Lecture by Hugues Bersini

Date: 16th June 2021

Speaker of the day: Hugues Bersini

Articles: Varela, “A Cognitive View of Immune System” (1994)

Abstract: Varela and colleagues have developed an alternative vision of the immune system where autoimmunity plays the central role. We have delved into the many implications this approach has for our understanding of cognitive systems, personal identity, various diseases, and life in general.

Keywords: idiotypic network, immune system, autoimmunity, circular causality, metadynamics, self/non-self, cognitive system

Summary

Presentation

In cell-recognition vision of immunology, B-cells produce antibodies (AB) which are supposed to match with the antigens (AG). The B-cells will then amplify the production of the AB, which will, to make it short, kill the AG. In their network-recognition vision of immunology, Francisco Varela and Antonio Coutinho were discussing the possibility for the AB to recognize not only the AG, but other AB as well. Using a connectivity matrix, they have demonstrated that in a small network of 28 species of AB, some of the AB matched with nearly all species of AB, others with several species, and still others with only one species of AB in a keylock manner. That indicated the existence of a network of AB, from which it may follow that the immune system (IS) is autonomously active in a manner independent from external impacts [however, that is not to be confused with being wholly independent from the environmental impacts – cf. point 2 in the discussion]. In an experiment later on, Francisco and Antonio compared the IS activity in mice in isolated environment vs. in mice that had been exposed to a pathogen or an AG. They sought for evidence of a strong IS activity even in a fully protected mouse. Despite the unfavourable research conditions (their inquiry was so different from the classical view of immunology that even the fellow scientists from the Paris immunology group declined to perform these experiments), they have managed to obtain some results, but it was hard to defend them – and it still is up to this day.

Classical immunologists tend to believe that in autoimmune diseases, the IS becomes deranged on its own and tries to destroy itself. Francisco published an article in PNAS where he was discussing the autoimmune diseases being testified by a different type of dynamics, where a perturbation of a network results in a so-called topological disease. In 1995, we wrote a paper Neural tolerance in a simple immune system (published in Journal of theoretical biology) where we, based on deep mathematics and computer simulation, demonstrated that an AG will be tolerated when perturbing a closed network of three clones (= AB), but destroyed if the network is open. This is an instance of circular causality – the choice between tolerance and immunization depends on the connectivity of the network. The network is constructed in time: the IS is refreshing itself at a very high rate. Francisco called this the metadynamics: the network is simultaneously growing by recruiting new nodes and shrinking due to disappearance of the existing ones (cf. Detours et al., Development of an idiotypic network in shape space). There is a key interdependency between dynamics and metadynamics: the dynamics decide on who should be recruited, and the metadynamics, because of the recruition of new nodes, will modify the dynamics.

We proposed the IS to be divided into the central immune system (CIS) and the peripheral immune system (PIS, which is the subject of classical immunology). CIS integrates the self-antigens, guaranteeing a tolerance but also shaping the reaction of the IS. PIS consists of clones that are not recruited in the CIS but are just floating around and are fully available to react to the antigens that are not hosted by the CIS.

The division of the IS into a tolerant and an immune zone shows that the determination of self and non-self is not imposed on the system from the outside: it is the very network that makes the decision. Immunologist Niels Kaj Jerne was the first to propose the idea of the AB interacting with each other in a so-called idiotypic network.

The concept of metadynamics has been an incredible anticipation of what was to become a successful idea 20 years later – namely, complex systems in biology. In one of the most important papers in physics of the last two decades, Barabási et al. presented scale-free networks, introducing the notion of preferential attachment: the network’s evolving topology determines the way that new nodes are recruited.

One important lesson to be taken from Francisco’s work is to reestablish the importance of circularity in biology. Most biologists still subscribe to a linear vision of biological systems: classical immunology presumes that between the impact and the IS’s answer there is a cascade of reactions that is, at all of its levels, still highly influenced by the first impact; it is the same with the neural networks and in genetics, where there is a linear path from DNA to RNA to a protein. If you accept the shift to circular causality, everything becomes much more interesting and complex, and also more realistic. There is a lot of internal feedback within the system – for instance, more than 80 % of connections in the brain are feedback loops, rather than external connections; epigenetics has an enormous role in genetics; also, circularity is the origin of autonomy and spontaneity in biology: in Life: An introduction to complex systems biology, Kunihiko Kaneko demonstrates that despite an identical external impact, the reaction of the system will be highly different, being dependent on the system’s state – so it is not enough to understand the impact, but the internal, network effect as well.

