The Perceptual Life of Cells

Over the summer, at ICPA 2019, I met a computational biologist called Katie Bentley. She is interested in angiogenesis, the cellular level process of new blood vessel growth. She was at ICPA because she has been developing a more dynamical systems approach to understanding how angiogenesis works. More specifically, her work suggests that the cells involved behave in very active, exploratory ways that are very analogous to the kinds of perception-action systems ecological psychologists study, and she was looking for people to help bring our insights to her work. We've been chatting, and the project of connecting our work is now on the go. 

The basic idea is simple. Mainstream cell biology is currently very gene-centric, in almost exactly the same way as mainstream cognitive science is very brain-centric. Like cognitive science, this has been methods driven - it's been hard to study cells in action, and imaging techniques have been static and structural. In general, when biologists see a cell doing something, the research goes looking for the genes that are making it do that thing. Bentley's work has instead started asking questions about time - how long does something take to happen? How are the component processes organised in time? What happens if that timing is perturbed? She's now in a good position to say that one of the key regulators of the dynamics of angiogenesis is timing, that this timing is not centrally controlled by a genetic clock, and that a temporal perspective should therefore move front and centre of the research on the process.

The analogy between the debates between information processing and ecological accounts of cognition and behaviour are uncanny, and her move is a very ecological one. The question we are now discussing is whether it makes sense to start considering this process as not just a dynamical process, but a perception-action process, which has become an option given her temporal perspective on things. 

In this post I'm going to discuss some of the specific empirical findings reviewed in her paper The Temporal Basis of Angiogenesis, which is a good read even if you aren't a cell biologist. I'm getting ready to go hang out in her lab this week and I'm working on figuring out the best questions to be asking to move this project forward in the best way. Remember, Gibson's theory is about what happens at the ecological scale of organisms like us, and there's no guarantee that it will have much to say about the perceptual life of cells. However, there is definitely enough evidence to make it worth checking out, and that's what I'll be up to for the next little while. 

Angiogenesis

When blood vessel cells detect hypoxia (low or insufficient oxygen) they begin to sprout new growths to expand the ability of the circulatory system to get oxygen to the deprived area. One cell becomes a tip cell that leads the way towards the affected area, and neighbouring cells become stalk cells that form the growth extending into the deprived tissue. The mobile tip cells make radical changes to their shapes so as to be able to move. The research question is, what is the mechanism by which a tip cell is selected, and what makes the cells move the way that they do?

The Central Pattern Generator Approach

Early research focused on the internal dynamics of the cells to explain these. When hypoxia is detected (via changes in signals such as vascular endothelial growth factor, VEGF), all the cells in the neighbourhood begin to decide who will be the tip cell via a process of lateral inhibition. VEGF detection triggers a genetic response which produces a protein D114; this then binds to Notch receptors on neighbouring cells, which has the effect of reducing that cell's ability to detect VEGF. The cell best able to keep detecting VEGF becomes the migratory tip cell (it starts to change shape to enable it to move) and it's neighbours become inhibited and become stalk cells. This central pattern generator (CPG) mechanism takes VEGF as input and runs according to the behaviour of the genes triggered by that input. Bentley calls this the decide-then-move model of angiogenesis. 

The main issue here is that the CPG takes too long to run; each cycle of detect-gene activity-cell response takes 4-6 hours, and multiple cycles are required; but the cell shape changes that allow it to move begin to happen much faster (on the order of minutes). So the CPG decision to become the cell that moves does not happen before the cells begin to move! It cannot be just the CPG in the driver's seat.

Bentley instead proposes a move to a move-while-deciding model, where the cells are all actively trying to change so as to move. This activity couples to the CPG, and this expanded process selects stalk cells, rather than tip cells. From this perspective, cells are not static entities that must change so as to move, they are dynamic entities that must change to stay still. 

Cell Selection as a Sensorimotor Process

There is a variety of evidence in favour of this more dynamic model

• Multiple cells in the stalk region remain active and mobile throughout the tip cell selection process
• During the process, the cell repeatedly extrudes and retracts filopodia and lamellapodia in what looks like exploratory behaviour (more on this later). Using her mechanistic model, Bentley has shown that removing these makes the switching between tip and stalk states slow and gradual (perhaps just driven by the internal CPG dynamics). In fact, anything that reduces the amount of filopodia activity slows the whole process down.
• Over time, during the selection process, which cell is the tip cell keeps switching as individual cells move around within the stalk. This jostling emerges from the fact that the cells vary in how adhesively they are connected to their neighbours, and this process is then calibrated to support useful cell migration by the Notch/VEGF CPG system.
• The CPG system is a lateral inhibition system that leads to an alternating pattern of active and inhibited cells. As cells change position, they need to change whether they are inhibited or not to preserve this pattern, and this does not seem to run on the clock defined by the CPG.
• The timing of all this activity can vary throughout the sprouting tissue, which has consequences on the branching of the tissue.
• Overwhelming the CPG with pathological levels of VEGF alters the cell's basic tendency to be on the move, which also affects the final shape of the tissue. Treatments (in mice) that target both the levels of VEGF and the cell's metabolism are required to restore normal cell rearrangement behaviours.

