Seminars in Cancer Biology 15 (2005) 405–412
Review
Gene expression perturbation in vitro—A growing case for three-dimensional (3D) culture systems Anna Birgersdotter 1 , Rickard Sandberg 1,2 , Ingemar Ernberg ∗ Microbiology and Tumor Biology Center, MTC, Karolinska Institutet, Box 280, 171 77 Stockholm, Sweden
Abstract Cells grown in vitro are dramatically perturbed by their new microenvironment. Analyses of genome-wide gene expression levels offer a first glance at which genes and pathways are affected in cell lines as compared to their tissues of origins. We have summarized available gene expression data and review how cell lines adapt to in vitro environments, to what degree they express markers of their tissues of origins and discuss how cells grown in three-dimensional (3D) cultures may have more physiological interactions with neighbouring cells and extracellular matrix. We will also discuss the interplay between malignant cells and stroma present in tumours but lacking in cell lines and how these differences might affect gene expression comparisons of cell lines to tumours. A model simulating impact of stromal cells on gene expression profiles is presented. Understanding the transcriptomes of cells grown in 2D and 3D cultures and how they compare to those of in vivo cells are important for improving cell line model systems and for the reconstituting of tissues in vitro. © 2005 Elsevier Ltd. All rights reserved. Keywords: Tumor stroma; Global gene expression; Transcriptome; Cancer cell lines; 3-D cell culture
Contents 1. 2. 3. 4. 5. 6. 7.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transcriptional changes associated with cell line adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lessons from 3D cultures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tissue identity present in cell lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tumour stroma interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model of cell type effects in comparisons of cell lines to tissues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction The culturing of animal and human cells outside of the body was successfully accomplished during the first half of the last century when the necessary growth medium condi∗
Corresponding author. Tel.: +46 8 5248 6262; fax: +46 8 319 470. E-mail address:
[email protected] (I. Ernberg). 1 These authors contributed equally. 2 Present address: Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02319, USA. 1044-579X/$ – see front matter © 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.semcancer.2005.06.009
405 406 407 408 408 409 410 410 410
tions were gradually established [1]. The cell growth medium, together with the solid glass or plastic support, is required to both replace the surroundings of the tissue components [2] and supply the nutrients necessary [3] for the cells to survive and divide. This knowledge has been instrumental for keeping cell lines alive, which made possible their decisive contribution to the advance of molecular biology. Although cell lines are often used as substitutes for tissues, it is obvious that cell lines only approximate properties of normal and tumour tissues. Moreover, this approximation is limited to the malignant
406
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
cell and does not take into account the impact of the tumour stroma. The paradigm for the development of cancer is a Darwinian multi-step process, during which a cell acquires multiple mutations. However, it becomes more and more apparent that the growth deregulation within a tumour can only be explained once we understand the contributions of and interactions with the microenvironment. The surrounding cells such as fibroblasts, endothelial cells, immunocells, neurons and others together with extracellular matrix are active participants in shaping the tumour [4]. Malignant tumour cells recruit or activate the local vasculature and stroma through production and secretion of stimulatory growth factors and cytokines. The locally activated host microenvironment, in turn, modifies the proliferative behaviour of the tumour cells, as well as supporting the inflammatory-like response. In fact, these interactions are likely to be as important in the tumour tissue as in any normally functioning organ or tissue, in which no one would dream of omitting contributions of different cell types. The difference lies in that the tumour tissue is orchestrated by this one mutated tumour cell and that the tissue is reorganized to support growth of that same cell. In tissues cells connect to each other as well as to the extracellular matrix (ECM). The ECM is a support structure that contains proteins such as collagen, elastin and laminin, and these gives the tissues their mechanical properties and help to organize communication between cells embedded in the matrix. Receptors on the surface of the cells, i.e. integrins, anchor to the ECM and also determine how the cells interprete biochemical cues from the immediate surroundings. Given this complex mechanical and biochemical interplay, important biological properties are missed if they are only studied in two-dimensional (2D) cell cultures. These 2D cultures, as conveniently used for the maintenance of cells and for biological studies, impose highly unnatural geometric and mechanical constraints many types of cells. However, 3D culture might be a tool to bridge this gap in behaviour. Most of the cell lines used in biomedical research have been derived from tumours (often even from a metastasis) and in many cases it is not clear which cells from the heterogeneous tumour tissue that ‘grew out’ in culture to establish the cell line. The cells that do grow in vitro adapt to the new microenvironment, by changes at the genetic [5], transcriptional [6] and protein [7] levels. Thus, the degree to which cell lines are representative of the tumour they are derived from varies [8,9]. Many cell lines have been in culture for several years or even decades, imposing a strong in vitro selective pressure on them. As they have been distributed in several labs, the same cell lines might also have undergone various selection steps due to different feeding techniques, etc. The limited availability of tissue samples and the restricted possibility to manipulate in vivo cells makes cell lines an instrumental tool also for future biochemical and molecular cell biology research and drug development. Detailed comparisons are, however, needed to assess how the genotypic and phenotypic characteristics of cell lines correspond to those of the tumour tissues they were derived from.
