Aneuploidy affects the expression not only of the genes located on the supernumerary chromosomes, but also the expression of multiple other genes across the entire genome (Torres et al. 2007; Stingele et al. 2012; Upender et al. 2004; Sheltzer et al. 2012; Gemoll et al. 2014). This global gene deregulation can originate from at least two sources. First, transcriptional regulators that are located on the aneuploid chromosome regions and therefore present in altered copy numbers can affect the expression of genes on the other chromosomes. In this case, the response to aneuploidy should be largely dependent on the specific extra chromosome. Intriguingly, a uniform transcriptional response was identified in all aneuploidy model cell lines (Sheltzer et al. 2012; Dürrbaum et al. 2014). This rather suggests a second model, where the expression changes throughout the whole genome are driven by aneuploidy per se and the resulting protein imbalance triggers a specific cellular response that feeds back to the transcriptional regulation. The gene expression pattern in aneuploid yeast strains of different origins resembles the transcriptional pattern of a previously described yeast environmental stress response—ESR (Torres et al. 2007; Sheltzer et al. 2012). In particular, the enriched gene ontology terms for differentially expressed genes suggested aneuploidy-driven alterations in RNA processing and energy metabolism. Similarly, aneuploid mammalian cells of different origins show a uniform aneuploid pathway response pattern (Williams et al. 2008; Stingele et al. 2012; Sheltzer et al. 2012; Dürrbaum et al. 2014; Foijer et al. 2013). The aneuploidy response pattern is similar also in cultured amniocytes from trisomic pregnancies (Sheltzer et al. 2012). In detail, gene ontology terms such as endoplasmic reticulum (ER), Golgi apparatus, lysosomes and vacuoles and membrane metabolism were consistently upregulated, whereas DNA and RNA metabolic pathways—e.g., DNA replication, repair, transcription and RNA splicing—were downregulated on the transcriptional level (Dürrbaum et al. 2014). These deregulated pathways were confirmed by proteome analysis of six different human aneuploid cell lines (Stingele et al. 2012).

What triggers these conserved changes? Recently, it was shown that aneuploidy affects molecular pathways linked to the maintenance of protein homeostasis (Torres et al. 2007; Stingele et al. 2012; Tang et al. 2011; Oromendia and Amon 2014). That is, gain of an extra chromosome may disturb cellular proteostasis by flooding the cellular system with proteins and exhausting the resources for their synthesis, thus causing proteotoxic stress (Donnelly and Storchova 2014; Oromendia and Amon 2014). As a consequence, protein degradation pathways might be activated to eliminate overexpressed proteins and accumulated misfolded proteins. Indeed, aneuploidy activates autophagy and upregulates annotations associated with lysosome, vacuoles and membrane metabolism (Dürrbaum et al. 2014; Tang et al. 2011; Stingele et al. 2013). In addition, the upregulation of ER and Golgi might be also attributed to autophagy, as these organelles serve as a membrane source for autophagosomes (Lamb et al. 2013). In concordance with the hypothesis that the transcriptional response reflects the cellular changes upon proteotoxic stress, autophagy inhibition in near- diploid HCT116 cells resulted in transcriptional changes similar to the aneuploid transcriptional response pattern (Dürrbaum et al. 2014) and aneuploid cells are sensitive to drugs triggering proteotoxic stress (Torres et al. 2007; Oromendia and Amon 2014). Taken together, the current data suggest that the aneuploid transcriptional response pattern arises as a direct consequence of proteotoxic stress due to karyotypic imbalance.

Another possibility is that the transcriptional pattern of aneuploid cells reflects their slower proliferation. Impaired proliferation and delayed progress through the G1 and S phases was observed in disomic yeasts as well as in trisomic and tetrasomic mammalian cell lines (Torres et al. 2007; Williams et al. 2008; Stingele et al. 2012). Two budding yeast strains carrying mutations in cell cycle regulatory factors that delay cell cycle progression show similar transcriptional changes as observed in response to aneuploidy (Sheltzer et al. 2012). Further, the differential transcription was changed when disomic and euploid yeast were grown at the same growth rate in the chemostat (Torres et al. 2007). However, slow proliferation cannot fully explain the aneuploidy response pattern, as complex human aneuploid cell lines exhibit the same transcriptional changes, yet their growth rate is similar to diploid cell lines (Dürrbaum et al. 2014). Moreover, similar expression changes were also identified in chromosomally unstable cancer cell lines of the NCI-60 panel that generally proliferate well (Roschke et al. 2008; Sheltzer 2013). This finding suggests that either both, CIN and aneuploidy, trigger a similar transcriptional response or that the aneuploidy response pattern is in fact a cellular response to CIN, as aneuploidy can promote CIN (Sheltzer et al. 2011, our unpublished data). Interestingly, despite the striking similarities among the aneuploid pathway response, there is no significant overlap of the individual deregulated genes between different aneuploidies (Williams et al. 2008; Gemoll et al. 2014; Dürrbaum et al. 2014). This indicates that there are common physiological changes in response to aneuploidy, but the specific molecular determinants are diverse, likely as a consequence of heterogeneous karyotypes (Dürrbaum et al. 2014). Taken together, aneuploidy triggers activation and inhibition of certain pathways and these global changes likely influence all other physiological consequences of aneuploidy.

