Aneuploidy involves changes in chromosomal copy number compared with normal euploid genotypes. Studies of gene expression in aneuploids in a variety of species have claimed many different types of responses. Studies of individual genes suggest that there are both structural gene dosage effects and compensation in aneuploids, and that subtle trans-acting effects across the genome are quite prevalent. A discussion is presented concerning the normalization procedures for studying gene expression in aneuploids. A careful documentation of the modulations of gene expression in aneuploids should provide insight into the nature of cancerous cells and the basis of aneuploid syndromes.
Aneuploidy refers to the gain or loss of a chromosome. In diploid organisms, the loss of a chromosome is called a monosomic and the gain of a chromosome is called a trisomic [1,2]. In humans, all monosomic conceptions spontaneously abort very early in development and most trisomics will also miscarry . Segmental aneuploids, which involve only part of a chromosome, usually also have detrimental effects [3,4]. Thus from the standpoint of birth defects, aneuploidy is a major contributor. Aneuploidy is also characteristic of cancer cells, although the distinction between causation and correlation is unresolved [5–10]. Therefore from a medical perspective understanding the molecular consequences of aneuploidy is an admirable goal. Achieving this goal, however, seems to be more challenging than one might think given that a wide range of results have been claimed for gene expression in aneuploids. As with phenotype, changes in gene expression are more prevalent in aneuploids than whole genome ploidy changes [11–13]. In the present review, I note the challenges involved and pose some options to overcome them.
Global studies of gene expression in aneuploids have resulted in a variety of claims about the overall patterns of gene expression in different organisms and conditions. Of course, one might expect different patterns for different aneuploids and there may be in fact organismal differences in response to altered global gene expression. However, some of these differences might arise from the fact that the changes in gene expression in aneuploids are subtle and from the types of data treatment employed in each case.
A variety of generalized effects have been claimed for the modulation of expression in aneuploids. For many investigators, the expectation is that there will be a gene dosage effect for the genes on the varied chromosome ([10,14]; Figure 1). At some level this is certainly a reasonable expectation, i.e. that changing the copy number of a gene will produce a linear response in the amount of gene product realized. Certainly with most transgene experiments, the fact that they work suggests that, on a single gene level, this type of response is found. It should be noted, however, that this expectation is so ingrained in the minds of some investigators that the failure to observe this result leads to the dismissal of other types of experimental findings or attempts to normalize the data to obtain such a final result. Indeed, in her historical account of the development of translocations to produce segmental aneuploids in Drosophila, A. Carpenter noted that when these stocks were used in attempts to find the structural genes for enzymes, there was sometimes a failure to find a dosage effect . In addition, other modulations were also found for regions known not to contain the gene of interest, to the extent that the approach lost favour . However, as noted below, this technique did work in a substantial number of cases and in retrospect revealed a large number of trans-acting effects, which were largely ignored at the time.
Diagrammatic representation of the types of effects observed for gene expression in aneuploids
The second claim is that genes on the varied chromosome essentially do not exhibit a gene dosage effect and instead produce a ‘dosage compensation’ of the genes on the varied chromosome arm (; Figure 1). Certainly sex chromosomes of some species exhibit dosage compensation for the majority of genes encoded on the sex chromosome and thus it is possible that a similar mechanism could operate in aneuploids . However, a caveat of such interpretations is whether the experimental procedures are sufficiently robust to distinguish changes in gene expression of the magnitude that are likely to occur in aneuploids. If the experimental method cannot distinguish a small effect, then a conclusion of dosage compensation might be made when none exists.
A third type of claim is that there is a ‘buffering’ mechanism, which provides feedback in order to ameliorate changes so that homoeostasis is maintained for gene expression . These claims have largely come from studies of heterozygous deletions that do not apparently exhibit a strict dosage response.
