Genome-Wide Gene Expression Analysis in Cancer Cells Reveals 3D Growth to Affect ECM and Processes Associated with Cell Adhesion but Not DNA Repair

Cell morphology determines cell behavior, signal transduction, protein-protein interaction, and responsiveness to external stimuli. In cancer, these functions profoundly contribute to resistance mechanisms to radio- and chemotherapy. With regard to this aspect, this study compared the genome wide gene expression in exponentially growing cell lines from different tumor entities, lung carcinoma and squamous cell carcinoma, under more physiological three-dimensional (3D) versus monolayer cell culture conditions. Whole genome cDNA microarray analysis was accomplished using the Affymetrix HG U133 Plus 2.0 gene chip. Significance analysis of microarray (SAM) and t-test analysis revealed significant changes in gene expression profiles of 3D relative to 2D cell culture conditions. These changes affected the extracellular matrix and were mainly associated with biological processes like tissue development, cell adhesion, immune system and defense response in contrast to terms related to DNA repair, which lacked significant alterations. Selected genes were verified by semi-quantitative RT-PCR and Western blotting. Additionally, we show that 3D growth mediates a significant increase in tumor cell radio- and chemoresistance relative to 2D. Our findings show significant gene expression differences between 3D and 2D cell culture systems and indicate that cellular responsiveness to external stress such as ionizing radiation and chemotherapeutics is essentially influenced by differential expression of genes involved in the regulation of integrin signaling, cell shape and cell-cell contact.


Introduction
The microenvironment is a fundamental regulator of cell behavior [1,2,3]. A large body of work has shown how interactions of cells with the extracellular matrix (ECM), as one of the key components of the microenvironment, contribute to the regulation of critical cell functions such as cell shape/architecture, survival, proliferation, and differentiation [2,4,5,6,7,8,9]. Cell-ECM interactions also control gene expression and chromatin organization in a growth and ECM-dependent manner [10,11]. Recent emerging findings show that especially the growth conditions play an essential role for the cellular responsiveness to external stimuli. This became evident in three-dimensional (3D) ex vivo cell cultures grown in ECM and in spheroid models [4,11,12,13,14,15,16,17]. Importantly, these 3D cell culture models better mimic a physiological microenvironment than conventional uncoated or ECM-precoated cell culture plastic [15,18].
In addition to the use of 3D cell culture models in tissue engineering [19,20,21] and studies on embryonic development and physiology [18], 3D cell cultures are increasingly employed in cancer research [7,8,9,12,16,22]. In the vast majority of cases, tumor cell lines of different origin show an enhanced resistance to radio-and chemotherapy in a 3D environment indicative by increased clonogenicity and decreased apoptosis [12,13,14,16,17,23,24,25,26,27]. Apart from a significant impact of integrinmediated cell-ECM interactions [28], a complex interplay of biochemical signaling pathways and biophysical/mechanotransduction-related factors is thought to confer this enhanced tumor cell resistance whose underlying mechanisms remain to be determined both on the gene and on the protein level [2].
With regard to gene expression, great efforts have been undertaken to identify specific diagnostic, prognostic and therapy-monitoring gene expression patterns in biopsies of various human malignancies [2,5,6,29,30,31]. Intriguingly, some of these studies demonstrated strong overlap between in vivo and 3D but not 2D cell culture data sets, which finally enabled the identification of gene signatures predictive for overall survival of cancer patients. It remains to be clarified whether changes in gene expression under 3D versus 2D growth conditions can explain or provide hints at certain stress or DNA repair pathways involved in the enhanced radio-or chemoresistance of 3D grown tumor cells. If so, targeted therapeutic approaches against key tumor promoters could be optimized.
To address this question, this study compared basal gene expression of two human cancer cell lines of different origin and with varying genetic background in a 3D ECM scaffold or under conventional 2D monolayer conditions with respect to their behavior upon radiation and chemotherapy.

Materials and Methods
Cell lines, culture conditions, and cell doubling times Lung tumor cell line A549 was obtained from ATCC (Manassas, USA). The squamous cell carcinoma cell line UT-SCC15 was a kind gift from R. Grenman (Turku University Central Hospital, Finland). For conventional 2D cell culture, cells were cultured in Dulbecco's Modified Eagle Medium (DMEM; PAA, Cölbe, Germany) containing glutamax-I (L-alanyl-L-glutamine) supplemented with 10% fetal calf serum (FCS; PAA) and 1% non-essential amino acids (NEAA; PAA) at 37uC in a humidified atmosphere containing 7% CO 2 . For 3D cell culture, cells were plated into a mixture of 0.5 mg/ml laminin-rich extracellular matrix (Matrigel; BD, Heidelberg, Germany) and complete DMEM medium upon a layer of agarose (Sigma, Taufkirchen, Germany) in a 24-well cell culture dish (BD) as published [12,13,14,16].
Doubling time of cells growing as monolayers in cell culture flasks were counted according to standard protocols. Briefly, single cells were plated in 2D or 3D and trypsinized and transferred to 10 ml of medium containing FCS. After gently mixing in medium, 10 ml of the cell solution was pipetted onto a Neubauer counting chamber using an appropriate dilution. Cells in 4 squares were counted microscopically (Zeiss, Jena, Germany) and the number of cells was calculated as previously described [32]. For cells growing in a 3D Matrigel environment, cells were isolated as previously described [23,33] using a Neubauer counting chamber.