In general, gene expression in regions with copy number variations as a result of aneuploidy is correspondingly altered in all types of cancer tissues. However, the estimations of the correlation between gene copy numbers and gene expression is strongly influenced by the methods used to integrate the DNA copy numbers with the gene expression data. In brief, there are three main approaches: (1) simple comparison of the fold changes and calculation of the percentages of genes with high/low copy number that is co-directional with mRNA expression changes; (2) gene-by-gene correlation of gene copy number with gene expression; and (3) correlation of the average gene expression with the averaged copy number across a chromosome arm or segment (Fig. 3; Table 1). Using the first approach the mRNA expression derived from trisomic chromosomes or more than twofold amplified regions is increased within a range from 1.14 to 3-fold compared to the normal diploid case (Schoch et al. 2005; Xu et al. 2010). This broad range already indicates that the results of a fold change comparison will vary according to the cutoffs applied. When no cutoffs were applied, 62.5–87 % of the genes on trisomic chromosomes in acute myeloid carcinoma showed increased expression (Schoch et al. 2005). Similar percentages were reported in breast cancer, when twofold amplified regions were considered (Pollack et al. 2002) and in colon cancer, when genes with the highest 20 % of expression were analyzed (Tsafrir et al. 2006). If considering only more than 2.5-fold amplified regions, 44 % of genes are correspondingly expressed in colon cancer (Hyman et al. 2002). In contrast, several studies report exceptions of amplified regions that do not show correlative changes in gene expression (Tsafrir et al. 2006; Habermann et al. 2007; Taylor et al. 2010). For example, the loss of 8p is common in cancer, but a study in prostate cancer did not detect a correlative decrease in gene expression for all genes in the corresponding region (Taylor et al. 2010). A silencing of chromosomal amplifications in colon cancer has been also reported, but the results are based on a very stringent analysis; here, only 3.8 % of genes located in twofold amplified regions showed a twofold change in mRNA expression (Platzer et al. 2002).

In addition to the statistical method applied, sample size might influence the results; thus, studies calculating the gene-by-gene correlation as in (Holland and Cleveland 2012) over a large tumor sample set give more robust results. The correlation of median gene expression and copy number was consistently reported to be between 0.52 and 0.63 in colorectal cancer (Tsafrir et al. 2006; Grade et al. 2007). Analyses suggest that 10–40 % of the mRNA expression of five different types of tumors can be explained by copy number changes (Xu et al. 2010; Chin et al. 2006; Gu et al. 2008). Moreover, plotting the distribution of gene-by-gene expression and copy number correlation results in a normal-shaped curve with the mean shifted towards positive correlations (Xu et al. 2010; Pollack et al. 2002). The weak positive correlation becomes stronger by applying the approach described by Thompson and Compton (2010), i.e. correlating the average gene expression and copy number changes of a chromosome arm or applying a segmentation algorithm. This approach identified high correlation in three meta-studies on preexisting data (Gu et al. 2008; Ortiz-Estevez et al. 2011; Fontanillo et al. 2012). Moreover, average gene expression along a chromosome has been shown to predict chromosomal aneuploidy (Hertzberg et al. 2007).

Taken together, the effect of aneuploidy on the transcriptome is less pronounced in cancer tissues compared to model aneuploid cell lines owing to the population heterogeneity as well as genetic and epigenetic changes that were selected throughout the tumor evolution. Thus, aneuploidy is only one of many factors influencing the gene expression. Despite the heterogeneity of the approaches and interpretations, there is a trend towards a weak but consistent correlation of chromosome copy number changes and abundance of transcripts in cancer, which is prone to noise at the gene level, but becomes robust when only distinct regions of copy number changes are considered. In future, single cell analysis of cancer cells may provide more insights into the transcriptome changes and their correlation with copy number changes within cancer cell populations.

