the elaborate procedures or growth conditions described for other model systems [8, 11, 12]. MAM-1 is also capable of growth in soft agar. However, if not in association with MAM-1 tumor cells, stromal cells will tend to form monolayers on top of or beneath the agar. One reason for the faithful renewal of this self-contained co-culture model can be attributed to the properties of syngeneic tumor associated stromal cells which have a strong influence on mammary tumor cell growth and gene expression [8, 11, 18]. Further studies are underway, to establish the genetic relationship between the cells of the stroma and tumor (i.e. karyotype analysis) and to determine the potential of each cell type to develop tumors in vivo.

As suggested by immunohistochemical analysis of the primary tumor (Fig. 1) we expected the tumor cells in these co-cultures to express HER2/neu and the stromal cells to express α-SMA. We observed that expression of HER2/neu was exclusive to the tumor cell nests with a high level of distribution in the membrane and cytoplasm (Fig. 2). HER2/neu expression was at background level in the stromal cells, and α-SMA was strongly expressed in the cytoplasmic actin-based microfilaments exclusively in these cells. In this co-culture setting, strong and consistent expression of both antigens, tumor-associated HER2/neu and stromal-associated α-SMA were stable out to 20 passages. Using these two antigens as markers to identify and separate tumor from stroma cells, we adapted our staining method for flow cytometric analysis that would enable us to evaluate responses in each cell type.

As shown in Fig. 3A, the MAM-1 co-culture is readily fractionated into a HER2/neu+, α-SMA-tumor subpopulation that accounts for 50–55% of the culture, a HER2-, α-SMA+ stromal cell population that accounts for 40–45% of the culture, a HER2/neu+, α-SMA+ population which may represent a stem cell or cells undergoing mesenchymal transition and typically accounts for 3–5% of the culture and finally a double negative population that appears to represent a fibroblast population that typically accounts for 3–5% of the culture.

Whether or not a "true" mammary stem cell exists in this co-culture is under investigation. A variety of candidate mammary stem cell markers are expressed by the different subpopulations. In mature MAM-1 co-cultures we have determined that the tumor cell population is HER2/neu⁺, CD24^lo/med, CD29^hi, SMA^- and the stromal population is HER2/neu^-, CD24^neg, CD29^hi, SMA⁺. Since CD24 negative, lo and hi populations correspond to nonepithelial, basal/myoepithelial and luminal epithelial cells respective [19], we consider our "stromal" cells to be nonepithelial and our "tumor" cells to be a mixture of luminal and basal/myoepithelial phenotypes. In so much as CD24⁺CD29^hi phenotype is enriched in its ability to reconstitute the essential elements of the mammary gland, a mammary tumor stem cell-like subpopulation may be present in these MAM-1 cultures. Since the "oncogenic" activated rat HER2/neu genetic lesion is expressed in the mammary stem cell of the donor BALB-NeuT transgenic mice, it is possible that this co-culture model may support a limited degree of cell differentiation from a stem-like progenitor that is in the HER2/neu positive subpopulation. Further support for a stem cell component in this self-renewing co-culture model is provided by the high level of CD44 in MAM-1 and enriched expression of vimentin and cytokeratin 19 in MAM-1 compared to the cloned Bam1a cell line described later.

