New drugs won’t shoot the messenger

Sharing is caringShare on Facebook0Tweet about this on Twitter0Share on LinkedIn0

(Cell, Oct 2013) New theoretical models suggest that targeting the dynamic features of a signaling network, rather than its individual components, will lead to drugs with reduced side effects.

Treating disease should be easier. Identify the genes causing the problem when mutated or corrupted by harmful pathogens, use a drug to fix them and – boom! – cure. We should be done before the game starts. But even in such utopia where drugs target one and only one gene, doctors and researchers would still face a simple problem with a fancy name: .

Our cells are constantly detecting and responding to signals from other cells and the environment, using what researchers call or networks. Of course, cells cannot afford having a single pathway for each one of the myriad of messages that it receives. Instead, there are components of these molecular networks that are highly interconnected to many other cellular components and are used to transduce a whole range of different signals. Perhaps not too surprisingly, the malfunction of these so-called signalling “hubs” is frequently associated with disease, making them the obvious target for pharmacological treatment. But because of pleiotropy, targeting hubs can unleash a suit of side effects.

There has to be a better way, and the key may lie in how signaling hubs process information. That is what Dr. Marcelo Behar and a team of researchers led by Prof. Alex Hoffmann at University of California, San Diego began exploring with computer models of signaling networks. In a recent Cell article, they report on their novel approach to predicting the outcome of pharmacological treatment based on a better understanding of the of a signaling cascade [1]. “A dauntingly complex problem turned out to have a simple solution when looked from the right angle” says Behar, while explaining that the trick was to think of transduction cascades “in terms of out-of-equilibrium, quasi-equilibrium, and steady states, building blocks for all intracellular signals”.

Let’s take a step back, to just before the problem became dauntingly complex.

Behar and colleagues knew that signaling hubs can often distinguish stimuli based on a signal’s , but no one really knew to what extent signaling dynamics could be exploited to improve pharmacological specificity. So they decided to conduct an to find out how a series of simulated perturbations would affect the capacity of a signaling hub to discriminate among different stimuli.

Behar and colleagues designed seven simple models of signaling hub topologies that should capture the functional characteristics of a large spectrum of biological networks. All the modules had a signal transducer X at their core, activated by the stimulus, and they differed in how X was turned off (from a simple spontaneous deactivation of X to more complex feedback mechanisms that involved an additional component Y and connections among them). The researchers then simulated the activation of these modules by ten signals with different dynamic properties, and compared their behavior before and after a series of “perturbations” that mimicked what drugs could do in cells.

For this screen, the authors decided to mainly focus on a module’s ability to discern between acute versus sustained stimulation, a capacity that is likely shared by many signaling hubs in real world cells. To this end, they characterized the “early maximum” (E) and “average late” (L) activation of X in response to a stimulus, hoping that they could identify perturbations that would specifically affect one or the other, and that would do so across different module topologies and signal dynamics. Such a correlation would have made it possible to start envisioning new classes of drugs with lesser side effects. However, they found that these perturbations often had signal-specific effects that were too tightly dependent on a particular hub topology (doing little to closely related topologies) or, even worse, that they had opposite effects on a hub’s performance depending on the particulars of the stimulus being applied. In other words, the screen didn’t reveal any “virtual drugs” that would work over a wide range of hub topologies and signals.

What failed? Behar and colleagues had to dive even deeper, and used hardcore math to tackle the problem. The solution came from solving the for each of the hub topologies, before and after a perturbation. This approach allowed them to more accurately predict if a perturbation could specifically affect the transient (early) or steady-state (late) activation of a signaling hub, and thus establish when a “virtual drug” could have a more restricted mode of action.

In order to test drive their modelling approach, the authors used a signaling pathway that they knew well: the Nuclear Factor kappa B, or . Turning what they knew about this pathway into the necessary mathematical equations, the authors could first establish an abstraction of the NFkB pathway that looked like one of their simulated hub modules, which allowed them to quickly predict a series of real pharmacological treatments that should affect the rapid or slow response of the pathway to two kinds of stimuli with different dynamics. And when they tried those perturbation in cultured cells, their predictions panned out satisfyingly well.

Behar warns that it won’t always be so easy, because “it’s not always possible to independently change time-scales and dose-responses. One almost always affects the other. The cleanest effects can be achieved when there is a strong between parts of the network”. Time-scales are intricately wired in our cells and drugs can’t easily change them.

Another concern is that targeting signaling dynamic controllers may be a little like kicking the can. “My sense is that targeting dynamics may help you avoid pitfalls of pleiotropy at some signal transduction levels, but ultimately some pleiotropic-derived effects at the gene level will remain”, recognizes Behar. However, he adds, “instead of shutting down a communication satellite, you are now blocking only some frequency chunks. As a result you are still going to lose a bunch of channels and perhaps let through some that you don’t want. But it beats the alternative of shutting down the whole satellite”.

Now Behar and colleagues are looking down a few paths down the road. “I would like to test the effects of the many drugs known to affect NFkB, and see if they affect the expression of its target genes in the ways that we would predict”. Because, as he explains, it is also necessary to know how the genes of interest react to changes in signal dynamics. “This is a very young area of research and will take a while and substantial experimental work to mature in human systems”, adds Behar.

The team is also excited by another corollary from their study: “Because perturbations, be them environmental, chemical, or genetic, affect the different dynamical parts of signal, we reasoned that we could use these backwards to infer the mechanistic mode of action of poorly understood perturbations. I am currently trying to build a conceptual and practical framework to do just that”.

