Eric YES! Bear with me for a moment ;)

We have recently described a new form of inter-neuronal communication using pairs of neurons from the DRG. When the DRG were isolated by a gentle dissociation method (developed for the first calyx-synapse preparation Brain Research 1989/PNAS1990/JN 1991 to retain cell contacts) I observed that some of the neuron somata remained in pairs. Stimulation of one neuron with a train of impluses led to a delayed and long sustained response in its pair (EJN 2012). However, the two somata were found to be separated by a thin glial membrane. Detailed analysis demonstrated that transmission is via a signalling pathway where CaV2.3 channels are activated in the stimulated neuron triggering ATP release; this activates P2Y2 receptors on the glial membrane and glutamate release by Ca2+ released from intracellular stores onto both the stimulated and passive neurons (EJN 13a, b; JP 13). Thus, we demonstrate transglial communication between the neurons and have termed this a 'Sandwich Synapse'.

I am particularly excited about this not only for its implications for communication within the DRG and possible roles in pain etc but, far more important, for the possibility that SS type transmission occurs in the CNS and is common. In theory, SS transmission could occur between any two neuron somata or processes that are separated by a common glial cell. Such slow transmission can not reasonably contribute to acute responses but so much of brain activity involves processes that occur in exess of seconds time scales. In theory at least, SS transmission could represent a far more extensive mechanism of information sharing than that mediated by action potentials - but as a homeostatic, modulating, synchronizing and even metabolic communication.

We are currently seeking an experimental model in the CNS to test these ideas using the principles learned in the DRG. If anyone has a suggestion where to look - ideally with large neurons that are typically very close to each other - I would greatly appreciate it.

And thanks for raising this and giving me a soap box!

ee

Astrocytes modulate neurotransmission by absorbing, modifying and releasing glutamate. They can modulate calcium levels, which also impacts neurotransmitter release at the axon. They are not simply supportive in the sense of providing structure. Glia also produce neurotrophic factors, such as CNTF, IGF-1 and bFGF, which may in turn impact neurotransmission. Astroglia likely produce many more undiscovered proteins.

Eric YES! Bear with me for a moment ;)

We have recently described a new form of inter-neuronal communication using pairs of neurons from the DRG. When the DRG were isolated by a gentle dissociation method (developed for the first calyx-synapse preparation Brain Research 1989/PNAS1990/JN 1991 to retain cell contacts) I observed that some of the neuron somata remained in pairs. Stimulation of one neuron with a train of impluses led to a delayed and long sustained response in its pair (EJN 2012). However, the two somata were found to be separated by a thin glial membrane. Detailed analysis demonstrated that transmission is via a signalling pathway where CaV2.3 channels are activated in the stimulated neuron triggering ATP release; this activates P2Y2 receptors on the glial membrane and glutamate release by Ca2+ released from intracellular stores onto both the stimulated and passive neurons (EJN 13a, b; JP 13). Thus, we demonstrate transglial communication between the neurons and have termed this a 'Sandwich Synapse'.

I am particularly excited about this not only for its implications for communication within the DRG and possible roles in pain etc but, far more important, for the possibility that SS type transmission occurs in the CNS and is common. In theory, SS transmission could occur between any two neuron somata or processes that are separated by a common glial cell. Such slow transmission can not reasonably contribute to acute responses but so much of brain activity involves processes that occur in exess of seconds time scales. In theory at least, SS transmission could represent a far more extensive mechanism of information sharing than that mediated by action potentials - but as a homeostatic, modulating, synchronizing and even metabolic communication.

We are currently seeking an experimental model in the CNS to test these ideas using the principles learned in the DRG. If anyone has a suggestion where to look - ideally with large neurons that are typically very close to each other - I would greatly appreciate it.

And thanks for raising this and giving me a soap box!

ee

Thank you both, Alisa and Elise.

Elise, I'm also interested in exploring the contribution of glial cells in pain procesess. What do you know about that?

Wish I was an expert (I'm really a presynaptic calcium channel/SV release person). The SS may well contribute, in particular because in the DRG they use CaV3.2 channels (low-threshold calcium channels) of much interest for analgesics. However, there are many on here that are - I was just corresponding with Emmanuel Bourinet on RG. Perhaps you could ask him?

ee.