Using flow cytometric analysis we were able to further subfractionate the HER2/neu (tumor) population based on their Forward Scatter Profile, which is an index of cell size. These different size tumor cell subpopulations may represent cells enriched at different phases of the cell cycle, smaller cells tend to be enriched for cells in G0 and larger cells tend to be enriched for cells in G2/M (not shown). We typically observe an equal distribution of small (S) and large (L) cell subpopulations in the mammary tumor cell fraction (Fig. 3B). We observed that the larger cell population (L) had approximately >3-fold higher level of HER2/neu expression on the cell surface. If this subpopulation represents a high proportion of cells in G2/M antigens and drug responses that are differentially sensitive to the phase of cell cycle can be detected/identified in the different sized tumor subpopulations (by forward scatter) and then confirmed and correlated with further analysis of the cell cycle distribution. For example, MAM-1 co-cultures that are treated for 1 hour with Iressa show a redistribution of p-c-Jun (Ser63) to the nucleus (Fig. 4A) and HER2/neu to the cytosol. In addition to redistribution from the cytosol, there was an overall decrease tumor cell associated phospho-c-Jun (i.e., in the α-SMA negative cells) but not in the α-SMA positive stromal cells (Fig. 4B). The observed 45% decrease in overall tumor cell phospho-c-Jun resulted from a 46% decrease that corresponded to the small cells and a 38% decrease in the large cells. Thus, small, non-dividing cells are about 20% more responsive to Iressa in terms of c-Jun phosphorylation. This further implies that large, dividing cells, which are enriched in HER2/neu receptors (Fig. 3B) and have a higher baseline level of phospho-c-Jun also have transient resistance to Iressa at this stage of the cell cycle. It has been widely documented that dividing cells often have a higher level of intrinsic resistance to a variety of chemotherapeutic agents.

We defined the MAM-1 transcriptome of highly expressed genes by comparing total RNA from a MAM-1 co-culture to a commercial preparation of universal mouse RNA which provides a balanced representation of normal and cancerous mouse tissues and cell lines. We used a cutoff value of >2-fold to be considered over-expressed and a p value cutoff > .05. A comprehensive listing of >2,000 known genes generated in this comparison is provided [see Additional file 1].

To identify genes associated with the tumor cell and stromal cell subpopulations in these cultures we compared the gene expression profiles of the Bam1a cell line, a cloned and characterized mammary carcinoma cell line developed from a BALB-NeuT mouse [6] and MAM-1 co-cultures. Bam1a was established from a soft agar colony and found to be immortal in vitro and tumorigenic in vivo and highly sensitive to Iressa and anti HER2/neu antibodies both in vitro and in vivo [6]. We identified clusters of genes that were highly expressed and common to Bam1a and MAM-1 co-cultures or differentially expressed between Bam1a and MAM-1. Table 1 is a representation of genes that are highly expressed by Bam1a and MAM-1 and part of the "ErbB-2 Signature" that is associated with many ErbB-2 expressing breast cancers [20‐23]. In most instances the relative expression level of these signature genes was similar (within 2-fold) between Bam1a and MAM-1. This two-fold difference is likely to represent the dilution of tumor cell RNA with stromal cell RNA in the MAM-1 co-culture.

Genes uniquely over-expressed by MAM-1 largely reflect the stromal signature of this breast cancer co-culture system. A select list of ~563 differentially expressed genes is provided [see Additional file 2]. We considered only differences greater than 3-fold to compensate for the dilution of tumor and stroma specific RNA in the MAM-1 cultures when compared to the cloned cell line, Bam1a. Certain genes that are uniquely over-expressed in Bam1a are likely to reflect the influence of co-culture on the gene-expression patterns. Paradoxically, for example, we observe >25-fold higher expression of EPSTI1 in Bam1a compared to MAM-1, which contradicts what is typically observed and expected [12]. A majority of the genes that are over-expressed in MAM-1 have been identified in tumor associated fibroblasts and stromal cells and represent genes involved in the fibrotic response and basement membrane synthesis [24, 25]. In particular, collagen genes involved in fibrosis and contraction and growth factors that stimulate the fibrotic response. In addition, genes involved in remodeling the extracellular matrix, including ADAM and MMP family members are highly represented.