Finally, a fourth type of claim is that there is a patchwork of dosage effects and dosage compensation for genes on the varied chromosome arm and that there are also positive and negative trans-acting dosage effects across the genome [11,12,19–21]. This type of result is often claimed in studies where individual genes are analysed, via measuring the enzyme activity, protein or mRNA levels, although some global studies have come to similar conclusions [22–27]. The types of trans-acting dosage effects are usually, but not always, within an inverse or direct correlation limit of the aneuploid chromosomal copy number relative to the euploid normal control ([11,12,21]; Figure 1). In other words, when a chromosome arm is varied in one, two or three copies, there can be trans-acting effects that are directly correlated showing reductions in the segmental monosomic to 50% of the normal diploid and increases in the trisomic to 150%. The other major type of effect is an inverse correlation with the varied dosage. In this case, the monosomic would produce trans-acting effects that range up to a 200% increase in monosomics and reductions to 66% in trisomics relative to the normal two-dose euploids [11,12,28]. Trans-acting effects that exceed these limits are also found but are relatively rare.
The reasons for the diversity of effects reported for aneuploid gene expression patterns could be varied. First we need to consider the normalization procedures used. Whereas most global studies to date have used microarray analyses, the same principles would apply to the recently developed RNAseq (RNA sequencing) technique used to quantify gene expression. Many investigators use an arbitrary 2-fold cut-off in establishing gene expression alterations that are of importance. In this case, if the inverse and direct trans-acting effects observed for individual genes operate globally, then such a cut-off method would basically ignore all of these effects. Although one might argue that the magnitude of these effects are small and therefore insignificant, it is important to realize that most aneuploids have rather dramatic effects on the phenotype, if indeed they are not dominant lethal [2–4]. However, these ‘dramatic’ effects are almost never of a magnitude greater than 2-fold and are usually much less. The trans-acting effects are of similar magnitude to gene dosage effects and thus would be expected to produce similarly detrimental consequences. If the magnitude of changes in gene expression found in viable aneuploids is characteristic of dominant lethal cases, then global modulations of less than a 2-fold effect could produce lethal effects.
The observation of dosage compensation in random aneuploids is unlikely to involve any process of natural selection, such as that which might operate on sex chromosomal dosage compensation, because the aneuploids are laboratory constructions and usually detrimental. Rather the detrimental effects most probably reflect an upset in the gene regulatory mechanisms which are manifested in the aneuploid condition [11,28–31]. At least in some cases, dosage compensation on the varied chromosome arm can result from the fact that specific individual regulatory genes, which are varied on the aneuploid chromosome, produce the trans-acting inverse effects on linked gene targets also located on the varied chromosome [19,32,33]. The change in dosage of the target gene in one direction, which would otherwise produce a gene dosage effect and modulation of its expression via a trans-acting regulator, hence produces an inverse effect which will cancel out to produce a total output essentially at the normal euploid level [32,33].
Another caveat with regard to normalization involves the nature of most statistical programs for global gene expression analysis; these correct individual gene expression values using the mean or median expression of all genes as a reference. Because of the variability in hybridization of microarray studies and the narrow range of the potential effects to be documented, the issue of normalization is a difficult one. In order to be able to detect very subtle effects, some type of normalization or validation procedures are required. The problem with the standard protocols is as follows. If, for example, the trans-acting effects for a particular trisomic were skewed toward a greater number of inverse effects than no effect or direct effects, then normalization by correction to the mean expression will raise the absolute gene expression values for the genome wide expression including those on the varied chromosome. Thus if the inverse effects alter the expression of many genes, as evidence from sampling a few genes might suggest  (at least for some aneuploids), then the global effects on the whole genome will be cancelled out to some degree and a predominance of dosage compensation on the varied chromosome is converted into dosage effects for the linked genes. Some authors note that trisomics have reduced expression and intentionally normalize in this manner to ‘correct’ for such effects with the result being that they realize the dosage effect for most genes on the varied chromosome (as their preconceived ideas have imagined would be the case). The fact that such normalization provides them with the presumed and expected results lulls them into rationalizing such normalization as the proper data treatment.
On the other hand, monosomics often exhibit trans-acting direct effects that are quite prevalent, although they also exhibit inverse effects, with increased expression across the genome when individual genes are assayed [11,12]. If, however, the global down-regulation predominants, then normalization to average or median gene expression will cancel, for the most part, the reduced expression across the genome and increase the treated value of the genes on the varied chromosome. This leads to the false conclusion of up-regulation, buffering or compensation when in absolute terms it does not occur to that degree.