There are only a few studies that have systematically investigated the influence of aneuploidy on the cancer proteome. Myhre et al. (2013) compared DNA copy number, mRNA and protein levels of 52 breast cancer-related proteins in order to determine at which level the protein expression in cancer is regulated. A large fraction of the analyzed proteins (48.07 %) exhibited no correlation to DNA or mRNA expression values, 28 and 26 % showed a correlation of DNA to mRNA and mRNA to protein, respectively, and 15 % showed a correlation on all three levels. However, the sample size is too small to assume a general genome-wide effect of gene copy number on protein expression. In-depth proteomic analysis of two breast cancer cell lines revealed only weak correlations of 0.22 and 0.28 between gene copy number and protein expression changes (Geiger et al. 2010). Comparison of average expression to average copy number of adjacent chromosome regions increased the correlation. These results indicate that the effect of copy number alterations on the proteome level is even smaller than the effect on the transcriptome level. However, it should be noted that protein levels are influenced not only by gene copy numbers, but also by regulation of transcription, translation and by protein stability. In fact, only 40 % of individual protein concentrations can be explained by mRNA abundance in normal eukaryotic cells (Vogel and Marcotte 2012), which may partially explain the rather minor effects of gene copy number on mRNA and protein expression observed in cancer samples. This also suggests that many gene copy number changes are not reflected on protein levels and extensive proteome analysis will be required to fully estimate the effects of copy number changes on cancer proteome.

Finding commonly deregulated genes associated with cancer subtypes, prognosis or response to treatment remains central to cancer research (reviewed in Chibon 2013). Since aneuploidy is mostly associated with poor prognosis (McGranahan et al. 2012), a predictive gene expression signature for aneuploidy might be a promising approach for tumor classification. A gene signature is a group of genes, for which the expression is associated with a characteristic feature of cancer such as prognosis, subtype, instability or aneuploidy (Chibon 2013). In most cases, gene signatures are derived from transcriptome data, where the gene expression is correlated with cancer features and genes with the highest correlation build the signature. A well-known signature is the CIN70 signature derived from the NCI60 cancer cell line panel of cells to characterize cells with high CIN (Carter et al. 2006). The signature consists of 70 genes strongly correlated with the “total functional aneuploidy” as a proxy of overall aneuploidy across the NCI60 cancer cell lines. Therefore, CIN70 correlates with both CIN and aneuploidy without clear distinction. CIN70 successfully predicts the clinical outcome in multiple cancers, as a high net expression of the 70 signature genes predicts poor prognosis in the majority of cancer samples tested (Carter et al. 2006; Birkbak et al. 2011; Muthuswami et al. 2013). Recently, it was suggested that the CIN70 signature does not correlate with numerical heterogeneity in the NCI60 cancer cell line panel, but with proliferation (Sheltzer 2013). Nevertheless, the high predictive potential of CIN70 was confirmed. Moreover, a novel “HET70” gene signature, whose expression correlates with karyotype heterogeneity in the NCI60 panel was also shown to predict poor prognosis (Sheltzer 2013). In addition, more signatures, mostly for very specific cancer subtype classifications were published in recent years, all either based on or enriched for CIN genes (Chung et al. 2013; Ying et al. 2013; Szasz et al. 2013; Al-Ejeh et al. 2014). In uveal melanoma, a direct transcriptional comparison of high aneuploid and low aneuploid tumors, defined as tumors with >20 and <5 % of nondiploid chromosome arms, respectively, identified a 54-gene signature (Ehlers et al. 2008). Significant overlap of this 54-gene signature genes with genes upregulated in breast cancer, multiple myeloma, keratinocytes treated with UVB irradiation, and with cell cycle genes was found, but the prognostic relevance remains to be proven.

All above described gene signatures are linked to proliferative cellular processes such as replication, chromosome segregation (CIN70) and centrosome function, cell cycle regulation, DNA damage repair (54-gene signature), and the high net expression of the signatures are predictive. In contrast, these gene ontology terms are largely downregulated in model aneuploid cell lines. Accordingly, a gene signature (TRI70) associated with aneuploidy in model aneuploid cell lines predicted good prognosis (Sheltzer 2013), whereas aneuploidy in cancer is generally associated with poor prognosis (McGranahan et al. 2012). These results reflect a general reverse transcriptional pathway deregulation observed when comparing the transcriptome of aneuploid model cell lines and aneuploid cancers (Sheltzer 2013 and our unpublished results). We hypothesize that aneuploidy per se acts as a cellular stress factor and impairs essential cellular pathways. This results in a strong selection pressure for enhanced expression of the impaired factors. Thus, cells that gained mutations, gene amplification, additional aneuploidies and other changes that reverse the aneuploidy response may overcome the adverse effects. Along this line, comparison of different colorectal cancer stages revealed a selection for specific chromosomal imbalances throughout tumor progression that was accompanied by correlating transcriptional changes affecting an increasing number of cellular pathways (Habermann et al. 2007). In future, this hypothesis should be tested by determining the effects of the evolutionary adaptation to aneuploidy on transcriptome and proteome in aneuploid model cell lines and aneuploid tumors.