DNA microarray technologies offer a possibility to analyze how the global gene expression patterns of cell lines reflect that of tissues. In this review we will summarize the results from comparisons of the transcriptional activities in cell lines to tissues. We will discuss the transcriptional changes associated with cell line adaptation to in vitro environment and tissue identity present in cell lines, considering the growing possibilities of 3D culture as a tool to bridge between the two worlds. After reviewing the literature on how stroma and bystander cells modulate tumour phenotypes, we will introduce a simplified model of how stroma and bystander cells influence the gene expression measurements.
2. Transcriptional changes associated with cell line adaptation Gene expression comparisons of tumours and normal tissues with immortalized cell lines have highlighted some of the transcriptional modifications that occur in response to the in vitro environment. Comparisons of cell lines to their corresponding tumour tissues and normal tissue of origin have been performed for many different tissues, including colon [10], breast [11], lymphoma [12], leukemia [6], lung origin [13], ovaries [14] and prostate [15]. Hierarchical clustering of gene expression profiles of both cell lines and tissue samples have repeatedly shown that cell lines are separated from tissue samples (Fig. 1) [11–16]. The number of differentially expressed genes in cell lines as compared to tissues has been estimated to approximately 30% [13,16]. Early pioneering studies identified important features in cell line gene expression that were different from tissues or primary cultures. Comparing colon cell lines to colon tissues revealed that cell
Fig. 1. The gene expression profiles of cell lines compared to tissues. Projection of cell lines and tissue samples in “singular value decomposition (SVD) space” drawn by the correlation of each sample to SVD eigenarray 1 (x-axis) and 2 (y-axis) after singular value decomposition. The cell lines were separated from tissue samples by the first SVD eigenarrays, showing that the largest gene expression pattern difference within the data was a separation of cell lines and tissues. For method description and details see Sandberg and Ernberg [16] from which the figure was adapted.
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
lines up-regulate ribosomal proteins [10]. A ‘proliferation cluster’ was shown to be up-regulated in cell lines of breast origin when compared to breast tissue [11]. The up-regulation of ribosomal proteins and proliferation-associated genes has been confirmed by later studies [6], some of which also identified metabolic genes to be up-regulated in cell lines [16,17]. The most detailed description of the changes in gene expression in cell lines was achieved by meta-analysis of 60 cell lines and hundreds of normal and tumour tissue samples of multiple origins [16]. From our analysis it was shown that upregulated genes in cell lines fall into three broad categories: (i) cell cycling; (ii) metabolism and; (iii) turnover of macromolecules. Genes that encode for proteins that are directly or indirectly involved in cell cycling was up-regulated in cell lines, e.g. regulators of different phases of the cell cycle, DNA replication, microtubule, mitotic spindle. The higher metabolism is evident from the up-regulation of genes that involve in the glucose catabolism, tricarboxylic acid cycle and oxidative phosphorylation. The higher turnover of macromolecules includes genes involved in distinct cellular processes, such as translation, splicing, protein modifications and degradation. In summary, the gene expression programs that are up-regulated in cell lines enable the in vitro cells to rapidly grow and divide and to respond to growth factors in the culture medium. Cell lines also repress the expression of genes limiting their growth potential [18] and genes not necessary for growth in the in vitro environment. Cell adhesion molecules, cell–cell contact with extracellular matrix and membraneassociated signalling molecules have been reported to be down-regulated in cell lines [16,17,19]. These proteins have probably little selective value in cell cultures and are therefore repressed during cell line selection and adaptation. Interestingly, the same types of genes were again up-regulated, when immortalized cell lines were transplanted into mice [17].
