Keywords

1 Introduction

Dynamic changes in the topology of genomes and transcriptomes are not a newly recognized phenomenon—plasticity in DNA and RNA has long been recognized as a key regulatory point in most biological processes. However, on a 24-h timescale, it is only recently that the extent of changes at a systems level has begun to be appreciated (Reddy and O’Neill 2010). Progress in this arena has largely been propelled by advances in technology, which have allowed interrogation of DNA and RNA over time, at a truly global level (Akhtar et al. 2002; Panda et al. 2002; Rey et al. 2011).

The numbers of genes and transcripts regulated in a circadian fashion are not trivial. Various studies have estimated that approx. 10–15 % of mammalian transcripts undergo circadian oscillation in tissues such as the liver or heart and similar numbers of proteins oscillate over the circadian cycle (Akhtar et al. 2002; Panda et al. 2002; Storch et al. 2002; Ueda et al. 2002). Thus, when cells and tissues are viewed at the genomic, transcriptomic, and proteomic levels, they are not only in a state of flux because of their need to maintain body homeostasis but also because of the background influence of the circadian clock on the production of these macromolecules (Hastings et al. 2003; Reddy et al. 2006). In the extreme case, as occurs in cyanobacteria, the entire genome undergoes rhythmic change, which in turn sculpts the gene expression profile of thousands of genes (Johnson et al. 2008; Vijayan et al. 2009; Woelfle et al. 2007).

In this chapter, I will discuss recent advances in our understanding of the clockwork at the genomic and transcriptomic level, highlighting the importance of considering the circadian clock in analyses of cells and tissues for pharmacological experiments. I will also touch upon the types of high-throughput approaches that have been utilized to probe the clockwork at a systems level, which will have relevance to the pharmacologically minded scientist.

2 Genomic Level Analyses of the Clockwork

The genomic landscape is traditionally regarded as static and only subject to change when cells need to undergo fundamental, and often terminal, changes such as end differentiation (Kouzarides 2007). However, this line of thought has been challenged recently by observations that genome-level changes to DNA occur in disparate organisms, from bacteria to mammals. For example, in the circadian biologist’s favorite cyanobacterium, Synechococcus elongatus, its entire genome undergoes rhythmic supercoiling over a day, directing rhythmic abundance of mRNA (Kucho et al. 2005; Vijayan et al. 2009; Woelfle et al. 2007). In mammals, the changes are not thought to be as far reaching, but have been demonstrated at specific genomic loci where clock-relevant transcription factors, such as CLOCK and BMAL1, bind and modulate chromatin structure in a rhythmic manner (Ripperger and Schibler 2006). This has obvious implications for understanding RNA dynamics but also underscores the need for careful regard to sampling time in any experiment, since it cannot be assumed that the genome is “static” in terminally differentiated tissues or in cells cultured under laboratory conditions.

2.1 ChIPing Away at Chromatin

There are several techniques that can be applied to investigate the state of the genome at a given time, but one of the most versatile is chromatin immunoprecipitation (ChIP). The premise of this technique is straightforward. As transcription factors (or other proteins such as histones) bind to their native targets in the genome, they are first “frozen” in time and space by using a cross-linking reagent (usually formaldehyde). The result is that all transcription factors remain attached to the DNA they were bound to prior to fixing with the formaldehyde, and any unbound protein is cross-linked to other proteins. Cross-linked chromatin thus consists of DNA with transcription factors bound to specific regions. The cross-linked can subsequently be reversed by overnight incubation at moderate temperatures (65 °C typically), and pure DNA is extracted from this subsequently (Farnham 2009).

This process can be performed on samples from different times in the circadian cycle such that a temporal map of transcription factor binding can be obtained. If a particular genomic locus is of interest, such as a promoter/enhancer region of a gene, the polymerase chain reaction (PCR) can be used to amplify the relevant target region, and its enrichment at different time points can be compared quantitatively using real-time PCR (qPCR). If, however, the target regions in the genome are unknown, or you wish to take an unbiased approach to finding transcription factor target sites, then other genome-level methods have to be used in combination with ChIP.