There is a growing interest in the idea of topological diseases – it is observed that the disease appears after the dynamics in the some organ/organic system has changed (e.g., in brain or cardiovascular conditions). Some immunologists go along with Francisco, e.g., Alfred I. Tauber and Irun R. Cohen – the latter, though they have never met, came to very similar conclusions as Francisco: “Rejection of the infectious agents depends more on the site and circumstances of the infected tissue than it does on the identity of the infectious agent.” Nevertheless, autoimmunity and tolerance have never been the mainstream immunology as compared to the defense vision: even if we need CIS to study PIS, it is still possible to study PIS without it. The autoimmune diseases are not well researched to this day, despite that a lot of them are known, like lupus, rheumatism and multiple sclerosis.

Another possible reason for Francisco’s ideas not being more widespread in the immunological community is the style of his scientific writing, which tends not to be favoured among the computer scientists and mathematicians. He liked to assume a certain kind of vagueness, seeing a richness in it: he used a lot of repetitions, describing the same idea with different metaphors. Also, when you put his concepts into code, the magic somewhat disappears, and you fall into the more classical notions. Take, for instance, the autopoiesis: the code connects it to a lot of works that are doing the same thing, namely the widespread research about the minimal cell in the artificial life community. Strong immunologists and strong biologists tend to tell me that though they find Francisco’s work to be fantastic, he is too much of a philosopher for them. Further, if you put it into code, by definition, any strong emergence will disappear: the code only allows for a weak form of emergence. In cellular automata, like in autopoiesis, there cannot be a top-down causality. The not one, not two really becomes one: there is no way to disambiguiate it. Nonetheless, it is a question whether Francisco’s ideas are always computerizable.

Discussion

1) Sebastjan Vörös: In retrospective, scientists tend to be captivated by the key historical scientific figures’ breadth of knowledge and interests, but their reaction is usually different when they are confronted with a brilliant mind whose novel ideas constitute a horizon of philosophy around the theories he or she is developing.

Hugues: Indeed. Ever since his collaboration with Francisco, Antonio, who used to be a mainstream immunologist, has been marginalized by most of the classical immunologists. This was one of the reasons why he left the Paris immune group, added that even his colleagues criticized him for the Varelian evolution of his work.

2) Sebastjan: How would one explain covid in a traditional way vs. from the Varelian perspective?

Hugues: Not everyone reacts in the same way to (the same mutation of) this virus. While some become so ill as to barely get through it, others remain asymptomatic, and this is a proof that the virus alone cannot explain the unfolding of the disease. It is definitely not the only culprit for an attack on a certain organ – while it is true that it has perturbed the system (likely its connectivity), the system on itself will interact with the body.

Among the most promising treatments when the covid reaches a very serious stage has been the injecting of monoclonal AB. By integrating the donor’s AB into the network of existent AB, you can reestablish a certain connectivity that will help the sick to recover. Similarly, Cohen treats some autoimmune diseases by stimulating the IS either with AB or lymphocyte clones. In autoimmune diseases, the T-cells are destroying a healthy organ; in order to prevent further damage, we can help the T-cells to destroy themselves when they are in the wrong. Cohen obtains healthy T-cells from a donor (or even from the sick person themself, amplifying them and injecting them back) and, by injecting them, perturbs the patient’s IS. This reestablishes the right systemic functioning of the IS, resulting in an improvement of a given autoimmune condition.

Natalie Depraz: Would Francisco have wanted to be vaccinated against covid or would he have let his organism and self resist?

Hugues: I think he would have been vaccinated. We know that the vaccine prevents the virus from getting into the system, since the vaccine will, to some extent, stimulate the PIS. Be careful – Francisco was not against the classical immunology, which includes the vaccines. The concepts of autoimmunity, circular causality, etc., still support and justify the classical immunology, and there is room for the peripheral/reactive/immune IS. But what is possible is that these vaccines may have a destabilization effect on the immune system that could lead to the consequences we have seen, such as tromboses in case of AstraZeneca. The IS is, after all, a system, and it is perturbed by the vaccine.

3) Sebastjan: The classical approaches have a problem with the notion of horror autotoxicus, a mechanism that has been proposed to exist in the immune cells that prevents them from destroying the body. That inspired Francisco to put an emphasis on the importance for the IS to establish a boundary between the self and non-self.

Hugues: Any immunological textbook will start with these seemingly obvious things. The simulations, on the other hand, show that the IS does not care about what self and non-self are. They are not there a priori; instead, there is an unfolding system which can be perturbed, and, should that occur, it is possible then for the IS to perform badly, like attacking some healthy organs. But even in regular times, the IS is destroying healthy cells – however, in an extent and manner that do not harm the body. Francisco maintained that the autoimmunity is not a disease but a natural functioning of the IS. When the classical immunologists asked him what the role of the CIS was if not to defend the organism, his reply was that it maintains viability. For instance, it maintains all the cells in the body in a range of a homeostatic concentration. Homeostasis can be destroyed in a case of cancer, when there is an explosion in concentration of some cell(s), or in AIDS, etc. If maintaining the viability is the key role of the IS, then sometimes it has to reject an impact and sometimes not. If the impact does not violate the viability/homeostasis, the IS can accept and host it.

