Epigenetics, Bistability, Noise, and Development

Bacterial persisters and the role of stochastic gene expression as a putative mechanism in at least some cases of persistence (Wakamoto et al. 2013) raise the broad issue of gene regulation in development and evolution in changing environments. Gene regulation was alluded to earlier (see Chap. 2) and in many respects is similar in prokaryotes and eukaryotes. In bacteria, a fundamental regulatory mechanism is the operon, a cluster of functionally related structural genes under the control of a promoter, which together operate as a single transcriptional unit ( Chap. 8 in Madigan et al. 2015).

Enzymes for an entire pathway can be synthesized concurrently under the control of a single ‘on/off ’ switch (operator). This design allows bacteria to adjust their metabolism rapidly to environmental fluctuations (Chap. 3). Multiple operons, in turn, may be controlled at a higher level by a functional unit called a regulon. Control mechanisms exist at even higher organizational levels as well as at multiple lower levels. In eukaryotes, genes are packaged in chromatin and the transcription machinery is significantly more complex. As noted in Chap. 2, eukaryotic gene regulation can occur at any or several of six steps from gene to protein (Table 2.4 and Alberts et al. 2015).

Of these, control at the transcriptional level (including chromatin changes as well as the initiation of transcription) is the most efficient and presumably the most important. Even in Archaea and Bacteria, expression is coordinated at the transcriptional level by activator and repressor proteins coded for by a regulatory gene outside a particular operon. These proteins bind to cis-acting regulatory sequences (so-called c/s-regulatory elements) on the DNA, as they do in eukaryotes, where they regulate the degree of transcription by controlling the attachment of RNA polymerase. In multicellular eukaryotes, a given cell expresses only a small fraction of its genes and gene regulation is a central feature of cell differentiation, as well as in the reception/transduction of external or internal signals (Doebley and Lukens 1998; Carroll et al. 2005; Peter and Davidson 2015).

Virtually from the outset of genetics, the dogma became entrenched that phenotypic variability is attributable to mutation (implicitly, mutation of a structural gene). It was not until the classic work of Jacob and Monod on bacterial genetics in the 1960s that changes in gene expression became increasingly recognized as being evolutionarily important. Expression could change either directly, through alteration of what we know today as the cis-regulatory DNA of the gene, or indirectly, by changes in the upstream transcription factors that regulate the gene. The Jacob-Monod model established the important regulatory principles that gene expression is managed by an ‘on/off’ switch controlled by a DNA-binding protein and that this protein recognizes a particular DNA sequence in the vicinity of the gene.

These concepts were extended by numerous theoretical and experimental papers in the 1970s on gene expression in eukaryotes (e.g., Britten and Davidson 1971; King and Wilson 1975). When nongenetic, chemical changes to histones or to DNA (i.e., without affecting DNA sequence) are heritable, they are referred to in current semantics as epigenetic (Griffiths et al. 2015) and the phenomenon is well documented both in bacteria (Casadesus and Low 2006) and eukaryotes (Gilbert and Epel 2009). Examples include some forms of phase variation in bacteria; prion- associated phenotypic change (below); genomic imprinting (gene inherited from one parent is not expressed because DNA is methylated); and various forms of histone modification. At the population level, epigenetically inherited traits may be functionally indistinguishable from allelic alterations; in fact, in the case of plants, several instances of heritable variation now attributable to chromatin architecture were originally interpreted as DNA variants (Johannes et al. 2008; He et al. 2011).

Some forms of epigenetically acquired traits may be relatively fluid. In interpreting the epigenetic influence of a prion-like protein, Jarosz et al. (2014a) noted two differences between this inheritance mechanism and conventional nucleotide variation: First, because cells can acquire and lose prions relatively quickly, such elements could provide for a more rapid and dynamic means of adjusting to environmental fluctuations than does the process of mutation and reversion to wild-type in acquiring complex traits; second, environmental stresses can actually increase the rate of acquisition or loss of these proteins. In the case of Saccharomyces cerevisiae, acquisition of the protein-based epigenetic unit allows the yeast to overcome glucose repression (i.e., the inability to use even trace amounts of most other carbon sources in the presence of glucose; see Chap. 3) (Jarosz et al. 2014 a, b).

Certain bacteria associated with the fungus secrete a chemical factor that induces the epigenetic element. The resulting yeast phenotype becomes capable of using alternative carbon sources, becoming metabolically a carbohydrate generalist rather than a specialist, as it has come to be known from laboratory culture. This is a survival trait arising spontaneously in some members of a yeast population and likely of metabolic importance in the wild. Of particular interest and presumed ecological importance, the prion-like element was experimentally induced in wild strains but poorly or not at all in laboratory strains of S. cerevisiae by a factor secreted by wild but not by almost all laboratory strains of E. coli and B. subtilis tested.

Presence of the protein element in evolutionarily diverse fungi showed that this and similar epigenetic mechanisms are conserved over at least 100 million years. As the authors point out (Jarosz et al. 2014b), the phenomenon really is a heretical example of ‘Lamarckian evolution’: a heritable, epigenetic trait elicited by a chemical signal secreted into the environment. As prions have usually been assumed to be pathogens, this work also indicates how they may have originated or been conserved by evolution (see similar remarks regarding pleiotropy, later).