Irradiation of cell cultures
Irradiation was delivered at room temperature using single doses (0-6 Gy) of 200 kV X-rays (Yxlon Y.TU 320; Yxlon, Copenhagen, Denmark) filtered with 0.5 mm copper. The doserate was approximately 1.3 Gy/min at 20 mA. The absorbed dose was evaluated using a Duplex dosimeter (PTW, Freiburg, Germany).

cDNA microarray
Gene expression profile was measured with RNA extracted from 5610 6 cells grown in 3D or 2D. Total RNA was prepared using NucleoSpin RNA II Kit according to the manufacturer's instructions (Macherey-Nagel, Düren, Germany). Procedures for cDNA synthesis, labeling and hybridization were carried out according to the manufacturer's protocol (Affymetrix) and as published [35]. All experiments were performed using Affymetrix human genome gene chip HG U133 Plus 2.0. First strand cDNA synthesis with 90 ng of total RNA, synthesis of biotin-labeled cRNA and clean up was carried out using the 39 IVT Express Kit Three replicates were used for microarray analysis and statistics. For SAM analysis, a Signal Log Ratio of 0.8 (20.8) was used as a threshold which equals a fold change of 1.6 and 21.6, respectively. Gene expression data are available at GEO Accession No. GSE17347. For gene list, functional enrichment we used the ToppGeneSuite [36]. Pathways and other functional groupings of genes were evaluated for differential regulation using the visualization tool GenMAPP (Ver. 2.1) as described previously [37].
SAM requires the input of a ''delta'' value. The ''delta'' value defines the threshold of the number of false positive genes in the validated dataset. To identify a list of potentially significant genes, we calculated the false discovery rate (FDR). The estimated FDR (the median number of falsely significant genes) for each given ''delta'' (''delta'' value) was determined according to Tusher et al. [38].

Semi-quantitative RT-PCR
For validating microarray data by PCR, total RNA isolated for gene expression profiling was used. cDNA was prepared with SuperScript TM III Reverse Transcriptase kit according to the instructions of the manufacturer (Invitrogen, Karslruhe, Germany). Briefly, cDNA was synthesized in a 20 ml volume containing 1 mg of DNase-treated total RNA, 1 ml of oligo(dt) 20 (50 mM), and 1 ml 10 mM dNTP Mix, 4 ml of 56 First Strand buffer, 1 ml of 0.1 M DTT, 1 ml of RNase OUT, and 1 ml of SuperScript III (200 U/ml). RNA, dNTPs and oligo(dt) primer were mixed first, heated to 65uC for 5 min, and placed on ice until addition of the remaining reaction components. Then, the reaction mixture was incubated at 55uC for 45 minutes and terminated by heat-inactivation at 70uC for 15 minutes. An identical reaction without the reverse transcriptase was performed to verify the absence of genomic DNA. Semi-quantitative RT-PCR was performed for the genes TXNIP, DUSP6, CEACAM1, NPC1 and BCL2A1 using primers listed in Table 1 (Eurofins MWG Operon, Ebersberg, Germany), 2 ml of cDNA and HotStar Taq polymerase (Qiagen, Hilden, Germany) according to standard PCR protocols. Forward primers for all genes were chosen by searching the original oligonucleotide sequence of the corresponding gene identification number of the Affymetrix gene chip on the Affymetrix web site (http://www.affymetrix.com/ analysis/index.affx). The reverse primers were designed based on the gene sequence provided on the Affymetrix web page. Annealing temperatures of 50uC were used for TXNIP, DUSP6 and BCL2A1 or 55uC for CEACAM1 and NPC1. For normalization, a b-actin fragment of 540 bp was amplified concurrently using primers listed in Table 1 [39]. The results of three independent experiments were analyzed using 1% agarose (Carl Roth, Karlsruhe, Germany) gels and densitometric analysis with Image JH software (National Institutes of Health, Bethesda, USA) after staining gels with ethidium bromide (Carl Roth).