Overall, there is now clear evidence that the internal cell dynamics (the CPG) do not determine the timing in the behaviour of the cells. Instead, those dynamics must be coupled to cells that are mobile, adhere to their neighbours to varying degrees, and that are actively exploring their environments with filopodia. Decide-then-move is wrong; moving-while-deciding seems to be the rule. 

Is This a Perception-Action System?

There's no doubt now that angiogenesis is a dynamical process in which stable behaviour emerges from a coalition of constraints, rather than a central gene-centric process. Some of those constraints are internal (e.g. the CPG), some are about cell embodiment (e.g. differential adhesion), and some are external (e.g. detected levels of VEGF). But is the cell perceiving and acting?

Bentley's primary reason for thinking that it is are the filopodia/lamallapodia. These behave quickly (on the order of minutes) and they extend and retract in and out of the cell's local environment, carrying VEGF detectors in, through and out of the VEGF that is non-uniformly distributed throughout that environment. She suggests that these are the cell engaging in active perception, analogous to a human moving their eyes or their heads or their bodies to create and interact with variables in optic flow. Given how important they are to the success of the process, and given that they are active, this seems like a good hypothesis.

Is It Behaviour?

But what does it mean for this kind of activity to be considered perception? In our Ecological Representations paper, we noted (pg 236) that in order for there to be a need for a psychological level to your explanation, there needs to be a gap in the causal chain of events as described by a more mechanical, physics based process. If you are sitting in a tree and the branch under you breaks, the physics of gravity is all that's needed to explain why you moved. However, if you are sitting under a tree and a branch falls close to you, we need a psychological level explanation for why you got up and moved away. Cognitive scientists fill the gap with representations; ecological psychologists fill it with information and the resulting perception-action processes.

This was effectively a swing at defining what counts as a behaviour that psychology should try to explain, vs something that it does not need to get into. Our hunch was that in the latter case, two things mattered. First, there is a clear gap in the physical chain of events, and second, this gap can be seen because there was no obligatory response. I noted you might get up and move away; but the branch falling near you might make you do lots of things, depending on your current state and whether you perceived the event as dangerous or not. If the branch supporting you breaks and falls, you don't have these kinds of options.

The key question seems to be, how mechanical is the cell's behaviour? It's clearly dynamical and complicated, but then, so is weather and no one feels the need to say the climate is perceiving anything. How optional is the cell's response to what it detects via these active filopodia? Maybe their motion is less about generating information and more about simply optimising the cell's sampling of the local conditions, in order to get the optimal input for the rest of the system to work on. The latter, while cool, doesn't quite seem like perception in the ecological sense. We need to more fully characterise the exploratory behaviour of the cells; is it some kind of optimised random-walk type procedure, or is it being structured with respect to the environment?

Is there Any Information?

To answer that question, we need to first answer the question of whether there is anything like ecological information in the cell's environment. In vision, we ask 'what is the geometric form of the optic array and what can it specify?', with the answer 'it's a spherical vector field of flowing elements and these elements can flow in a large variety of ways specific to the dynamical property that caused that flow'. 

Translating that question to the cells, I want to know 'what is the geometric form of the VEGF array and what can it specify?'. If it's just a simple gradient field with a simple more-is-better relationship to the cell, then it's not clear that the cell needs to do anything called perception, in the ecological sense. If, however, there's more to it than that, and if the behaviour of the filopodia is structured so as to reveal or access this higher-order structure, then we may have a ball game. 

Conclusion

There is much to do, and one of the most important things to do is to not over-extend the ecological approach into a domain where it is asking the wrong questions. However, as it stands, it's clear that angiogenesis is the kind of self-organising, emergent dynamic operating under multiple constraints that we observe at the scale of organism behaviour. While one of those constraints is internal (the CPG, analogous to the brain being in charge), the differential adhesion constraint is clearly an embodiment constraint, and there is the clear potential for the VEGF constraint to be an information constraint that requires active perception. I therefore currently think that there is a case to explore here, and I'm looking forward to getting right into it. 

References

Bentley, K., & Chakravartula, S. (2017). The temporal basis of angiogenesis. Philosophical Transactions of the Royal Society B: Biological Sciences, 372(1720), 20150522.