Astrocytes also impact neuronal function through release of kynurenic acid which influences the release of dopamine, glutamate and GABA, and also modulates the release of other neurotransmitters like ACh. Manipulation of brain kynurenic acid levels (as well as other kynurenines) have behavioral/functional consequences as well. Check out the excellent work of R. Schwarcz for more on this.

Thank you very much to each and everyone of you!

Now, from a engineering point of view, what would be the implications at the level of computational modeling of the nervous system?

that's whats so interesting. For the standard 'tripartite synapse' the basic influence would be parallel inputs that modulate the strength of synapses. Thus, neuronal connections remain the same but their strength varies. The SS provides a potentially astronomical complexity since, in theory, it would permit transmission between neurons that have no direct synaptic contacts, vastly increasing the signalling network. Indeed, it might be very difficult to generate a wiring diagram. On the other hand, from what we've seen thus far transglial transmission would be expected to at most alter general excitability, and perhaps not lead to action potential initiation by itself.

ee

Hoi Erick and participants,

it's a pleasure to join!

As the astrocytes unite by means of Gap Junctions (with Ca++ waves) they form a syncythial territory.

Would someone know how many neurons would be involved in one territory in order to have an estimation for a computational model of Erick?

From any region, cortex, subcortex, brainstem, spinal level? Perhaps from studies of the hippocampus?

Jean Pierre.

I would like to know more about Superficial Dorsal Horn targets from descending projections (my original question). But let me explain myself:

Lu & Perl (2007) illustrated the connectivity patterns between adrenergic and serotonergic projections from brainstem and superficial dorsal horn interneurons (islet, central, etc.). On the other hand, experimental evidence shows that 2/3 of RVM descending fibers are GABA/glycinergic (Aicher et al, 2012), whereas 1/5 are serotonergic (Moore, 1981). Besides, adrenergic neurons don't come from PAG-RVM system, but from others supraespinal structures, such as locus coeruleus (Purves et al, 2007).

Anyone could give me new insights about this issue?

Dear Erick,

exact efference from rVMM to specific interneurons of the spinal cord are difficult to find in literature, but I will call for help form researchersat the univ.of Calif. (Irvine), who do studies on the rostral medulla.

In the mean time, for your interest in a computational model I suppose the effect from astroglia in PAG & rVMM is also from interest?

-> chronic restraint stress provokes atrofy in astroglia in the vl.PAG (known for behavioural inhibition): facilitating aggression (1); and same in the rVMM (known for sensitization and neuropathia) (2).

I suppose the sandwich synapses of Elise would be disrupted under the condition of chronic restraint.

(1) Neuroscience. 2012 Oct 25;223:209-18. doi: 10.1016/j.neuroscience.2012.08.007. Chronic restraint stress decreases glial fibrillary acidic protein and glutamate transporter in the periaqueductal gray matter.

Imbe H1, Kimura A, Donishi T, Kaneoke Y.

(2) Neuroscience. 2013 Jun 25;241:10-21. doi: 10.1016/j.neuroscience.2013.03.008. Effects of restraint stress on glial activity in the rostral ventromedial medulla.

Imbe H1, Kimura A, Donishi T, Kaneoke Y.

Thank you again Jean!

Actually, I'm focused on neurons and their connectivity patterns. I hope to include glial contribution in computational modeling of pain as soon as possible.

and yes! It is very difficult to find related material...

The chronic restraint model is very interesting - relevant to casts and long-term hospital care etc. Of course we don't know if all glial functions are affected equally its possible that some (such as non-specific glial extensions) are more vulnerable than others. Thus, while its possible SS contacts would be disrupted it may not be so.

Thanks for that Jean - I was not aware of that model.

Descending central pain facilitation:

influence on the precise type of spinal interneuron is difficult to find, but following article (1) states:

rVMM: 5HT + -> 5HT3-R on spinal [neuron-glial-neuron complex]

with increased glial reactivity & neuronal hypersensitivity.

Sounds close to the sandwich concept within spinal medulla.

And usefull for computational model.

There is a review article available on gliopathy in spinal cord above and under spinal cord injury level (SCI model) (2).

(1) J Neurosci. 2011 Sep 7;31(36):12823-36. doi:10.1523/JNEUROSCI.1564-11.2011.

Spinal 5-HT(3) receptor activation induces behavioral hypersensitivity via a neuronal-glial-neuronal signaling cascade.Gu M1, Miyoshi K, Dubner R, Guo W, Zou S, Ren K, Noguchi K, Wei F.