3. Lessons from 3D cultures This data supports the need for molecular studies of cells in 3D cultures including gene expression comparisons from cell lines grown in 2D and 3D cultures. Given the complex mechanical and biochemical interplay, biological properties might be missed if they are only studied in two-dimensional cell cultures. 2D cultures impose highly unnatural geometric and mechanical constraints to many types of cells. Fibroblasts in 2D cultures differ dramatically in behaviour from those in a 3D environment of a multicellular organism. The lack of dorsal receptor anchorage in 2D cultures tips the balance between spreading and retraction and creates an overall stimulatory environment for the organization of lamellipodia, stress fibers and focal adhesions. This imbalance would then cause cells to spread out in an extreme manner. The spatial arrangement of ECM receptors is in 2D cultures concentrated to the ventral surface while in 3D cultures they are spread over the entire surface. Fibroblasts in 3D cultures may
407
enhance a global retraction signal, causing the cells to adopt a highly elongated morphology [20]. Focal and fibrillar adhesion of fibroblasts differed depending on if they were grown in 2D substrates as compared to 3D substrates in the content of ␣51 and ␣V3-integrins, paxillin, other cytoskeletal components and phosphorylation of focal adhesion kinase (FAK) [21]. Other data suggest that decreased cell spreading and the consequential dephosphorylation of FAK leads to the up-regulation of p21, which may be responsible for the lower proliferation rate seen in smooth muscle cells when they are grown in a 3D matrix [22]. A hallmark of 3D epithelial tissue architecture in vivo is polarity, which is essential for the maintenance of tissue function. The loss of tissue polarity and increased proliferation is characteristic of breast tumour phenotype. When transformed mammary epithelial cells (MEC) were grown in 3D culture they regained polarity and proliferation was suppressed. By culturing cells in 3D it could be demonstrated that the PI3K pathway plays a distinct role in maintaining tissue polarity and suppressing cell proliferation [23]. It has been shown that tumourgenic breast cancer cells (T4-2) have increased integrin 1 expression, which is associated with loss of growth control and perturbed morphogenesis. A reduction of cell surface integrin 1 activity by inhibitory antibodies was found to be sufficient to revert this tumour phenotype in 3D cultures. Also, T4-2 cells in 3D cultures showed a reciprocal interaction between integrin 1 and epidermal growth factor receptor (EGFR). Antibodies directed to either of them also decreased the activity of the other receptor. This data suggest that in 3D cultures, growth and adhesion are coupled and mutually excluding [24,25]. Enhanced integrin 1 expression was shown to be associated with increased tumour aggressiveness, invasiveness and metastatic potential and was found to correlate with decreased patient survival. Integrin ␣64 interacts with intermediate filaments and facilitates hemidesmosome assembly and regulates tissue polarity. Polarity is associated with a basal localization of integrin ␣64 heterodimers in keratinocytes, which has been shown to direct intermediate filament organization, while its activity correlates with an increase in the level of p21 (waf-1). Polarized integrin ␣64 in breast cancer cell acini are consistent with these findings and emphasize the existence of coordinate integrin 1/4 pathway interactions [24]. Cancer cells undergoing metastasis are involved in coordinating adhesion and proteolytic interaction with the ECM substrate, resulting in the degradation and remodelling of interstitial tissue barriers. During tumour progression, multiple classes of ECM-degrading enzymes are up-regulated and activated including matrix metalloproteinases (MMPs), serine proteases, and cathepsins. Yet, in clinical trials, inhibitors targeting MMPs and serine proteases have not been beneficial. This might be explained by observations from studies made in 3D cultures. When ECM degrading enzymes were blocked, the cells changed shape into an amoeboid morphology and squeezed themselves through gaps in the matrix. This
408
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
switch is termed mesenchymal-amoeboid transition (MAT) [26]. Gene expression profiles of cells grown in 2D and 3D support some of the molecular findings. Expression profiles comparing vascular smooth muscle cells (SMCs) grown in 2D versus 3D, differed for 99 genes. p21 had a higher expression in cells cultured in 3D, correlating to their decreased proliferation rate (although c-jun and c-fos had a higher expression). Other genes included matrix proteins collagen 1 and fibrinogen indicating that SMCs were more active in ECM production in 3D matrix. Some genes involved in matrix remodelling such as lyosyl oxidase and MMP-1 did not change with matrix geometry. However, there was no significant change in integrin ß1 and ß3 subunits [22]. 3D culture of melanoma cells profoundly affected the gene expression, with 173 genes differing between cells grown in 2D and 3D. Genes that were strongly up-regulated were mostly chemokines, but also laminin and c-Jun. Genes with lower expression in 3D cultures included integrin 4 and fibroblast growthfactor 2, FGF-2. [27].