When we compared genes differentially expressed between Bam1a and MAM-1 to genes clusters used to determine the stromal signatures of breast cancers, we found that relative to Bam1a, MAM-1 over-expressed 70% of the genes associated with the desmoid-type fibromatosis signature described by West et al. [24] including WISP2, COL1A1, COL5A1, COL3A1, COL6A1, MMP23, MMP19, CNN1, CTGF, ADAM19, FBN1 and ADAM12 (Table 2). This cluster of stroma specific genes, also identifies subgroups of breast carcinomas with a more favorable outcome when compared to the solitary fibrous tumor cluster [24] Further analyses revealed similarities between MAM-1 and an invasion specific cluster that is associated with the desmoplastic response to invading breast cancer [25] (Table 2). In particular, we observe >20-fold over-expression of COL1A1, IGFBP7 and SPARC in MAM-1 which are correlated with panstromal and juctatumoral stromal cell responses involved in matrix remodeling and angiogenesis, respective. These data are consistent with the histological features of the primary lesion used to establish MAM-1 and demonstrate the preserved expression of invasion specific genes in the MAM-1 co-culture. We also observed upregulation of a variety of stroma specific genes associated with tumor progression, invasion and the malignant phenotype in MAM-1. These include, MMP2 [10, 18]ADAM12 [26], HAS2, [27] and CLIC4 [28, 29]. Others have also demonstrated that heterologous co-cultures of breast tumor cells with tumor-derived fibroblasts (but not normal skin fibroblasts) was required for myofibroblast differentiation (i.e., expression of alpha-SMA) and resembled the advanced stages of desmoplastic carcinomas [11], similar to what we observe in the homotypic MAM-1 co-culture model. In these studies, expression of PMP22 was identified as a candidate gene in modulating tumor cell interactions with fibroblast. In MAM-1, PMP22 is upregulated approximately 15-fold and thus, may also play a role in the MAM-1 model.

Western blot and microarray analyses revealed differential expression of the HER2/neu and EGFR in the respective tumor and stroma fractions of MAM-1 co-cultures. HER2/neu was exclusively expressed in the mammary tumor cell fraction and EGFR was approximately 8-fold higher in the stromal cell population. In so much as the EGFR is approximately 50-fold more sensitive to Iressa (IC₅₀>0.080 μM) compared to the c-erbB-2 tyrosine kinase (IC₅₀~3 μM) in vitro, we anticipated that EGFR in the stromal cell population was a likely target for the inhibitory activities of Iressa. Indeed, it has been suggested that targeting tumor associated fibroblasts [30] and the stromal EGFR may be a viable approach for preventing tumor progression and impeding the angiogenic switch. It has been shown that Iressa can differentially modulate fibrosis in various models of injury and disease [31‐33]. It is important to distinguish, that in most injury models, the fibroblasts that are targeted as effectors of fibrosis are normal stromal fibroblasts and not phenotypically similar to the cancer associated fibroblasts/myofibroblasts that represent the stromal cells in the MAM-1 model. We expected that the response to Iressa in the MAM-1 stromal cells could be effected by their intrinsic sensitivity to Iressa and the microenvironment. Thus, we used the MAM-1 co-culture model to determine the selectivity and specificity of Iressa for the HER2/neu over-expressing mammary tumor cells versus the EGFR-expressing stromal cells by evaluating signal transduction downstream of these RTKs.

In MAM-1 co-cultures that are stimulated by replacing the culture medium, strong cytoplasmic expression of phosphorylated p44/42 MAPK is readily observed (Fig. 5A). Inclusion of 1 μM Iressa eliminates this response. By flow cytometric analysis we determined the dose-response for p44/42 MAPK and pMEK1/2 in the ErbB-2+ (tumor cell) and ErbB-2-(stromal cell) subpopulations in MAM-1 co-cultures. We observed a dose-dependent decrease in pp44/42 MAPK and pMEK1/2 phosphorylation in tumor cells with maximal decreases of 90% and 40%, respective (Fig. 5B). We also observed a modest (38%) decrease in stromal cell phospho-pp44/42 MAPK at all doses of Iressa but no effect of pMEK1/2 phosphorylation, suggesting a small inhibitory effect of Iressa on the EGFR in the stroma. To determine the long term impact of these effects on cell growth and survival we treated MAM-1 co-cultures with Iressa for longer periods of time.

