Perspective

Recent research has changed the way we envisage the control of gene expression. Instead of thinking in sequential modular control elements such as enhancers, promoters, 5′ and 3′ UTR elements assembled in linear arrays; the full genomic landscape, with both coding and noncoding transcription needs consideration. Not just in the way we design our experiments, but also how we interpret them [Fig. 1 and reviewed: (Grzechnik et al. 2014)]. We recently published research demonstrating a propensity for aberrant 3′ UTR dynamics associated with heterologous 3′ UTRs in budding yeast (Swaminathan and Beilharz 2015). Specifically, we showed that the CYC1 3′ UTR, often used in ectopic expression plasmids could drive abundant convergent antisense transcription. Moreover, genomic integration of the common epitope tags TAP and GFP (Ghaemmaghami et al. 2003; Huh et al. 2003) resulted in locus-specific and transcription state-dependent truncating alternative polyadenylation and cryptic antisense transcription. In complementary research, modification of 3′ UTRs by introduction of MS2 stem-loops (used in localisation studies with fluorescently tagged MS2 coat proteins), was shown to lead to changes in mRNA stability and an accumulation of spurious 3′-truncation products (Garcia and Parker 2015). Add to this the common (but rarely reported) expression perturbation to neighbouring genes induced by integration of reporter and disruption cassettes (Ben-Shitrit et al. 2012; Pena-Castillo and Hughes 2007; T. Beilharz, unpublished); these studies suggest that cellular machineries do not always interpret our ‘on paper’ designs for recombinant expression as intended.

Fig. 1
figure 1

Transcription in the context of gene-loops. In addition to the synthesis of mRNA, transcriptional activity within a single genetic locus can include multiple RNA isoforms derived from noncoding transcription. In this schematic, transcription from the dominant promoter driving mRNA synthesis is indicated in black (1). However, antisense transcripts in red, often proceed from cryptic promoters within 3′ UTRs (2) including those used in heterologous expression cassettes (Swaminathan and Beilharz 2015). These run antisense to bona fide 3′ UTRs in what is termed convergent antisense transcription. Their transcription is linked to the transcriptional state of promoter (1). Other cryptic promoters generate stable unannotated transcripts (SUTs) often emanating antisense from within coding RNA (3). And many promoters are bidirectional (4) resulting in cryptic unstable transcripts (CUTs) that are rapidly destabilised by nuclear surveillance machinery under normal conditions (Neil et al. 2009; Xu et al. 2009). But, do all these forms exist in the same cells at the same time? And if not, what controls which transcripts are expressed, and when?

Full size image

A further non-technical complication to interpretation of gene expression is that we typically study cells in populations. In extreme cases these are communities of functionally diversified cells such as those in stationary phase cultures or in mature yeast colonies (Aragon et al. 2008; Cap et al. 2012; Traven et al. 2012). Even standard steady-state cultures represent an aggregate of cells in different stages of the cell and metabolic cycles. Thus, an apparent low level of certain RNA in mixed populations can correspond to very high levels in just a few cells. Nowhere is this more confusing than in the expression of noncoding RNA. It is now clear that many yeast promoters are capable of bidirectional transcription, and that directionality of such promoters is controlled (Fig. 1, Tan-Wong et al. 2012). In the case of convergent antisense transcription, simultaneous forward (marked No. 1 in Fig. 1) and reverse transcription (Nos. 2 and 3) results in polymerase collision (Prescott and Proudfoot 2002). Yet evidence that sense-antisense pairs co-exit at least transiently comes from RNAi reconstitution experiments that depend on double-stranded RNA duplexes (Alcid and Tsukiyama 2014). The open question is whether transcription from within a local genomic landscape is fixed in mutually exclusive states between cells in the population, or stochastically toggles between transcriptional states within individual cells. Experiments from the Fink lab suggest that the coding/noncoding circuitry around the FLO11 locus results in fixed, but variegated expression between cells (Bumgarner et al. 2009). During the cell cycle on the other hand, coding/noncoding circuits seem to switch between states in the same cells (Granovskaia et al. 2010). Understanding how these circuits are established and controlled will be a major challenge for the future. We suggest that the dominant regulatory information stems from the promoter of the coding transcript because its transcriptional state can rewire 3′-end dynamics associated with heterologous 3′ UTRs (Swaminathan and Beilharz 2015). However, this will require much additional research before a consensus mechanism can be reached.