Variation in gene copy number is seen among species, individuals and cell types and makes an important contribution to cell and organismal fitness. Three studies have shed some light on the complicated relationship between gene dosage and protein expression levels, function and disease.

Springer and colleagues asked a simple question: if you knock out one copy of a gene from a diploid genome, is the amount of expressed protein also halved? The authors tried this experiment on each of 730 yeast genes (more specifically, GFP-fusion genes) and found that the answer is 'yes' for 80% of genes across five environmental conditions. This result is surprising, as positive or negative feedback would be expected to either exacerbate or buffer the reduction in gene dose, respectively. As the fitness of the hemizygous strains is not lower than the fitness of wild-type yeast, these proteins must normally be expressed at about twice the required level. However, in <5% of genes the level of protein expression in the hemizygote was close to the wild-type level, so sensitivity to gene dosage varies across genes. This issue was also explored in two other papers.

Vertebrates have undergone two rounds of whole-genome duplication (WGD), and remnants of these events are visible in the copy number of some genes. What determines which ancient duplicates (ohnologues) survive? Makino and McLysaght tested the hypothesis that ohnologues involve genes that are more sensitive to dosage. They show that human ohnologues are half as likely as non-ohnologues to experience copy-number variation or undergo small-scale duplications (SSDs). By contrast, genes that have experienced SSDs are more likely to show copy-number variation. Dosage-balanced ohnologues are also strongly associated with disease: for example, three-quarters of the 16 candidate genes for trisomy 21 are dosage-balanced ohnologues, which suggests that the evolutionary history of genes could be used to identify candidate genes for disorders of dosage imbalance.

Following a different line of enquiry, Gout and colleagues sought to connect ohnologue retention with gene-expression level. Microarray experiments in Paramecium tetraurelia — across 40,000 genes under 58 conditions — revealed that the level of expression of a gene before WGD is positively correlated with its retention in duplicate after WGD. A mathematical model — which also incorporates data from yeast and animals — provides a general explanation: high gene expression bears a cost that puts gene function and expression under constraint, which increases the selective pressure against changes in terms of the encoded protein sequence, its expression level and its gene dosage. As a consequence, more highly expressed genes are over-represented among retained genes because highly expressed genes are generally more constrained.

These papers highlight the difficulties in extracting meaningful information from complex patterns of gene-copy evolution, but several widely applicable experimental, bioinformatic and mathematical tools are now available and are providing testable hypotheses.