Given the potential problems that might occur with normalization as outlined above and at the same time the need for accurate measurement to document subtle effects, how does one address these problems? In performing global measures of gene expression, the ability to determine absolute levels of expression are lost once homogenization of a tissue is performed when isolating protein or RNA for expression studies. Quantitative PCR for selected genes using a standard RNA that itself might be changed, as it is potentially among the globally affected genes, as a validation is subject to the same caveats. One potential means to address this problem would be to use isolation procedures that allow one to determine total protein, total RNA and total DNA from the same sample. Although this might address the issue to some degree, if a large number of minor proteins are varied in the same direction but have little impact on the total protein amounts, then the normalization to average expression calculated from each protein or mRNA will not be addressed. For RNA, basically all mRNAs are minor in a total RNA isolation and the problem will remain. One approach that is potentially valid is to compare the levels of specific mRNAs with rRNA, which seems refractory to modulation of total levels per cell [34–36], and then in turn to compare this with total DNA. Another potential way to address the problem would be to normalize global expression in microarray experiments to the mean of a collection of genes that have been shown by independent measurements not to vary relative to the DNA. Alternatively, validation of genes could be accomplished by performing quantitative RNA in situ experiments using the cells from which the measured sample was taken. The quantity of RNA per cell in the aneuploid and control samples, ideally measured under identical conditions, could be compared with the treated data for the respective gene from the microarray analysis. If the two types of experiment produce the same result, one can have more confidence in the normalization procedures performed on the global analysis. It is important in such experiments to sample all classes of modulation including increases, decreases and cases of presumed dosage compensation. If enhancer trap collections are available for genes included in the global analysis or if a phenotypic assay for an affected gene is possible, these approaches could be used to validate the response of specific genes in aneuploidy assayed via global analyses. Finally, comparative genome hybridization of aneuploid to euploid values could be used as a normalization standard from the same source as the tested mRNA population.
IS THERE A GENERALIZED ANEUPLOID CELLULAR RESPONSE?
Torres et al.  examined mRNA and protein expression in an extensive set of disomic chromosome lines of otherwise haploid yeast. Their interpretation of the data was that the varied chromosome produced a gene dosage effect at the RNA level for most of the genes on the aneuploid chromosome. Some minor trans-acting effects were reported. Certain stress proteins were induced in many of the aneuploid genotypes leading the authors to conclude that a generalized aneuploid cellular response existed.
The apparent dearth of trans-acting effects in yeast  is distinct from the aneuploid reaction in plants and Drosophila [11,37,38], although there are similarities to plants in that biomass is inversely related to the amount of the genome that is aneuploid and trisomics in diploids show less severe effects than disomics in haploids . It is possible that yeast has a unique response to aneuploidy, although it is not yet clear from this study whether the normalization issues raised above can be ruled out. Clearly, in maize  and barley  examination of protein profiles of aneuploids, both trisomics and monosomics, indicate that there is a radical upset of the relative protein levels for many regions of the genome. Sampling of mRNA levels for specific genes provided data that were consistent with the global protein studies demonstrating many changes in gene expression in aneuploids .
In Drosophila, there are clearly numerous trans-acting aneuploid modulations that can be visualized phenotypically and therefore bypass problems of normalization. Sabl and Birchler  scanned the autosomal regions for aneuploidy that modulated various leaky alleles of the w (white) eye colour locus located on the X chromosome. Numerous positive and negative aneuploid effects within the direct and inverse limits were observed. The w eye colour gene is an excellent reporter for subtle modulations. A trans-acting inverse effect for alcohol dehydrogenase was also phenotypically visualized using a reporter construct .
Examination of the older literature from Datura and Drosophila on screening for the location of structural genes using segmental trisomics shows that the inverse effect is quite prevalently the major trans-acting effect [39–48] although these effects were almost universally ignored. The maize findings using corresponding monosomics and trisomics, in which the monosomics showed increases for the gene products reduced in trisomics, provided evidence that this response was a reaction to aneuploidy rather than a secondary artifact of the aneuploid state [11,12,19]. These observations indicate that many genes on the varied chromosomes exhibit dosage compensation and that trans-acting dosage effects are quite common in multicellular organisms. Whether aneuploidy in multicellular organisms triggers a generalized stress response has not been addressed and would be an interesting question for investigation.