(2) Exp Neurol. 2012 Apr;234(2):362-72. doi: 10.1016/j.expneurol.2011.10.010. Spatial and temporal activation of spinal glial cells: role of gliopathy in central neuropathic pain following spinal cord injury in rats. Gwak YS1, Kang J, Unabia GC, Hulsebosch CE.

Dear Jean,

The first article was retracted (here is the link: http://www.jneurosci.org/content/31/36/12823.full.pdf+html)

I don't really know what does it mean...

I'm sorry, I think we deviated from the original topic (glial contribution to computational capabilities fo the nervous system).

Jean (and everyone), I think we can go back to the previous topic to continue our discussion of descending nociceptive control...

Thank you very much!

OK Erick,

we are looking for computational elements of neurons, interneurons and astrocytes (and perhaps also microglia) in the dorsal horn with particular interest of descending modulation with pain facilitation and -inhibition. The pain facilitation increases the probability of sensitization, allodynia, chronicity and neuropathia. Whereas the pain inhibition refers to central analgesia, gate theory and diffuse noxious inhibitory control. The balance between descending facilitation and inhibition is called (contextual) prioritization.

Inducing astroglia into the computational model is, to my knowledge, new. Your computation of this would be amongst the first, I believe:

- astroglia in this computational model can be inhbitory (ATP) and excitatory (Glu);

- inhibiton and excitation refer here to a syncythial territory of wich article (2) gives us an idea as to the extension: it could cover easily three segmental levels;

- this synchronization depends on frequency and intensity of Ca-wave oscillation within the territory.

Study literature on satelite cells in the DRG, you will be amazed.

I'm a little late to the party, but wanted to mention that we have a paper in press that describes a neuron-astrocyte-neuron retrograde circuit in the rat hypothalamus, not unlike the sandwich synapse described by Elise Stanley (above), but different. We find that neuroendocrine (vasopressin) neurons in the hypothalamus communicate with astrocytes via the dendritic release of vasopressin, which elicits a calcium response in the astrocytes. The astrocytes then release ATP onto GABA neurons that at a distance from the vasopressin neurons and are presynaptic to the vasopressin neurons. The glial ATP release causes a P2X receptor-mediated excitation of the GABA neurons and the activation of a GABAergic projection back to the vasopressin neurons, which happens to be excitatory in these cells due to unique Cl- homeostasis. This represents an interesting retrograde neuronal-glial circuit in which the astrocyte is integrated into the circuit and transmits information to upstream neurons. We have not yet begun to address the electrotonic coupling of the astrocytes and the spatial domains that this circuit may cover, but the presynaptic GABA neurons are quite distant from the postsynaptic vasopressin neurons, being located, we think, outside the nucleus that holds the vasopressin neurons (i.e., the paraventricular nucleus).

You are not late Jeffrey!

The dynamic you described resembles a feedback-loop control system. Let me take some time to rewrite that into a block diagram, and then we could discuss about it.

Is your paper already published?

Jeffrey - that is very interesting - a kind of open-faced-sandwich-synapse :D If you have any ideas where we might find the tight ones that are similar to those in the DRG we'd love to know.

Erick - yes, one might expect a positive feedback loop, since the return circuit activates the vasopressin neurons, which would be expected to cause further dendritic release of vasopressin and restart the cascade of signals. We have not yet tested this. The paper is not out yet, but we just returned the proofs, so it should come out electronically in the next couple of weeks in Journal of Neuroscience.

Elise - I prefer to think of it as a trans-neuronal-glial circuit, with the glia integrated as full players in the circuit. We think the signaling from the glia to the upstream GABA neurons occurs at the GABA cell bodies, since the increase in GABA inputs to the vasopressin neurons is blocked by TTX. Unfortunately, I don't know of any evidence for a sandwich-type synapse that you describe, but there is a lot of neuronal-glial plasticity in the system, with glial processes retracting from between neurons under certain conditions and allowing membranes of adjacent neurons to come together in direct juxtaposition. Also, there is evidence for taurine release from astrocytes onto vasopressin and oxytocin neurons (Deleuze C, Duvoid A & Hussy N (1998). Properties and glial origin of osmotic-dependent release of taurine from the rat supraoptic nucleus. J Physiol 507, 463–471.)