When treated for an extended period of time with Iressa, the morphology of the MAM-1 co-culture recapitulated a fibrotic response such that tumor cell nests and islands gradually eroded away and stromal cells increased in density forming multi-cell layer nests (Fig. 6A). Our primary observation was that the morphology and cellularity of the co-cultures was dramatically altered. Within 24 h of treatment with 1 μM Iressa, there was a decrease in the cellularity of tumor cell nests and an increase in the cellularity and density of α-SMA reactive material associated with the stromal cell layers (Fig. 6B). Decreased cellularity of the tumor cell nests is accompanied by significant tumor cell rounding and apoptosis as evidenced by nuclear fragmentation shown with DAPI staining (Fig 7A) as well as positivity for Annexin V binding and cleaved caspase 3 (not shown). In addition to apoptosis, when probed for PCNA, there was a marked reduction in tumor cell PCNA and robust staining of PCNA in the stromal cells (Fig. 7A). Basic flow cytometric evaluation of these cultures demonstrated a 44% reduction in the tumor cell population within 24 h of treatment with 1 μM Iressa and a >3-fold increase in the stromal cell population when compared to control cultures (not shown). When we evaluated the PCNA, phospho-p44/42 MAPK and phospho-MEK1/2 levels in the ErbB-2 positive and ErbB-2 negative subpopulations we observed a 62%, 54% and 27% reductions in tumor cell PCNA, phospho-p44/42 MAPK and phospho-MEK1/2, respective (Fig 7B). Interestingly, the larger subpopulation of tumor cells in these treated co-cultures were approximately 2-fold less responsive to Iressa in terms of PCNA and phospho-p44/42 MAPK levels attesting to the transient resistance afforded to cells likely in G2/M (not shown). In contrast to ErbB-2 positive tumor cells, ErbB-2 negative stromal cells had a robust increase in all of these proliferation markers following treatment with Iressa. PCNA levels increased by 205%, phospho-p44/42 MAPK increased 219% and phospho-pMEK1/2 increased by 279%. These data are in agreement with our qualitative observations that used immunofluorescence to document the relative density of the tumor and stroma cell populations in the MAM-1 co-cultures (Fig. 6B). The resilience of the stromal cell population is not limited to EGFR antagonists. We have also observed that while both the tumor and stroma cells express comparable levels of the TRAIL death receptor DR5, only the tumor cells are sensitive to TRAIL-mediated apoptosis (data not shown). Thus, targeting the stroma to effectively "sterilize" the microenvironment and inhibit tumor growth and progression presents a major challenge.

To generate a better understanding of the global response of MAM-1 to Iressa, we compared the gene expression profiles of MAM-1 co-cultures treated for 24 h with fresh media containing diluent or 1 μM Iressa. We observed a strong upregulation of inflammatory genes involved in the fibrotic response, consistent with the morphology of the treated MAM-1 co-cultures (Fig. 6A). Interestingly, none of the genes that were modulated by Iressa in Bam1a cells [6] were altered in MAM-1. We observed 2.5- to 6-fold increases in the expression of genes associated with inflammatory, fibrotic, pathological processes that are observed in a variety of diseases [34‐37]. Most notably, we observed increases in inflammatory genes, (C9, CCR7, TNFRS25, IL1F9, CCL22, TLR8, KLRA23, CLEC4N, CD22), cell adhesion genes, (COL4A3, SIGLEC5, LGALS12, CLDN5) growth factor signaling and transcription genes, (FGF22, CTGF, WNT8A, POU4F3, HEY1) proliferation and differentiation (CCNB3, MYB, TFF1, PROX1) and proteolysis genes (GZMK, TLL2, ADAM5, SERPINA1D).