Results
3D and 2D cell growth characteristics and cell survival after radio-or chemotherapy Cell shape and morphology were different between 3D and 2D cell growth conditions (Fig. 1A). Four days after plating, cells were exponentially growing with similar doubling times (Fig. 1B). This similarity in cell doubling times provided comparable experimental conditions for the whole genome gene expression analysis. Importantly, previous experiments evaluating the radiation survival of A549 and UT-SCC15 cells were repeated thus forming the rationale for the presented study to link growth-dependent radio-and chemoresistance with growth-dependent gene expression modification [11]. Under 3D conditions, both cell lines showed significant higher clonogenic survival upon exposure to increasing single doses of X-rays or increasing concentrations of Cisplatin as compared to 2D (Fig. 1C and D). These data suggest a connection between growth conditions and cell morphology and cellular radio-and chemoresistance.
Whole genome gene expression analysis of A549 and UT-SCC15 cultured in 3D or 2D This genome wide gene expression analysis was performed under untreated conditions and intended to identify specific gene expression patterns of single genes or functionally associated gene groups that can be linked to tumor cell radio-and chemoresistance. From gene profiling, hierarchical clustering and statistical analysis using SAM, it was obvious that growth conditions affect gene expression patterns ( Fig. 2A and B). The estimated false discovery rate (FDR) was 0 down to the set ''delta'' (''delta'' = 2.873 in A549; ''delta'' = 2.445 in UTSCC15). A higher number of genes were differentially expressed in A549 (376 transcripts) than in UT-SCC15 (178 transcripts) cells ( Fig. 2A, Table 2; Table S1 and S2). In 3D, A549 and UT-SCC15 cells showed more genes up-regulated than down-regulated as compared to 2D ( Fig. 2A and B).
According to SAM, A549 cells had 242 transcripts up-and 134 down-regulated while UT-SCC15 cells had 125 transcripts upand 53 transcripts down-regulated ( Fig. 2A and B, Table S1 and S2). Intriguingly, these two different cell lines showed an overlap of cellular components and biological functions to be affected under 3D growth conditions with regard to gene ontology (Fig. 2C, Tables S3 and S4). UT-SCC15 cells, in general, demonstrated less genes modified by 3D growth as compared to A549 cells ( Fig. 2A, Table S2). From a one-to-one gene expression comparison, it became obvious that the modifications occur in different genes within the same groups of cellular components or biological functions in a cell line-dependent manner. On top of SAM, t-test analysis was performed in which 856 genes were up-regulated and 721 down-regulated in 3D A549 cell culture relative to 2D (to note: SLR threshold +/2 0.8). In UT-SCC15 cells, 429 genes were up-regulated and 277 down-regulated in 3D as compared to 2D. The overlap of differentially expressed genes upon SAM and t-test for the 3D-to-2D comparison revealed 238 transcripts upregulated and 128 transcripts down-regulated in A549 cells and 119 transcripts up-regulated and 52 transcripts down-regulated in UT-SCC15 cells (Table S5 and S6).These findings indicate that processes associated with cell adhesion, biological adhesion, immune system responses, defense responses, tissue development and response to organic substance are predominantly regulated via differential gene expression when cells are transferred from a 2D to a 3D matrix-based microenvironment. Without pronounced expression changes of genes involved in DNA repair, it can be speculated that radio-and chemoresistance of 3D grown cells results from particular, yet unidentified, changes of cell architecture and hereby induced modifications of the cellular interactome in terms of signal transduction and protein-protein interactions.
The signal log ratios of selected genes from 3D or 2D cell cultures are displayed in Figure 3. TXNIP and DUSP6 present genes increasingly expressed in both cell lines when cultured in 3D relative to 2D (Fig. 3). CEACAM1 induction and NCP1 and BCL2A1 repression were only observed in 3D A549 cell cultures (Fig. 3). Semi-quantitative RT-PCR on RNA samples used for microarray analysis confirmed induced TXNIP expression in 3D while increased DUSP6 expression could only be confirmed in UT-SCC15 cells ( Fig. 4A and B). Enhanced CEACAM1 mRNA levels were detectable in both 3D A549 and 3D UT-SCC15 cell cultures, which was confirmatory for A549 and borderline for UT-SCC15 cells with regard to DNA microarray data ( Fig. 4A and B). The genes NPC1 and BCL2A1 showed reduced or stable levels in the genome wide analysis in A549 or UT-SCC15 cells, respectively; results fully confirmed by semi-quantitative RT-PCR ( Fig. 4A and B).