4. Tissue identity present in cell lines Cell-specific gene expression is of major importance in establishing tissue-specific functions. Systematic studies of gene expression have been pursued in a large number of human and mice tissues [28,29], revealing that between 3 and 6% are ubiquitously and highly expressed in all tissues, while 3% are tissue-specific (only detected in a single tissue). It has been debated how well cell lines keep the tissue identity and/or tumour characteristics while growing in vitro [8]. When transplanting cell lines from diverse tumour origins into nude mice, most cell lines formed tumours and the histopathology of the tumours correlated with the histopathology of their origins [30]. The expression of breast-specific genes (e.g. estrogen receptor) has been detected in breast cancer cell lines [31,32]. Analysis of the gene expression in 60 different cell lines (NCI60) showed that cell lines from six of the nine tissues of origins were clustered into independent terminal branches with few exceptions [6]. The melanoma-derived cell lines showed the most tissue similar gene expression pattern with many genes involved in melanocyte biology being up-regulated in the cell lines [6]. Comparing lymphoma and leukemia samples from patients to immortalized cell lines showed that some lymphoma-derived cell lines have maintained a germinal centre like gene expression [12], despite the fact that these cell lines have grown in vitro and separate from the germinal centre environment for considerable time. Many gene expression studies have, however, shown that cell lines in general have lost most of the tissue-specific expression of the in vivo tumours. When comparing 21 prostate cancer-derived cell lines to normal and malignant prostate tissue [15], we concluded that: “Overall, we found only a small number of genes with concordant expression in cell lines and malignant
tissues, which suggests that these cell lines have lost many features that characterize prostate cancer in vivo.” A study on ovarian cell lines using SAGE and tissues arrived at a similar conclusion [19]. Gene expression in different lung tumours and lung cancer cell lines have indicated that certain tumour subtypes may be better captured in cell line models than other subtypes. The tumour-specific gene expression seemed to be better preserved in cell lines from squamous cell carcinoma and small-cell lung cancers than from adenocarcinomas [13]. Comparisons of gene expression in primary breast cell cultures, immortalized breast cells and breast tumours showed that primary cultures were more similar to the gene expression of the original tumour [18]. In summary, most of the global gene expression studies demonstrate that immortal cell lines only reflect a limited part of the tissue-specific expression of genes of the malignant in vivo tumour. Immortalized cell lines seem to better capture aspects of lymphomas, leukemias and melanomas [6,12]. For solid tumours, the difference between tumours and cell lines is larger. It is possible that leukaemia- and lymphoma-derived cell lines are not perturbed by the in vitro environment to the same degree as solid-tissue-derived cell lines are. It is interesting that a specific tumour subtype may be better represented by cell lines than other tumours of the same tissue (e.g. in lung). It is possible that the malignant cells of some tumour types do not easily grow in vitro, or that cell lines derived from particular tumour origins are more prone to lose the tissue-specific gene expression. Since, cell lines have been propagated through decades there is always a risk of contamination. In fact, when systematically analysing over 550 cell lines from researchers it was demonstrated that 15% of lymphoma and leukaemia cell lines used were false [33]. Thus, it is important to authenticate cell line identities [34,35]. We have recently introduced a practical tissue similarity index that assesses how each cell line reflects the gene expression of different tumours [35]. We demonstrated the usefulness of our index by examining the NCI60 cell lines towards their corresponding tumour origins. Thus, we identified the cell lines that had gene expression profiles that closely resembled those of their corresponding in vivo tumour. Moreover, we identified cell lines that did not show any similarities towards their presumed tumour origin, as well as putative misclassifications of cell lines [35]. We believe that methods, such as the tissue similarity index, will prove very useful in the selection of cell lines for particular experiments (a high scoring cell line which is at the transcriptional level more similar to its original tumour than low scoring cell lines). Even more importantly, these types of genome-wide quantitative methods highlight that the choice of cell lines is important.
5. Tumour stroma interactions There is no doubt that the tumour stroma is an active participant in tumour development. Functional studies in various cancer types, including breast, colon, prostate and lung can-
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