The fundamentals of gene expression in prokaryotes and eukaryotes summarized above are relatively well known in very broad terms and comprise standard fare in biology textbooks. However, historically, the biochemical and molecular genetic processes have been inferred from studies on large populations of cells and macromolecules. Events are viewed as deterministic and interpretations are based on averages compiled across countless individual events. Only recently with the advent of techniques to examine living, single cells has it become possible to examine transcriptional and translational mechanisms at the singlemolecule level (Eldar and Elowitz 2010; Xie et al. 2008; Li and Xie 2011; Chong et al. 2014).

The biochemistry of individual cells is quite unlike that of mass solution chemistry of cell populations. The former is characterized by macromolecules in low copy numbers, stochastic reactions, nonequilibrium conditions, and holistic complexity rather than reductionist simplicity (Xie et al. 2008). In bacterial cells in particular there are only one to a few copies of a particular gene, few of a particular mRNA (due to short lifetimes), and few of many important proteins such as those regulating gene expression (Li and Xie 2011). Events such as gene regulation therefore need to be viewed as single-molecule processes and as such are inherently noisy (stochastic), potentially leading to phenotypic heterogeneity within an otherwise identical, isogenic cell population.

Noise. Various types of dynamic behavior in eukaryotic, multicellular macroorganisms as well as in single-celled microorganisms occur as a result of stochasticity (Eldar and Elowitz 2010). These include the role of noise: (i) at the shortest timescales, in enabling certain physiological regulatory mechanisms such as the coordination of gene networks; (ii) in permitting a range of stochastic differentiation options not available in a deterministic system; and (iii) over the longest timescales, in facilitating adaptive evolution (some examples follow and for others see Eldar and Elowitz 2010). A specific case with a microorganism noted above is persistence of certain cell lineages of Mycobacterium smegmatis in the presence of the drug isoniazid (Wakamoto et al. 2013). Adaptation to the drug could potentially take the form of a lineage of cells characterized by infrequent stochastic pulses of the enzyme KatG. Although the pulsing mechanism is not clear, it appears due to noise-induced transcriptional bursts of KatG. Here the pulses, in turn, evidently result from reversible gyrase association and dissociation with a DNA segment, which changes the degree of supercoiling (Chong et al. 2014), though sources of molecular noise are various (Maheshri and O’Shea 2007; Potoyan and Wolynes 2015).

Other well-known bacterial phenomena ascribed to phenotypic bifurcation of a population (bistability) have been revealed in studies of the soil bacterium Bacillus subtilis, including genetic competence, spore formation/cannibalism, and swimming/chaining (Dubnau and Losick 2006). Noise can also lead to coexistence of distinct, bimodal states without strict bistability as such (To and Maheshri 2010). Transcriptional noise has been investigated in eukaryotes of varying complexity such as Saccharomyces (Blake et al. 2003) and mammalian cells (Suter et al. 2011), where it can be a significant variable influencing differentiation and development. Noise, while unavoidable and of functional value (Eldar and Elowitz 2010), may be subject to natural selection for minimization above a baseline level (Blake et al. 2003; Fraser et al. 2004) such as by modifications to promoter architecture (Kaern et al. 2005; Sanchez et al. 2013).

Molecular changes in multiple transcriptional control mechanisms undoubtedly played a major role in the evolution of metabolic pathways in bacteria and in the evolutionary divergence of bacterial species. Analogously, the importance of changes in regulatory as opposed to coding DNA has increasingly been emphasized in plant (Doebley and Lukens 1998) and animal (Peter and Davidson 2015) developmental biology. Carroll and colleagues (see e.g., Carroll et al. 2005; Prud’homme et al. 2007; Carroll 2008) attribute the greater role of regulatory DNA in morphological evolution to three factors: (i) the higher ‘degree of freedom’ in ris-regulatory sequences to mutational change (no need to maintain a particular reading frame); (ii) modularity of the ris-regulatory elements (individual elements can evolve independently); and (iii) combinatorial action of the transcriptional factor repertoire in animal cells. Doebley and Lukens (1998) view plant development as a cascade of events beginning with internal signals at embryogenesis, followed by both internal and external environmental signals conveyed through a hierarchically arranged modular system that is progressively activated. The components include transducers of the signals (e.g., ligands, receptor kinases); transcriptional regulators such as the ris-regulatory sequences and associated proteins noted above; and the target genes at any point in the relay system.

Depending on location, some target genes have only a local impact while others near the start of the cascade (e.g., signaling) can have far-reaching effects. An allele affecting several organism properties is generally termed pleiotropic. Pleiotropy is a good example of the developmental category of evolutionary constraints discussed in 7Chap. 1. Pleiotropy constrains the kinds of changes possible in morphological evolution (Prud’homme et al. 2007). A bacterium can be assembled faster and more easily than can an elephant. Therefore, the constraints of informational relay and sequential gene action in a cascade are presumably much less for bacteria than for elephants. Importantly, however, pleiotropic and related effects can nevertheless in some cases be beneficial in facilitating coordinated rather than individualized change of multiple phenotypic traits. Thus, despite its potentially adverse consequences, and depending on trade-offs, pleiotropy may be favored overall by natural selection (Cheverud 1984; Guillaume and Otto 2012). Changes in ris-regulatory regions rather than in coding sequences tend to circumvent or minimize pleiotropic effects (Prud’homme et al. 2007).