And what does this all mean for the future of pharmacology? Should we fundamentally change our search for new drugs? Maybe, at least to some extent. “The data to generate the equilibrium surfaces are rarely available” explains Behar, and he argues that it will be critical “to add time profiles of important signaling pathways to the readouts of stimulus-based screens. Though quite challenging from a technical perspective, one would also like to have data for different stimulation regimes (slow rising vs. fast, etc). Not all of these tests need to be done at a full-screen scale, but having some baseline data would go a long way”. But perhaps more importantly, he explains, we need to change our way of analyzing data from pharmacological screens. “Statistical analysis is still key, but familiarity with the system dynamics of signal transduction networks is now a necessity. If you see such skill set in a job advertisement for a pharmaceutical company, let me know because I want to start buying that company’s shares!”, says Behar.

One thing is for sure: the field of signal transduction faces a very dynamic future. No pun intended.


Mariano A. Loza-Coll, PhD

[1] Behar M, Barken D, Werner SL, Hoffmann A. The dynamics of signaling as a pharmacological target. Cell 2013 Oct 10;155(2):448-61. doi:10.1016/j.cell.2013.09.018

Sharing is caringShare on Facebook0Tweet about this on Twitter0Share on LinkedIn0
Pleiotropy is a term used in genetics to indicate that a gene affects several physical characteristics of a cell or organism (from ancient Greek, “pleion”: more + “tropos”: turns). How pleiotropy is a problem for modern medicine can be illustrated with the following analogy: in order to prevent the broadcast of a very dangerous message, the best that biomedical research can sometimes offer is to take out the satellite used to relay it, effectively silencing the harmful message but also eliminating countless other ones.
When cells detect a signal, they use a series of relay molecules to communicate the message to other cellular components through a cascade of chemical modifications. The molecule that first detects the signal, or” receptor”, alters the molecular structure of a second one, which in turn alters the structure of a third, and so forth. These molecular games of tag relay the signal until a final effector becomes “it” and carries out a cellular response to the stimulus. For example, some of these transduction cascades culminate in changes to the expression of a whole battery of genes, due to the chemical alteration of the cellular components in charge of controlling gene expression.
Scientists often describe a system (for instance, a signaling network) in terms of interactions between its components and what they do to each other. This is normally referred to as the “topology” or “structure” of a network, and it’s simply a description of the hardwiring of a system. To fully understand a system, however, we also need to know the timing of its inner motions and interactions (some of those interactions can happen in an instant, while some may take a little longer). This is what experts normally refer to as the “dynamics” of a system.
Here, too, “dynamics” refers to timing. Signals not only differ in how intense they are, but how long they last, how quickly they achieve maximum intensity and the way they reach maximum intensity and turn off (do they ramp up gradually and steadily, or suddenly blast off/shut down?). All of these characteristics of a signal are described by what scientists call its dynamic properties.
Scientists frequently talk about ‘in vitro‘ and ‘in vivo‘ experiments. in vitro is Latin for ‘’in glass’, and it’s a historical remnant from the time when researchers did their experiments in glass vials, tubes and culture plates. By contrast, in vivo refers to experiments done with live animals or plants. A recent addition to this list of experimental set ups is ‘in silico‘, which originates from the core component of computer chips and refers to computational simulations of biological systems or processes. In this case, an ‘in silico screen’ means that the researchers used computers to test a comprehensive series of signaling scenarios by altering numerical parameters in their models.
These are maps that scientists can draw to predict the behavior of a system following stimulation. Imagine a world map, with its continents and oceans. A given stimulus and a few other parameters provide the coordinates of where on the map the system “falls” during stimulation. If on “dry land”, the system is said to be in “equilibrium” or “quasi-equilibrium” and many of its variables will remain more or less the same (for instance, the intensity of the signal transduced by the system). Contrarily, if the system falls on an ocean, it is said to be “out-of-equilibrium”, and it will change following a predictable trajectory (as if following an oceanic current) until it reaches dry land. More importantly, researchers can also predict how perturbations change the shape, size and positions of the “continents” in these maps.
NFkB, the core component of this pathway, responds to external signals by moving into the nucleus of a cell and binding DNA at specific locations, to turn on the expression of its so-called target genes. One of NFkB’s jobs is to alert cells about the presence of potentially harmful invaders, which it does by responding to molecules produced by cells of our immune system that have detected their presence, or molecules produced by the harmful microbes themselves.
This refers to processes within the cell that take much shorter or longer than others, and that are part of the same feedback loop within a signaling network. For instance, proteins can get very rapidly activated by simple chemical modifications that occur in seconds within a cell. But the inactivation of the same protein may involve shuttling it to a different cellular compartment to have it destroyed. This can take minutes, which means that there is a strong time-scale separation between the two parts of this loop (activation and inactivation of the protein).
Phenotype, a word derived from the Greek phainein (to shine, show or appear), is a term historically used to describe the physical manifestation of a given group of genes (or “genotype”). Nowadays, the term applies to a wider range of physical manifestations, like behavior, and scientists often use “phenotype” to refer to any observable change in a system. For instance, it is common to hear them say “treatment X showed a phenotype”, indicating that treatment X caused observable changes in the system under study. In this case “dynamical phenotypes” refers to changes in the dynamical properties of a signal relayed by the hub.

Leave a Reply

Your email address will not be published.