cer, have confirmed the concept that fibroblasts can determine the fate of epithelial cells, since they are able to promote malignant conversion as well as to revert tumour cells to a normal phenotype. Both in vivo and in vitro studies have demonstrated that fibroblasts contribute to tumour formation and growth rates [36]. Fibroblasts have been inoculated with breast or bladder tumour cell lines in nude mice, and this resulted in shortened tumour latency and increased tumour growth [37]. Fibroblasts cultured from malignant tumours have stimulatory effects on MCF-7 cells, whereas fibroblasts cultured from normal tissue are inhibitory [38]. In invasive ductal breast carcinoma, the presence of stromal myofibroblasts correlates significantly with pathological parameters associated with a poor prognosis [39]. Phase I and II clinical trials are currently using an antibody called Sibrotuzumab to target fibroblast activation protein (FAP) in metastatic lung and colon cancers [40]. FAP is uniquely expressed by tumour stromal fibroblasts in epithelial carcinomas, but not by the epithelial carcinoma cells, normal fibroblasts, or other normal tissues. The most studied stromal impact on tumour development is angiogenesis. The current theory is that tumours cannot grow beyond 1–2 mm unless they are supported by a rich vascular supply. The angiogenetic switch is considered to be a part of tumourgenesis [41]. Tumour angiogenic activity (TAA) is an important prognostic factor in many human tumours. In general, carcinomas that are highly vascularized take a more aggressive clinical course than carcinomas of low vascularization. Genetic and functional experiments indicate that macrophages, neutrophils, mast cells, eosinophils and activated T cells contribute to malignancies by releasing extracellular proteases, pro-angiogenic factors and chemokines [42]. Both chemokines and neurotransmitters affect the metastatic properties of tumour cells. Macrophages produce a number of potent angiogenic and lymphangiogenic growth factors, cytokines and proteases, all of which are mediators that potentiate neoplastic progression. Macrophages are recruited through the local expression of chemoattractants such as colony stimulating factor 1 (CSF-1) and macrophage chemoattractant protein 1 (MCP-1). Overexpression of both of these factors is correlated with poor prognosis in breast cancer [43]. However, tumour-associated macrophages (TAMs) have been also shown to display positive effects on survival in gastric cancers [44]. T cells have been implicated to play a role in Hodgkin’s lymphoma (HL). T cells from nodes affected by Hodgkin lymphoma are anergic, and can profoundly inhibit perhipheral immune cell responses, and probably inhibit immune cells in the tumour tissue as well [45]. Mast cells accumulate around tumours and could either promote or inhibit tumour growth depending on the local stromal conditions. The presence of mast cells in HL is associated with poor prognosis [46]. Mast cells have been shown to be responsible for angiogenesis in squamous epithelial carcino-
409
genesis in mice, by activating MMP-9 and therefore tilting the angiogenic switch [47]. Kinzler and Vogelstein [48] have broadened the nomenclature that exists for cancer-associated genes into genes that monitor growth referred to as “gatekeepers”, genes that indirectly suppress neoplasia by regulating genomic instability called “caretakers”, and a third set of genes recognized as enabling genes or “landscapers” and who might affect nontarget cells. For example, patients suffering from juvenile polyposis syndromes (JPS) or ulcerative colitis (UC) develop hamartomas that consists of proliferating stromal cells. The epithelial cells associated with the polyps are more likely to undergo neoplastic transformation, as a result of an abnormal microenvironment. One of the features of the gastrointestinal polyps seen in Apc-Smad4 mutant heterozygote mice is the increased proliferation of stromal cells, which is one of the characteristic features of juvenile polyps seen in humans [49]. In colon cancer the abundance of the PINCH protein in stroma increased from normal mucosa to primary tumour and metastasis, and was more intense at the margins of invasion than it was in the intratumoural stroma. Thus, stromal inflammation and proliferation eventually may lead to the development of malignant epithelial transformation because of an altered microenvironment. The microenvironment can sometimes be mimicked in vitro by co-culturing experiments. Although IL-6 has been identified as a major growth factor in multiple myeloma (MM), it is believed that maintenance of tumour growth in vivo depends on one or more additional stroma-derived factors. Treatment of human MM cells lines (IL-6-dependent MM cell line INA-6 and primary MM cells) with the IL-6 receptor antagonist Sant7 or with an anti-gp130 monoclonal antibody (mAb)-induced apoptosis if the cells were cultured in the absence of bone marrow stromal cells (BMSCs). In contrast, apoptosis could not be observed if the MM cells were co-cultured with BMSCs [50].
6. Model of cell type effects in comparisons of cell lines to tissues The enrichment of malignant cells and the lack of stroma in cell lines may bias gene expression comparisons between cell lines and tissues. We find it useful to use a simplified model to estimate the bias artefacts when comparing cell lines to tissues. Our model considers three cell types: a malignant cell, a tissue-specific cell and an additional cell which represents the additional stroma, infiltrating cells and bystander cells (Table 1). The cell lines, tumours and normal tissue are composed of different percentages of the three cells present in the model. We will examine the effect of different cell compositions on a set of idealized genes, representing oncogenes, tumour suppressor genes, stroma-enriched genes (Table 1). In our model, a cell line is assumed to be only composed of malignant cells, albeit adapted to in vitro con-