Viktorija Lipič and Toma Strle: The theory of biological expression of personal identity states that the immune self (i.e., the immune network) is defended by the IS. What are the characteristics of the immune self? What is its individuality? What is the treshold of activity that differentiates between what is in and what is out?

Hugues: You have to take into account the system that creates its own treshold. It is hard to figure how this treshold is defined. We know that the system is not of a stable, definitive shape – its identity is not stable. When I did the autopoiesis simulation, I found out that it contained the exact same kind of idea: a new chemical can be put into a cell and become a part of its network, while some other chemical will be rejected. When you have a bunch of cells, like in the IS, this principle found in a simple cell is just scaled up.

4) Primož Vidovič and others: What are we supposed to call a “cognitive system“? According to Varela’s criteria, we can call the IS a cognitive system, we can also call the nervous system a cognitive system, and I feel like that makes the organisms, such as vertebrates, a composite of multiple cognitive systems. Can we treat all these separate systems as cognitive systems of their own right, or are they really cognitive systems only in the context of the organism as a whole?

Adnan Sivić: Say you have different levels of autonomy, the IS being one of them, the neural system another – what would be the relation between them? Is there a hierarchy, and if so, is it an emergent structure? What does that mean? Is the organism as such a separate autonomous unit?

Hugues: Collaborating with Francisco, I mainly played with three kinds of software: the autopoietic cell, the IS and the neural networks. What they have in common is a lot of circular causality, autonomous activity and the capacity of the system to manage its interaction with the environment. When I saw code for the autopoiesis for the first time, I was impressed by the fact that the membrane was catalyzing the cell’s chemistry, and the same chemistry was creating the membrane. All these things could be defined as cognitive. I remember that we had a discussion on whether autopoiesis is a living system. You can beyond that and connect the IS with the nervous system: we know that a lot of connections exist between them – for example, the nervous system strongly influences the concentration of a number of hormones. So do these two systems comprise a bigger unity?

Francisco wanted the system to be able to create a top-down emergence, influencing the cells coming in by what was happening inside the system. In the code, you can get the impression that the system itself is, in fact, doing that – but that is a wrong vision. Yet still, despite being a reduction, a semantic compression, it is currently the best way to phrase and to explain the system’s behaviour.

Evan Thompson: Francisco and I tried to formulate exactly what the requirements for the emergence would be, either in a computational model or in a living system. The first criterion would be the existence of a network of interrelated components (as seen in the nervous system or the IS) with a global dynamic that arises out of non-linear interactions of the components. And this is where it gets tricky. The second feature would be some kind of way that the global pattern channels, constrains, shapes and limits the degrees of freedom of the local interactions – something like collective variables or order parameters in the dynamical systems theory. And then the question is whether this is just epistemological. If you talk about a global order parameter, you can form a strong argument, but that is really just for the convenience of you as a modeler. It makes your language and description more compact and enables you to circumnavigate the impossibility of calculating everything that is happening yourself. But really, all the causal action, strictly speaking, is local. In case of the code, that is the most plausible view. Now in the case of a living system, it gets much more complicated, because then you want to know whether all the interactions really are local, or is there (the third criterion we introduced for emergence) an irreducible relationality.

I pushed Francisco on this a lot. In cellular automata, you arguably do not have any of these, so how feasible is it that this kind of relational holism, this downward influence, exists in an actual cell or brain? And there, he was very open. He said, “I don’t know, but I’d like to think so [that it is possible].”

Hugues: Quantum mechanics can accomodate some form of holism; the same holds true for the mind-body problem – the term emergence has in fact surfaced with the philosophical problem of mind and body.

Observing birds in flight, physicists will try to understand how the shape of flock emerges out of the local interactions of birds. Biologists will aim for a story related to the natural selection (e.g., the birds are flying this way, using these rules, because it helps them to save on energy or to evade being an easy prey). They loop together environment and natural selection. These two are not part of physics in general but they could strenghten the notion of emergence. (Yet still they are subject to neglect, since the emergence is to some extent explainable by the classical science.)

5) Evan: Is the autopoiesis entirely computable? Robert Rosen’s metabolic closure is analog to computational systems. He claims that it can be shown mathematically that the metabolic closure is not entirely computable. He takes this to be an important clue for our further conceptualization of life.

Hugues: John Stewart was fond of Rosen’s work: he found it to be greatly supporting of the idea that the autopoietic systems are not computable.