Discussion
Loss of functional and phenotypic characteristics occur when cells are cultured ex vivo in a 2D microenvironment [1,4,18,20]. This can be prevented by culturing cells in physiological 3D ECM scaffolds shown for various endpoints such as protein expression, morphology, and prediction of gene signatures for clinical outcome [2,5,6,29,30]. Considering these points, we focused on the mechanisms modulating tumor cell sensitivity to radio-and chemotherapy. To assess the impact of basal gene expression profiles on the more physiological 3D versus the artificial 2D cell culture models, this study was performed on human epithelial cancer cell lines originating from two of the most frequent cancers worldwide, i.e. lung (A549) and head and neck squamous cell carcinoma (UT-SCC15) [40]. Selected from various cell line models tested in our laboratory, these two cancer cell lines vary in their origin and genetic background and are prime examples showing the prosurvival effects of growth in a 3D ECM scaffold in comparison to conventionally used culture plastic [11,12,13,14]. We show significant changes in gene expression profiles of 3D versus 2D cell cultures. While genes involved in DNA repair pathways stayed unmodified, the majority of altered genes in both cell lines were associated with biological functions like tissue 3D cell cultures are increasingly employed in cancer research as well as tissue engineering, developmental and cell biology. For therapy development, cell responsiveness is the key issue and dramatically different in a 3D physiological environment as compared to Petri dish conditions. With regard to clonogenic cell survival after exposure to X-rays or the widely applied chemotherapeutic drug Cisplatin, both A549 and UT-SCC15 cells show increased survival rates in 3D relative to 2D. As these are only two examples out of several [12,13,14,15,23,24], the paradigms ''cell adhesion mediated radioresistance'' and ''cell adhesion mediated drug resistance'' mainly examined in conventional ECM-coated cell culture dishes have to be re-evaluated by taking severe modifications in e.g. signal transduction, DNA repair processes, etc into account when cells grow in 3D. Importantly, differences in parameters such as hypoxia, proliferation and radiation dosimetry that are well known as essential determinants of cellular radiosensitivity could be excluded in a previous in-depth comparative analysis between 3D and 2D cell culture conditions [11]. Thus, the 3D ECM-based cell culture model used here is a reasonable and feasible method for investigations of tumor cell radio-and chemosensitivity and the underlying mechanisms.
Despite our knowledge about severe alterations of gene expression patterns by separating cells from tissues for ex vivo 2D cell cultures [1,41,42], these findings have been widely neglected. In 2D, both the number of differentially expressed genes and the gene expression levels in cell lines from tumor and normal tissues were diminished by up to 70% and had a fluctuation range of about 1.5-fold as compared to the originating tissues, respectively [43,44]. Comparing our microarray analysis with the work of others, we found similar gene expression profiles. For example, both A549 and UT-SCC15 showed distinctive basal patterns of genes encoding for proteins with functions in the modulation of the ECM like Laminin, Fibronectin, Collagen, a finding similarly observed in a basal gene expression analysis of 60 cancer cell lines and their corresponding cancer tissues and in a study done on melanoma [30,42]. These data further support our notion that 3D cell cultures are more physiological and tissue-like than 2D. Despite similar growth rates of our tested human tumor cell lines, cellular differentiation processes associated with modifications in e.g. ECM proteins could pronouncedly influence the responsiveness of tumor cells to various therapies. Of upmost importance for us was the finding that 3D cell cultures differentially express genes involved in the 'extracellular matrix' and 'cell adhesion' as well as 'defense response' but not in 'DNA repair'. As a modulation of genes involved in the regulation of cell size and adhesion is not astonishing when cells are transferred from monolayer to a 3D environment, these findings strongly suggest that the level of expression of particular genes associated with DNA repair is not necessarily linked functionally to an observed cell behavior. In our case, irradiated or Cisplatintreated A549 and UT-SCC15 showed a higher clonogenic survival under 3D growth conditions as compared to 2D; however, a corresponding increase in the expression of e.g. DNA repair genes such as PRKDC (protein kinase, DNA-activated, catalytic polypeptide), ATM (ataxia telangiectasia mutated) or MDC1 (mediator of DNA-damage checkpoint 1) was absent. Conclusively and in contrast to study assessing the prediction of radiotherapy and chemotherapy responsiveness, it may be speculated that the major determinant of this improved survival is likely to be the protein interactome and not the transcriptome. Concerning transfer of data from bench-to-bedside for therapy, it should be emphasized that data generated in tumor samples with conserved phenotypes, including gene expression profiles, is likely to be more relevant than 2D data and can be used to identify essential signaling hubs for target therapy.
In summary, the presented data clearly demonstrate that growth conditions have profound impact on gene expression. Since 3D cell cultures reflect physiological growth conditions and confer therapy resistance to cancer cells, these findings suggest that, on the transcriptome level, cell shape and cell-cell contact are two of the major determinants of cellular responsiveness to external stress. This notion can be propagated to a more realistic, higher complexity in which structural and communicative changes in the proteome add substantial cues for the regulation of cellular behavior, in particular therapy resistance. Future studies in optimized cell culture models such as the 3D ECM model are warranted to address the transcriptomic and proteomic mechanisms of tumor cell biology, cancer therapy responsiveness and evaluation of novel potential cancer targets.