410
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
Table 1 In silico model of stromal impact of gene expression profiles
Oncogene Tumour suppressor gene Stroma-enriched gene
Relative gene expression in model cells
Observed fold changes
Malignant cell
Tissue-specific cell
Stroma and bystander cell
Cell linesa vs. tumours (50% malignant cells)
Cell lines vs. tumours (80% malignant cells)
Tumours (80% malignant cells) vs. normal tissueb
5 1 1
1 5 1
1 5 5
1.7 (5/3)c 0.3 (1/3) 0.3 (1/3)
1.2 (5/4.2) 0.6 (1/1.8) 0.6 (1/1.8)
4.2 (4.2/1) 0.4 (1.8/5) 1.0 (1.8/1.8)
a
Cell lines are assumed to be composed of only the malignant cell type (albeit adapted to in vitro environment). Normal tissues are assumed to be composed of 80% tissue-specific cells and 20% additional stroma, infiltrating cells and bystander cells. c A calculation example: cell lines are made of 100% malignant cells (1 × 5 = 5), tumour is in this comparison assumed to be composed of 50% malignant cells and 50% additional stroma and bystander cells (0.5 × 5 + 0.5 × 1 = 3). The fold change is subsequently 5/3, which equals 1.7. b
ditions. A tumour tissue would consist of malignant cells, non-malignant tissue-specific cells and additional stroma and bystander cells, while a normal tissue would be comprised of tissue-specific cells, stroma and bystander cells. In different tissues and tumours, the percentages of these cells vary [51]. By determining different percentages of cells within a specific tumour, the model can be used to analyze the effects of cell composition on the observed fold change values when comparing the tumours to cell lines. When analyzing the expression levels of genes in a tumour sample, the expression levels represent the ‘cellular average’ of the sample. We will now give examples of how the cell compositions might influence the observed gene expression profiles. The cell model demonstrates that if malignant cells constitute at least 50% of the tumour cell population (a standard criterion used in different studies), the genes up-regulated in malignant cells could appear up-regulated in vitro and the degree would depend on the real fold change difference between the malignant cell and non-malignant cells. This ‘artificial fold change’ could, however, never exceed 2 as long as the tumour tissues are comprised of at least 50% malignant cells. A true fold change of 5 would appear as a cell line up-regulation with a fold change of 1.7 (assuming 50% malignant cells in tumour) or 1.2 (assuming 80% malignant cells in tumours). There is a possibility that genes up-regulated or only expressed in stroma and bystander cells appear downregulated in cell lines due to the lack of these cells in culture. This effect could potentially result in an apparent downregulation of gene expression in cell lines. Again, using a simplified model we can get some rough estimates on how an up-regulation in a specific stroma cell would affect cell line to tissue comparisons. The cell model demonstrates that for a gene up-regulated five times in stroma cells which constitutes 20% of the tumour tissue, the observed fold change (down-regulation) in cell lines would be 1.8 (Table 1).
7. Conclusions The interactions between tumour cells and stroma affect tumour development and phenotypes. The relevant tumour
environments are hard to mimic in vitro. Therefore, the cell lines represent a model system in which only the malignant cells that have adapted to the 2D in vitro microenvironment are studied. Polarity is a tissue characteristic. Cells grown in 2D lose their polarity which affects their intracellular signalling pathways including those involved in proliferation. It seems that restored cell polarity, provided by 3D cultures, have effects on intracellular signalling and this effect is independent of how long time the cells have been in 2D culture. Therefore, it might be recommendable to consider 3D cultures in future cancer therapeutic screening experiments. As cells grown in 3D cultures show reduced proliferation rate, changes their adhesion structures, and changes the transcription levels of some adhesion molecules, it will be interesting to learn if cells grown in 3D cultures also express a higher degree of tissue-specific markers reflecting their origin.
Acknowledgements This work was made with support by the Swedish Cancer Society, Swedish Children Cancer Society, the Board for Internationalization of Science (STINT) and Karolinska Funds. A.B. has been the recipient of a fellowship from the Karolinska Ph.D.-Company research Training School.
References [1] Carrel A. On the permanent life of tissues outside of the organism. J Exp Med 1912;15:516–28. [2] Sanford KE, Earle WR, Likely GD. The growth in vitro of single isolated tissue cells. J Natl Cancer Inst 1948;9(3):229–46. [3] Eagle H. Nutrition needs of mammalian cells in tissue culture. Science 1955;122(3168):501–4. [4] Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100(1):57–70. [5] Roschke AV, Tonon G, Gehlhaus KS, McTyre N, Bussey KJ, Lababidi S, et al. Karyotypic complexity of the NCI-60 drugscreening panel. Cancer Res 2003;63(24):8634–47. [6] Ross DT, Scherf U, Eisen MB, Perou CM, Rees C, Spellman P, et al. Systematic variation in gene expression patterns in human cancer cell lines. Nat Genet 2000;24(3):227–35.
