I have compared tumours with ecosystems in previous occasions and one thing about ecosystems is that some people try to study and characterise them using networks (or graphs if your background is mathematics).
Basically that involves selecting key species and their interactions, be it predator/prey or mutualistic. Once a network is defined, it is possible to study a number of things such as network dynamics and robustness.
That is because network theory has been used to study all sorts of things including communications networks. Once a network has been laid out it is possible to study how fragile or how resilient it is. How easy is to include new nodes or add new links is a fundamental property of a communication network. Network analysis has proven how robust and scalable internet is against most disruptions…and also how fragile could it be against a determined attack against a few but selected servers.
There are parallels between ecological and non ecological networks. A (not that recent) review paper I recently found in Nature is precisely addressing this parallelisms. The paper is Ecological Networks and their fragility by Montoya et al. Nature Vol 422, July 2006. Can be found here
Some interesting things coming out from the paper is that some models suggest that species that have a big overlap in their diets can’t coexist. So basically if two species depend on basically the same nutrients to survive and live in the same ecosystem, eventually one of them will eventually become extinct. On the other hand the effect of one species on the numbers of another diminishes with the separation in the network or food web.
For a cancer I might speculate that this could imply that there is a limit to how many different phenotypes could there be simultaneously in a tumour (feeding on the same nutrients and competing for the same space) but Bob points out that the time constraints in which tumour evolution takes place are a better explanation of this potential lack of diversity I predict.
Unfortunately the paper does not get into the details of how to study the effects of taking down (or adding up) a new node or link in the network in terms of robustness. I am sure it would be very interesting to be able to characterise a tumour ecosystem in terms of a network and then identify nodes (tumour phenotypes) or links (cooperation or predation) that could be potential target for therapy
This idea is known as Gause’s Principle of Competitive Exclusion. Available from any good ecology textbook near you! The basis is easy to understand – if the fitnesses of different phenotypes are constant, then the one with the largest fitness will out-compete the rest. The same argument follows for species, except that there is variation within species, so what we can (and sometimes do) see instead is that the species evolve to become more distinct.
There is one problem with this – it might tell you how many there will be eventually, but that point may only be reached after the patient has died. I suspect cancers are so dynamic that equilibrium theory wouldn’t work too well. A lack of diversity may be more because there isn’t the time for it to be built up (I know hardly anything about within-cancer diversity, so feel free to tell me I’m wrong).
Hi Bob,
I take your point (I’ve updated the post to reflect that). I am no biologist but from my conversations with biologists and some physicians I gather that there is quite a bit of phenotypic diversity in a tumour. Also, there is mathematical work that correlates tumour phenotypic diversity with the surrounding microenvironment “here”:http://www.cell.com/content/article/fulltext/?uid=PIIS0092867406013481). Finally there are many cancers that take years (or decades) to develop so maybe there are occasions in which there is time to reach some sort of (not THAT stable) equilibrium? No idea, though, if that’d represent a sizeable number of cancers.
I suspect a cancer is a relatively simple system, compared to ecosystems, so your questions should be answerable. Actually, it looks more like population genetics than ecology – the different cancer genotypes will still be relatively similar.
Hmmm…
I’m sitting here wondering – yet again – about the relationship between tumors and ecosystems. We can visualize a tumor as a complex mix of cell types with distinct niches. Indeed we can imagine any organ in this way and think of the tumor as a caricature of the organ (I know it’s not an original thought, but it’s mid-afternoon and I need a caffeine fix!)
So during tumorigenesis we see changes in the phenotypes of the cells present and in the niches available. Some of these cellular changes represent cell types not normally seen in the local (for example chronic inflammatory cells, or tumor associated macrophages) which could be thought of as new “species” in the ecosystem. However other cell phenotypes might well be a result of changes induced by the environment. So cells can change shape and type to quite an amazing degree (EMT being a good example of a possible change). In the ecosystem analogy would these represent new species or newly evolving sub-populations (per Bob’s comment above)?
I suppose where I’m going with this is the idea that there are multiple populations which interact. They do give rise to distinct new niches within the tumor and there are both new cell types coming in and changes in the phenotype of resident cells. As you point out many of these things can occur over many years so stable, or at least metastable, tumors can arise. Certainly an understanding of signaling in such a network is valuable to identify the key nodes that can be targeted to make the whole thing collapse.