2.2 ChIP-chip and ChIP-seq Allow Whole-Genome Analyses of Transcription Factor Binding

Having the ability to temporally map binding sites of transcription factors (or other proteins that bind to DNA) across the genomic landscape has only recently become possible with the advent of new technologies. The first breakthrough technology was the DNA microarray. When using ChIP coupled with microarrays (also known as a DNA “chip”), the technique is termed “ChIP-chip.” An important aspect is to ensure adequate genomic coverage. This is, however, difficult given the limited capacity and packing density of probe DNA sequences on the surface of microarrays and the necessity to synthesize a plethora of probes to cover the entire genome (Buck and Lieb 2004; Scacheri et al. 2006; Wu et al. 2006). To get useful coverage, multiple microarrays have to be used, which is both expensive and experimentally time-consuming. Initially, promising studies were confined to the detailed analysis of individual chromosomes, but low-resolution studies at the “whole genome” level have been performed (Bieda et al. 2006; Horak et al. 2002).

Microarrays have now been effectively usurped, as will no doubt be the case soon for transcriptomics studies, by high-throughput sequencing approaches. This technology has revolutionized genomic-scale analyses. Instead of having a defined set of probes that cover the entire genome, it is now possible to instead simply sequence all of the ChIPed DNA sequences that represent binding regions across the genome (termed “ChIP-seq”). After some relatively demanding bioinformatics to align each sequence to the reference genome (e.g., mouse or human), each sequence, and the relative number of sequences at a particular genomic locus, can be determined at a very high resolution (Farnham 2009). At binding sites, enrichment of overlapping sequences yields “peaks” where the target protein was bound (Fig. 1). Further bioinformatics can subsequently determine, with ever-increasing precision, binding motifs within the genomic DNA (Park 2009; Pepke et al. 2009).

Fig. 1
figure 00161

Protein–chromatin interactions are first cross-linked in situ using, typically, formaldehyde. Specific DNA fragments are co-immunoprecipitated and sequenced to identify genome-wide sites associated with a factor or modification of interest (Adapted from Illumina Web site, http://www.illumina.com/technology/chip_seq_assay.ilmn)

Full size image

Recently, some investigators have applied ChIP-seq methodology to investigate the function of the core transcription–translation feedback oscillator at a genomic scale. For example, by using an antiserum directed against BMAL1, Rey and colleagues have mapped BMAL1 binding sites across the genome of mouse liver, over the circadian cycle (Rey et al. 2011). This convincingly showed that BMAL1 rhythmically binds to over 2,000 target sites in the genome, with peak occupancy occurring in the middle of the circadian day. Functionally, these targets were diverse, but their results pointed towards carbohydrate and lipid metabolism loci as major targets for BMAL1’s action in vivo. Furthermore, using a combination of bioinformatics and modeling, these authors were able to show that E-box motifs were strongly correlated with the presence of BMAL1 binding sites and to rhythmic transcription of these loci (Rey et al. 2011). Recent comprehensive studies have further developed these principles and determined a circadian “chromatin landscape” by assaying other components of the “clock complex” (Koike et al. 2012). The true power, however, of genome-level approaches comes from the ability to relate changes at the level of DNA to transcripts and eventually to proteins, the effectors of cellular physiology.

3 Circadian Transcriptomics

Early studies in mammals implicated a “core” set of clock genes in the molecular clockwork (Buhr and Takahashi 2013; Hastings et al. 2003; Reddy and O’Neill 2009). Amongst these were the Period and Cryptochrome gene families, as well as the transcription factors driving their expression, Clock and Bmal1. It became apparent quickly that the majority of these genes were expressed rhythmically and that there were close parallels between the mammalian clockwork and what had been extensively investigated in Drosophila previously (Hastings et al. 2003). Initially, however, it was thought that relatively few genes (and their respective transcripts) were under circadian clock control. However, with the advent of microarray technology, this hypothesis became eminently testable.

The first study to map the circadian transcriptome was performed in plants by Harmer and colleagues (Harmer et al. 2000). This demonstrated the pervasive nature of rhythmic transcription and the clear anticipatory advantage that clock control over plant homeostasis might have. It did not take long for similar results to emerge in other eukaryotic systems, most notably in mammals.