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412 [7] Irish JM, Hovland R, Krutzik PO, Perez OD, Bruserud O, Gjertsen BT, et al. Single cell profiling of potentiated phospho-protein networks in cancer cells. Cell 2004;118(2):217–28. [8] Masters JR. Human cancer cell lines: fact and fantasy. Nat Rev Mol Cell Biol 2000;1(3):233–6. [9] Jacks T, Weinberg RA. Taking the study of cancer cell survival to a new dimension. Cell 2002;111(7):923–5. [10] Alon U, Barkai N, Notterman DA, Gish K, Ybarra S, Mack D, et al. Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays. Proc Natl Acad Sci USA 1999;96(12):6745–50. [11] Perou CM, Jeffrey SS, van de Rijn M, Rees CA, Eisen MB, Ross DT, et al. Distinctive gene expression patterns in human mammary epithelial cells and breast cancers. Proc Natl Acad Sci USA 1999;96(16):9212–7. [12] Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403(6769):503–11. [13] Virtanen C, Ishikawa Y, Honjoh D, Kimura M, Shimane M, Miyoshi T, et al. Integrated classification of lung tumors and cell lines by expression profiling. Proc Natl Acad Sci USA 2002;99(19):12357–62. [14] Schaner ME, Ross DT, Ciaravino G, Sorlie T, Troyanskaya O, Diehn M, et al. Gene expression patterns in ovarian carcinomas. Mol Biol Cell 2003;14(11):4376–86. [15] Welsh JB, Sapinoso LM, Su AI, Kern SG, Wang-Rodriguez J, Moskaluk CA, et al. Analysis of gene expression identifies candidate markers and pharmacological targets in prostate cancer. Cancer Res 2001;61(16):5974–8. [16] Sandberg R, Ernberg I. The molecular portrait of in vitro growth by meta-analysis of gene expression profiles. Genome Biol, in press. [17] Creighton C, Kuick R, Misek DE, Rickman DS, Brichory FM, Rouillard JM, et al. Profiling of pathway-specific changes in gene expression following growth of human cancer cell lines transplanted into mice. Genome Biol 2003;4(7):R46. [18] Dairkee SH, Ji Y, Ben Y, Moore DH, Meng Z, Jeffrey SS. A molecular ‘signature’ of primary breast cancer cultures; patterns resembling tumor tissue. BMC Genomics 2004;5(1):47. [19] Stein WD, Litman T, Fojo T, Bates SE. A serial analysis of Gene Expression (SAGE) database analysis of chemosensitivity: comparing solid tumors with cell lines and comparing solid tumors from different tissue origins. Cancer Res 2004;64(8):2805–16. [20] Beningo KA, Dembo M, Wang YL. Responses of fibroblasts to anchorage of dorsal extracellular matrix receptors. Proc Natl Acad Sci USA 2004;101(52):18024–9. [21] Cukierman E, Pankov R, Stevens DR, Yamada KM. Taking cell-matrix adhesions to the third dimension. Science 2001;294(5547):1708–12. [22] Li S, Lao J, Chen BP, Li YS, Zhao Y, Chu J, et al. Genomic analysis of smooth muscle cells in 3-dimensional collagen matrix. FASEB J 2003;17(1):97–9. [23] Liu H, Radisky DC, Wang F, Bissell MJ. Polarity and proliferation are controlled by distinct signaling pathways downstream of PI3-kinase in breast epithelial tumor cells. J Cell Biol 2004;164(4):603–12. [24] Weaver VM, Petersen OW, Wang F, Larabell CA, Briand P, Damsky C, et al. Reversion of the malignant phenotype of human breast cells in three-dimensional culture and in vivo by integrin blocking antibodies. J Cell Biol 1997;137(1):231–45. [25] Wang F, Weaver VM, Petersen OW, Larabell CA, Dedhar S, Briand P, et al. Reciprocal interactions between beta1-integrin and epidermal growth factor receptor in three-dimensional basement membrane breast cultures: a different perspective in epithelial biology. Proc Natl Acad Sci USA 1998;95(25):14821–6. [26] Wolf K, Mazo I, Leung H, Engelke K, von Andrian UH, Deryugina EI, et al. Compensation mechanism in tumor cell migration:
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34] [35]
[36]
[37]
[38]
[39]
[40]
[41] [42] [43]
[44]
[45]
[46]
[47]
411