Not wanting to take this analogy too far, but what then is a metastasis – is this cells migrating out of a hostile environment “in search” of a new home? Or is it just random shedding of cells without the ability to hang on which move to metastatic sites in a quasi stochastic manner?
Enough – I need coffee!
Bob, thanks for your comment again. I am not sure whether it is straightforward to compare the complexity of a conventional ecosystem to that of a tumour ecosystem. My impression is that a tumour ecosystem is complex enough that a typical population genetics approach would be insufficient to answer some interesting questions. Consider the healthy tissue cells with various types of tumour phenotypes all of them interacting in different ways with different types of stromal cells, some of which have been co-opted by the tumour to aid it in its progression. On top of that, as Simon mentions, invading species such as immune cells and inflammatory cells. Add to that a complex environment that includes extra cellular matrix to which cells are normally attached, signalling and growth factors…
Simon, nice post. I am not surprised that you like the network analogy :)
The main question would probably remain if the tumour network (and I don’t mean exclusively signalling networks but also cell-to cell interaction networks, maybe between tumour phenotypes and carcinoma associated fibroblasts) is stable enough that is possible to target its weak points before those change. But I think that in many slow growing tumours, it is a mathematical tool worth considering. Hope you enjoyed an espresso!
Commenting on Gause principle… I was in Schola Biotheoretica XXXIV (of Estonia) about a month ago and there was some discussion on how the formulation of biological principles tends to reflect the background of the people who formulate them. As of Gause principle, it reflects that competitive exclusion in such a form is mostly studied in animal systems, because for plants the applicability of that principle is a lot less clear. Plants, in principle, they said, almost all live on the same nutrients… Though, of course, “niche” is more than only nutrients. But still, there are interesting forces like pest pressure that can work against one plant overcoming others that have similar requirements…
Hi Taivo, glad to see you here!
It is a good point: the exclusion principle implies that the species are competing for the same resource and I imagine that one important factor is that the resource has to be finite. I imagine (speculate) that the difference is that some resources for which plants fight for, are for all practical purposes unlimited (sunlight) with the limitation being stricter in terms of how the resource is acquired than in the possibility that the resource might be depleted by overconsumption.
Great thread. I am glad to see more attention paid both to the microenvironment and to interaction networks.
I would propose a cautionary note though. While it’s one thing to acknowledge that there are different types/species of cell interacting in tumorigenesis, it’s another to be able to create meaningful and predictive models of specific types. If new mutant “types” are evolving rapidly all the time, given enough (epi-)genetic heterogeneity there will always be types that have never been seen before and that are fit for the specific microenvironment in which it was born (the snowflake analogy comes to mind). This can be the case even if there are some pathways (or hallmarks) that are more likely to appear than others. Also, when you identify a “type”, you necessarily discount most of a phenotype’s actual and potential behavior, and thus you may not have really captured the important essence in any particular typology. So a question arises of how important, relatively speaking, is typology vs pathology vs hallmark-ology, and can any one of these be ignored or considered subsumed?
With respect to networks, I have been disappointed to see so much of the research focus on a network’s structure independent of what I would call its computational dynamics (i.e. the actual patterns of activation and interaction). Such dynamics are almost universally ignored from what I have seen. Additionally, there’s a lot more going on in structural dynamics of cell networks than preferential attachment, but that seems to be the model that network theory focuses on exclusively. Hopefully this all is changing, and I would love to hear of work that is not business as usual in network theory, especially as it applies to biological systems…
Hi Rafe, thanks again for your (well thought) comment. I take your point that focusing too much on the structure of the network and not that much on the dynamics is dangerous and might not capture the rapid increase in the number of cell types at certain stages of carcinogenesis. Also I agree that the concept of phenotype is a bit more fluid than the ones I normally consider in my research and that all sorts of factors and circumstances can effectively alter the behaviour of a cell type.
I don’t know much about research done in network dynamics so I cannot comment on its limitations although the people I know do use the preferential attachment that you were mentioning.