HOW DOES ONE DEFINE DOSAGE EFFECT OR DOSAGE COMPENSATION?
Another issue that is obscure in the literature on gene expression in aneuploids is a consensus on what is defined as a gene dosage effect and what is considered dosage compensation. One might suggest a criterion that a ratio between an aneuploid and its euploid control that is significantly greater than 1.0 be considered a dosage effect and that any ratio that is significantly less in a trisomic than 1.5 be considered dosage compensation. However, such criteria could lead to the nonsensical conclusion that a trisomic ratio of 1.1 is a dosage effect and a ratio of 1.4 is dosage compensation. We propose that the most appropriate terminology would be would be to refer to cases in which the observed value does not differ significantly from the predicted value for a dosage effect be called such, and when the observed value is not significantly different from the predicted value for dosage compensation, that these cases be termed dosage compensation. Cases that are significantly different from both the predicted value for the gene dosage effect and for gene dosage compensation should be classed as intermediate between a dosage effect and compensation.
TRANSCRIPTION FACTOR DOSAGE EFFECTS: WHY THE LACK OF TRANS-ACTING EFFECTS IN ANEUPLOIDS IS UNLIKELY
The finding that heterozygotes for mutations in transcription factors are usually associated with haplo-insufficiency phenotypes in humans [49–52] and Drosophila [53–57] would indicate that it is very likely that aneuploidy in these species would exhibit extensive trans-acting effects. Whereas it is true that some transcription factors present on the varied chromosome would be dosage-compensated and would therefore not have an impact on their target loci throughout the genome, those transcription factors that do exhibit any magnitude of change would alter the expression of their targets in the genome. For aneuploids not to produce a significant number of trans-acting effects from changes in the dosage of transcription factors would require that transcription factors respond differently (i.e. with no response) in the context of aneuploidy compared with changing the gene copy number alone, e.g. in heterozygous-null mutations or when adding copies via transgenes. At present, there is no evidence to suggest that this would be the case.
The dosage effect of transcription factors has been attributed to the effect of an altered stoichiometry in multisubunit complexes, which then affect the kinetics and mode of assembly and thus the function of the whole [50,51,58–60]. Classical genetics has shown that aneuploidy produces much more drastic effects on the phenotype than changes in whole ploidy [1,2]. Molecular studies on gene expression of individual genes also show a similar molecular pattern, with more effects produced from aneuploidy than ploidy changes . These observations, among other evidence [29–31,60], support the idea of the stoichiometric effects of transcriptional machineries.
This principle is further supported by evolutionary evidence from whole genome duplications compared with CNVs (copy number variants). Most organisms have experienced repeated polyploidization events followed by deletion of most genes back to the two copy diploid level over evolutionary time, especially in plants and protozoa. Genes involved in multisubunit molecular complexes, which include regulatory genes, are preferentially retained for longer periods [61–63]. In contrast, genes involved with molecular complexes are preferentially depleted from segmental duplications [64,65]. These results are consistent with a requirement for a stoichiometric relationship of regulatory factors and suggest that varying them individually, either by deletion from whole genome duplication or by duplication in a CNV, produces a dosage effect that is detrimental. These results support the idea that the subtle global modulations caused by altered transcription factor stoichiometries have significant biological relevance.
As the number of global studies of gene expression in aneuploids or cancer cells increases, there is great potential to gain significant insight into the causative effects that these conditions confer on to organisms and cells. Databases of the aneuploid changes in cancer cells will provide knowledge of the genomic arrangements, as well as the mutational differences that do or do not foster cancerous growth. In parallel, it is important to document the gene expression changes that operate in such cancerous cells relative to the regulation properties of normal cells. Such studies must realize that potentially small changes in gene expression could have significant effects on cell and organismal phenotypes and set-up the necessary experimental conditions to document them correctly.
Research on this topic was supported by the National Science Foundation (Division of Biological Infrastructure) [grant number 0733857]; and by the National Institutes of Health [grant number RO1GM068042].