mesenchymal-amoeboid transition after blocking of pericellular proteolysis. J Cell Biol 2003;160(2):267–77. Ghosh S, Spagnoli GC, Martin I, Ploegert S, Demougin P, Heberer M, et al. Three-dimensional culture of melanoma cells profoundly affects gene expression profile: a high density oligonucleotide array study. J Cell Physiol 2005. Su AI, Wiltshire T, Batalov S, Lapp H, Ching KA, Block D, et al. A gene atlas of the mouse and human protein-encoding transcriptomes. Proc Natl Acad Sci USA 2004;101(16):6062–7. Su AI, Cooke MP, Ching KA, Hakak Y, Walker JR, Wiltshire T, et al. Large-scale analysis of the human and mouse transcriptomes. Proc Natl Acad Sci USA 2002;99(7):4465–70. Fogh J, Fogh JM, Orfeo T. One hundred and twenty-seven cultured human tumor cell lines producing tumors in nude mice. J Natl Cancer Inst 1977;59(1):221–6. Wistuba II, Behrens C, Milchgrub S, Syed S, Ahmadian M, Virmani AK, et al. Comparison of features of human breast cancer cell lines and their corresponding tumors. Clin Cancer Res 1998;4(12):2931–8. Wistuba II, Bryant D, Behrens C, Milchgrub S, Virmani AK, Ashfaq R, et al. Comparison of features of human lung cancer cell lines and their corresponding tumors. Clin Cancer Res 1999;5(5):991–1000. Drexler HG, Dirks WG, Matsuo Y, MacLeod RA. False leukemialymphoma cell lines: an update on over 500 cell lines. Leukemia 2003;17(2):416–26. Masters J. False cell lines. Carcinogenesis 2002;23(2):371. Sandberg R, Ernberg I. Assessment of tumor characteristic gene expression in cell lines using a Tissue Similarity Index. Proc Natl Acad Sci USA 2005;102(6):2052–7. Cheng JD, Weiner LM. Tumors and their microenvironments: tilling the soil. Commentary re: Scott AM et al. A Phase I doseescalation study of sibrotuzumab in patients with advanced or metastatic fibroblast activation protein-positive cancer. Clin Cancer Res 2003;9:1639–47. Clin Cancer Res 2003;9(5):1590–5. Camps JL, Chang SM, Hsu TC, Reeman MR, Hong SJ, Zhau HE, et al. Fibroblast-mediated acceleration of human epithelial tumor growth in vivo. Proc Natl Acad Sci USA 1990;87(1):75–9. Adams EF, Newton CJ, Braunsberg H, Shaikh N, Ghilchik M, James VH. Effects of human breast fibroblasts on growth and 17 betaestradiol dehydrogenase activity of MCF-7 cells in culture. Breast Cancer Res Treat 1988;11(2):165–72. Yazhou C, Wenlv S, Weidong Z, Licun W. Clinicopathological significance of stromal myofibroblasts in invasive ductal carcinoma of the breast. Tumour Biol 2004;25(5–6):290–5. Scott AM, Wiseman G, Welt S, Adjei A, Lee FT, Hopkins W, Divgi CR. A Phase I dose-escalation study of sibrotuzumab in patients with advanced or metastatic fibroblast activation protein-positive cancer. Clin Cancer Res 2003;9(5):1639–47. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med 1971;285(21):1182–6. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002;420(6917):860–7. Leek RD, Lewis CE, Whitehouse R, Greenall M, Clarke J, Harris AL. Association of macrophage infiltration with angiogenesis and prognosis in invasive breast carcinoma. Cancer Res 1996;56(20):4625–9. Shunyakov L, Ryan CK, Sahasrabudhe DM, Khorana AA. The influence of host response on colorectal cancer prognosis. Clin Colorectal Cancer 2004;4(1):38–45. Marshall NA, Christie LE, Munro LR, Culligan DJ, Johnston PW, Barker RN, et al. Immunosuppressive regulatory T cells are abundant in the reactive lymphocytes of Hodgkin lymphoma. Blood 2004;103(5):1755–62. Molin D, Edstrom A, Glimelius I, Glimelius B, Nilsson G, Sundstrom C, et al. Mast cell infiltration correlates with poor prognosis in Hodgkin’s lymphoma. Br J Haematol 2002;119(1):122–4. Coussens LM, Raymond WW, Bergers G, Laig-Webster M, Behrendtsen O, Werb Z, et al. Inflammatory mast cells up-regulate
412
A. Birgersdotter et al. / Seminars in Cancer Biology 15 (2005) 405–412
angiogenesis during squamous epithelial carcinogenesis. Genes Dev 1999;13(11):1382–97. [48] Kinzler KW, Vogelstein B. Landscaping the cancer terrain. Science 1998;280:1036–7. [49] Takaku K, Oshima M, Miyoshi H, Matsui M, Seldin MF, Taketo MM, et al. Intestinal tumorigenesis in compound mutant mice of both Dpc4 (Smad4) and Apc genes. Cell 1998;92(5):645–56.
[50] Chatterjee M, Honemann D, Lentzsch S, Bommert K, Sers C, Herrmann P, et al. In the presence of bone marrow stromal cells human multiple myeloma cells become independent of the IL6/gp130/STAT3 pathway. Blood 2002;100(9):3311–8. [51] Mueller MM, Fusenig NE. Friends or foes—bipolar effects of the tumour stroma in cancer. Nat Rev Cancer 2004;4(11): 839–49.
