My lab is interested in the control of cellular behaviors, with the word control used in its engineering sense: the execution of strategies for achieving useful ends, such as precision, robustness, efficiency and fast response. Biological development and regeneration provide impressive examples of tight control of growth, differentiation and patterning. Numerous developmental processes have been extensively studied from a mechanistic perspective (What happens? What molecules are involved? How do they work?), but only recently have serious efforts been directed toward understanding the logic of control (Why is the system configured the way it is? What purpose does it serve?).
We seek to address control questions in the context of long-standing biological problems in the areas of pattern formation, growth regulation and regeneration. We are particularly interested in how selection for tight control imposes order on biological systems, as a result of tradeoffs that naturally arise among competing objectives. We are interested in the connection between such order and the tendency of biological systems to be combinatorially fragile, i.e. to break down when faced by combinations of small perturbations that would, on their own, be innocuous.
We try to address these sorts of questions using a variety of traditional biology approaches: model organisms (mice, flies, fish), genetic manipulation, and genomic analysis. But because control problems are inherently systems-level–i.e. they deal with the behavior of an entire system in its environment–we utilize many of the tools of Systems Biology in our research, including mathematical modeling, computational simulation, large-scale data collection, and high-resolution live cell imaging. To facilitate this work, we collaborate extensively with mathematicians, computer scientists, physicists and engineers.
Some of the specific areas we focus on now, or have recently worked on, are discussed in detail below.
Morphogen gradients and their role in pattern formation. Morphogens are diffusible molecules that are released at one location in a tissue and, by virtue of transport (e.g. diffusion), create spatial gradients to which surrounding cells respond differentially, according to the amount of morphogen they interact with. It is now well established that many cells in developing organisms use morphogen gradients to receive positional information, which enables them to grow or differentiate appropriately for their locations. How the mechanisms that create such gradients achieve the right levels of morphogen at the requisite points in space, despite the presence of large intrinsic and environmental disturbances, is still a mystery. We believe that a great deal of the molecular machinery utilized in the formation of morphogen gradients exists because it is needed to counteract or filter out such disturbances. We already have many clues about different ways in which individual classes of perturbations may be resisted, but mathematical modeling (and the real world experience of engineers) tells us that strategies designed for single purposes often interfere with each other. The need to be robust to many types of disturbances drives engineered systems to become quite complex, and we have every reason to believe the same is true for biology. Put in other words, there is every reason to believe that much of the complexity of biology only makes sense in the context of its role in control.
So far we’ve studied robustness and control of morphogen gradients in fruit flies, zebrafish, and mice, using both mathematical modeling and a variety of genetic tools for manipulating morphogen production and response.
Proliferative control and its relationship to cancer. The sizes of tissues and organs are specified with extraordinary reliability, mainly by getting cells to stop proliferating after just the right number of them have been made. Because proliferation is an exponential (autocatalytic) process, even small errors should compound, making it all the more impressive that good control is achieved. Even more remarkably, experiments show that, in most cases, control is local (distant tissues are not coordinating with each other), is not achieved by counting cell cycles or measuring elapsed time, and the final sizes of tissues tend be sensitive neither to the number of cells from which they were founded, nor the rate at which cells turnover.
We use experiments (in mice) and theory (in silico) to identify biological circuits that mediate such feats of control. The principle of integral negative feedback turns out to be essential. One way such feedback is achieved is when diffusible factors produced by differentiated cells influence the probability that stem cell progeny differentiate into more mature cells (“renewal control”). In continually self-renewing tissues, however, this strategy is subject to undesirable oscillations, which can be prevented by using feedback to direct progenitor cells down lineage branches (“fate control”). Other lineage modifications can allow tissues to stabilize at sizes much larger than the spatial ranges over which diffusible factors can act.
Overall, through the investigation of how tissue size is controlled, we uncover novel justifications for common features of real cell lineages. This is consistent with the view that fundamental control principles shape the architectures of biological systems.
Recently, our work on proliferative control has branched out into cancer biology. Cancer is normally thought of us a total loss of growth control, prompting cancer biologists to pay little attention to the mechanisms that normally keep cells in check. Yet, as discussed above, growth control mechanisms are likely to be elaborate, and serve many objectives, from regulating steady state size, to influencing growth dynamics (e.g. regeneration speed), to managing competition between cells, to dealing the consequences of noise and stochastic behavior. Cancer cells need only modify these pathways just enough to escape homeostasis, so we believe they are very likely still carry around a lot of “growth control baggage”, which could be exploited for therapy. One way to learn about such baggage is to study benign lesions, many of which start out growing abnormally but eventually stop. Common moles (nevi), for example, carry the same oncogene mutation that is the most common driver of melanoma, yet they reliably stop growing through mechanisms that, we believe, are very similar to those used during development to stop the growth of tissues at fixed sizes.
“Transcriptomopathies” and the multifactorial origins of birth defects. A relatively common, severe birth defects syndrome, Cornelia de Lange Syndrome (CdLS), is most commonly caused by a small reduction in the level of the product of the NIPBL gene. NIPBL regulates cohesin, a structural component of chromosomes, and together with cohesin influences long-range cis-regulatory interactions throughout the genome. We developed a mouse model with which to understand the pathophysiology of CdLS, and to develop new diagnostic and therapeutic modalities. Analysis of this model indicates that small (10-50%) changes in the expression of a large number of genes (~1000) collectively give rise to the heart, limb, gut, brain, kidney and other defects observed in CdLS.
It has recently been recognized that many birth defects syndromes have an analogous etiology–they arise from subtle but pervasive dysregulation of the transcriptome, due to mutations that affect molecules broadly involved in transcriptional regulation. Such disorders have been termed “transcriptomopathies”. One thing that is particularly interesting about the transcriptomopathies is that, by producing reproducible phenotypes through the combinatorial effects of many small changes in gene function, they can serve as a model of the etiology of polygenic traits (traits caused by the combined effects of multiple, presumably-small-effect, alleles).
We are actively working on the Systems Biology of CdLS, as a way to understand principles governing the emergence of complex, polygenic phenotypes. We are focusing on the early development of the heart, which is abnormal in CdLS, as well as Nipbl-deficient mice. Doing this work requires monitoring gene expression at the single cell level, which we currently do with single-cell RNA sequencing.
The evolution of combinatorial fragility. The phenotypes in transcriptomopathies provide an example of combinatorial fragility, whereby multiple, small, seemingly innocuous changes collectively manifest as large deleterious traits. Results of genome-wide association studies suggest that many human traits result from groups of small genetic effects, interacting either additively or synergistically. Interestingly, evolutionary simulations strongly suggest that synergistically fragile interactions ought to arise spontaneously in almost any biological system that is under strong selection for robustness. We’ve used both mathematical modeling and population genetic simulations to explore this interesting idea, with the hope of finding better ways to mine genetic data for the causes of serious, common diseases.