Understanding the emergence and maintenance of stable cellular phenotypes and the switching of phenotypes in response to environmental changes is at the forefront of biological research in diverse areas of study such as cancer and development. It is well known that identical genotypes can develop into diverse phenotypes. Moreover, isogenic cells in the same environment may exhibit some degree of phenotypic variability 
and can even switch between two well-determined phenotypes 
. However, given the apparent complexity of genetic networks, an open question is: what are the processes leading to the stabilization of a particular phenotypic state of a cell—its morphology, metabolism and function? Many years ago, Waddington coined the concept of “epigenetic landscape”, in an attempt to construct a useful metaphor for the underlying complex mechanism of genotype-to-phenotype transformation 
. Waddington proposed that “strategic” principles, beyond individual molecular interactions, were required for bridging the gap between the short-term physiological responses and the long-term evolutionary processes and this, he believed, will come through understanding the epigenetic landscapes underlying developmental processes (e.g., cell differentiation). Notwithstanding the impressive progress in molecular biology over the last decades, the strategic relation between the process of gene expression and the emergence of specific phenotypes has remained elusive.
On the one hand it is commonly thought that the emergence of a stable phenotype is the result of a programmed system, in which, for a given genome and environment, a set of “instructions” for the phenotype is instilled (via evolution) in the underlying genetic regulatory circuits 
; phenotypic variability is then the result of unavoidable “noise”. On the other hand, given the huge combinatorial phase-space spanned by the gene-expression degrees of freedom 
and the significant levels of intracellular and environmental fluctuations 
, it has been proposed that stable phenotypic states emerge as attractors in the phase-space determined by the concentrations of expressed proteins; given a genetic network architecture (connectivity), the finite number of attractors guarantees the stabilization of specific phenotypes by dynamically directing the initial vector of expressed proteins into one of its stable steady states 
. This attractive concept, a modern version of Waddington landscape metaphor, was developed theoretically within the framework of specific models and for certain classes of networks it was shown that attractors do emerge naturally in the system, i.e. they are properties of the network's connectivity and structure 
. However, many questions related to the attractor idea remain open: what intracellular processes do actually determine the stable attractors and their basins of attraction 
? Do these attractors reflect the intrinsic dynamic response of genetic networks to environmental signals? What is the level of degeneracy in the phase-space of expressed genes? Do many different attractors result in similar cell phenotypes? The experimental basis necessary to tackle these key issues is still lacking 
. In this paper we attempt to advance our understanding on these matters by studying the relation between the emergence of a stable phenotype in response to an environmental switch and the underlying gene expression dynamics.
Biological cells are history-determined systems, so understanding their intrinsic dynamics requires us to discriminate between necessity and contingence, between inevitable and accidently-instilled intracellular processes in evolution. Toward this end, we have studied the gene expression response of cell populations adapting to surmount a severe unforeseen challenge. Since it is unlikely that genomic circuits are “pre-programmed” to respond to any arbitrary novel perturbation, an unforeseen challenge has the potential of exposing gene expression patterns beyond the specific ones shaped by evolution 
. Here, we utilize this approach as an effective tool to probe the expression dynamics underlying the emergence of a novel stable phenotypic state of the cell. To gain further insight on the emerging dynamic patterns of expression we study for comparison, “wild-type” cell populations responding to a common environmental switch. The fascinating process of adaptation to a severe unforeseen challenge was discussed in our previous publications 
and is not the focus of the current paper. This setting of populations adapting to a challenge however, provides us with an effective model system to study long-term population expression dynamics throughout phenotypic changes that were unlikely to be “pre-programmed” into the cellular regulatory system.
Experiments were performed on unicellular yeast cells in which the essential gene HIS3
from the histidine biosynthesis pathway was detached from its natural regulatory system and was placed under the exclusive regulation of the GAL system responsible for galactose utilization 
. “Wild-type” cells with identical genomic background but deleted of the gene HIS3
were used for comparison. The arbitrary HIS3
rewiring, linking the foreign histidine and GAL systems, was shown to be stressful and challenging by creating incompatibilities in gene expression 
. In particular, a switch from a galactose-based to a glucose-based, histidine-lacking medium presented a severe unforeseen challenge to the cells since the GAL system and the GAL-controlled HIS3
were initially strongly repressed in glucose. Note that cells deleted of HIS3
could not survive in a medium lacking histidine 
. Recently, we have shown that a cell population carrying this GAL-HIS3
rewired genome could rapidly adapt (within ~10 generations) to grow competitively in this medium despite the strong initial repression of HIS3 
. Similar adaptation of genome-rewired cells to glucose was shown for different culture techniques: chemostats, batch cultures as well as for cells grown on agar plates 
. Once established, the adapted state had been propagated stably for hundreds of generations. Our previous work showed that the inherited adaptation was not due to selection
; every cell in the population had, in principle, the potential ability to adapt 
. Indeed, we have shown that the adaptation was due to a response of many individual cells to the glucose medium and not due to selection of rare advantageous phenotypes.
Intriguingly, underlying the adaptation process was a global re-organization of gene regulation. We have previously shown that the adapting cell populations exhibited genome-wide expression dynamics involving a sizable fraction of the genome and presented strong correlations between genes across functional modules 
. These results revealed that co-expression does not necessarily imply co-functionality. Moreover, the observed crosstalk between functional modules presumably played an important role in enabling the emergence of a proper metabolic state. We also observed the simultaneous induction and repression response of genes residing within the same functional metabolic module. Thus, co-functionality does not necessarily imply co-expression and there is no simple connection between transcriptional patterns and metabolism. Importantly, the global gene expression response was found to be non-reproducible between repeated experiments that nevertheless showed similar population growth dynamics and metabolism 
. This is a surprising result, since the irreproducibility in expression patterns was global and spanned the entire set of metabolic genes participating in the emergence and maintenance of a stable adapted growth phenotype. These results indicate that a spectrum of different gene expression patterns can potentially arise in populations under the same experimental conditions.
Gene expression response and its relation to the phenotypic cell state depend both on the environment and the history of the population. Thus, it is difficult to exclude the possibility that the variability in gene expression, even between isogenic populations grown in the same environment, results from their different histories. In this paper we overcome this problem by an experimental approach that enabled us to probe the gene expression patterns underlying phenotypic order through studies of the population dynamics while controlling for environmental conditions and population history. To compare the dynamics of gene expression between populations with identical histories, we developed a novel experimental setup in which two populations with a joint history could be separated at a defined time point and examined under identical environmental conditions. Our genome-rewired cell populations were grown in chemostats under severely challenging conditions in which cells fiercely competed for limited resources. Thus, the relevant phenotype that integrates essential metabolic functions was that of growth rate and proliferation and this phenotype was highly constrained for the adapting cells in our experiments.
The results of the present paper advance our previous work in two important aspects. First, we show here that chemostat populations with identical histories nevertheless demonstrated variable expression dynamics of essential genes. Second, by utilizing high temporal resolution, low-noise gene expression measurements, the set of experiments presented here show that the observed variable gene expression patterns were not due to cellular “noise”. Rather, these patterns of expression reflected collective dynamics resulting from synchronization of the expression response of the cells within the population. Thus, the population itself was the proper level of organization determining the cellular gene expression response via its collective dynamics. We demonstrated the generality of this mechanism by showing collective dynamics of gene expression also for “wild-type” cells.