Several studies illustrated the extensive influence of the circadian clock on tissue transcriptomes, most notably in the liver, brain, and heart (Akhtar et al. 2002; Panda et al. 2002; Storch et al. 2002; Ueda et al. 2002). Subsequently, a plethora of tissue transcriptomes has been mapped over the circadian cycle, including those of adipose tissue, gut, and bone (Polidarova et al. 2011; Zvonic et al. 2006, 2007). Together, these studies and others highlight that in excess of 10 % of the transcriptome in any individual tissue could undergo robust rhythmic change over the circadian day and night (Hughes et al. 2009). The functional consequences of this widespread control over gene expression by the clock are perhaps obvious but are only beginning to be recognized more widely outside the circadian clock field (Fig. 2).

Fig. 2
figure 00162

Microarray analysis of gene expression over a circadian cycle (i.e., in the absence of external time cues) in mouse liver. The top panel shows a heat map with genes that oscillate in a similar pattern clustered together. In this case, transcripts peaking in the middle of the cycle, CT12, are shown (CT circadian time; where CT0 is subjective dawn and CT12 is subjective dusk). The bottom panel shows the same data as a graphical representation for each gene. The left side of the heat map and graph assayed expression by an autocorrelation method. See Akhtar et al. (2002) for further details

Full size image

As microarray technology has matured, high-throughput sequencing seems set to take over its reign over transcriptomics research, in a similar way in which ChIP-chip has given way to ChIP-seq. The power of sequencing the transcriptome (after reverse transcription and processing into DNA) is clear (Hawkins et al. 2010; Marguerat and Bahler 2010; Wang et al. 2009). RNA-seq (the term used to describe this technique) is so powerful because it cannot only interrogate messenger RNA (mRNA) but can also be used to assay small RNAs, such as microRNAs (Cheng et al. 2007; Gatfield et al. 2009; Yang et al. 2008) and other noncoding RNA species (e.g., large noncoding RNA—lncRNA). Furthermore, RNA-seq can be used to perform systems-level analyses of alternative splicing, which may add further tiers of regulation to RNA processing by the clock (Licatalosi et al. 2008; Wang et al. 2010).

4 Interferomics and Manipulating the Clockwork

Interferomics is a novel area within systems biology that aims to study the biological impact of perturbing post-transcriptional (but pre-translational) processes (Baggs and Hogenesch 2010). The main tool that is becoming used increasingly in this area is RNA interference (RNAi). Using this technique, it is possible to silence a specific gene with a cell line using a small interfering RNA molecule (siRNA). With a suitable screening platform for circadian clock function, a collection of these siRNAs could be applied onto cells and phenotypes screened for using a suitable screening assay.

The pre-eminent assay system used by circadian biologists to assay the clockwork employs bioluminescent reporter constructs (Yamaguchi et al. 2001). These consist of “clock gene” promoters driving the expression of luciferase expression (e.g., Bmal1::luciferase and mPer2::luciferase) which act as markers for circadian oscillation within the cell line. Once reporters are introduced into cells stably, siRNAs can be transfected into reporter cells and their effects on the clock determined using real-time bioluminescence monitoring (Hastings et al. 2005).

This type of approach has recently been put into practice in two major studies (Maier et al. 2009; Zhang et al. 2009), with similar paradigms also used for chemical compound screening (Hirota et al. 2010). Interestingly both kinds of approach have delineated links to canonical kinase pathways, including casein kinases (Hirota et al. 2010; Maier et al. 2009). This highlights the power of complementary approaches to dissecting components of the transcription–translation feedback loop.

5 Beyond Transcription

Proteins are of course the final effectors of cellular function; what do we know about the impact of rhythmic transcription on protein levels following translation? Intuitively, this would seem to be quite a straightforward question. However, the data in the clock field and in other domains highlights that transcriptomics datasets do not correlate well with proteomics datasets from the same samples (Hanash 2003; Reddy et al. 2006), emphasizing the importance of mapping protein abundance in addition to mRNA expression. This point is further highlighted by recent data from Selbach and colleagues, who took a systems approach to determine the “flow” of mRNA to protein quantitatively (Schwanhausser et al. 2011). More detailed descriptions of post-translational aspects (e.g., proteomics and metabolomics) are considered in other chapters within this volume (Robles and Mann 2013).