Thank you Enrico, I was indeed impressed by the satellite glia cells in hte dorsal ganglion. I think it confirms what Elisa told us.

May I ask Jeffrey his opinion about the role of cortical astroglia connecting different parts of the cortex and their significance in conscious feeling put forward by Pereira-Furlan (in comparison to the previous pure neuronal concept).

For Erick: I found 4 figures of astroglia connections that might be interesting to convert in a computational model? It is from the same authors (Pereira-Furlan).

-> http://www.msmonographs.org/article.asp?issn=0973-1229;year=2011;volume=9;issue=1;spage=183;epage=192;aulast=Pereira

Jeffrey - thanks for the perspective. What I've noted however, is that if you look at just about any study where astroglia have been identified by EM in the CNS you can find examples that are at least consistent with SS-type transmission: two neuronal processes that are separated by a common glial cell. This is tantalizing - but what we need are functional studies, testing if stimulation of one neuron can lead to delayed responses in a neighbour. Of course, it then has to be shown that this is not via a neuron-neuron(-neuron...) pathway.

The astrocentric hypothesis described in the Pereira-Furlan paper is very interesting with respect to neural networking and whole-brain activity in complex cognitive processes. I have also come across a paper modeling astrocytic optimization of neural processing at the synaptic level that may be helpful: Astrocytes Optimize the Synaptic Transmission of Information (Nadkarni, Jung and Levine)

http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1000088

Dear all, look into my chapter on peripheral nervous system topics, in The human nervous system Mai and Paxinos, part on the DRG functions and find the articles of Chris de Zeeuw who showed that a group of neurons surrounded by glia can behave different of the next group, although the same input is given.

Enrico - that is sensational! Will message...

Jeffrey, If you're right, we must expect a negative feedback loop that compensates the positive one, thereby stabilizing the process you described (just like voltage gated sodium channels and voltage gated potasium channels). On the other hand, we could have something alike to long term potentiation, which is involved in memory and pain processes, according to experimental evidence.

Jean, thanks for the pictures. Please, give me some time to take a look...

Dear all, we have done some experimental and modeling studies about the role of astrocytes in epilepsy, you might find the following journal papers useful:

Amiri, M., Bahrami F., JanAhmadi M., "Modified thalamocortical model: A step towards more understanding of the functional contribution of astrocytes to epilepsy," J Comput Neurosci (2012) 33:285-299, DOI 10.1007/s10827-012-0386-8.

Amiri M., Hosseinmardi N., Bahrami F., JanAhmadi M., "Astrocyte-neuron interaction as a mechanism responsible for generation of neural synchrony: a study based on modeling and experiments," J. Comput Neurosci (2013) 34:489-504, DOI 10.1007/s10827-012-0432-6.

The role of glia in the processes of intercellular communication unchallenged. Glia plays a role in depositing neurotransmitters in the synaptic cleft. Glia need to "neutralize" excess neurotransmitters. For example, for capture of glutamate. If this process is disrupted, it is close to the overexcitation and toxicity in the brain. Glia need for efficiency of the diffuse signal transmission in the nervous system. So, glia be taken into account in models of interneuron communication, brain plasticity, pathological processes in the brain (epilepsy).

I wonder if this gives us a mechanism for dementia, if some insult reduces the population of glial cells, can toxicity in the brain be increased to the point where neuron cell death follows?

Thank you all for your contributions!

Now, Would be reasonable talk about the computational capabilities of glial cells? Or should we consider these cells only as gain blocks?

Dear Erick! Firstly, glial cells varied. Glia differ in structure and function. Secondly, it is necessary to take into account the known (and little known) functionality glia. Thirdly, it is important to choose the functions of glia. For example, as participation of glial cells in the brain signaling. Fourth, by participation in the signaling function to understand the specific mechanisms rather than general words. Indeed, in the transmission of information in the brain is involved also the circulatory system. If there is no blood flow, it will not transmit signals.

Ok Vladimir, let´s talk about astrocytes.

Are they merely gain blocks? Are they capable of information processing?

Erick, although I'm not a modeler, I think it's going to be challenging to incorporate glia into a computational model at this point, as alluded to by Vladimir, since glia are heterogeneous and even the same types of glial cells are mediating mutliple functions, including classical housekeeping functions that we've known about for a while, but also the signaling functions of the different types of glia that are increasingly being discovered. Also, our state of knowledge of the signaling functions of glia is still pretty rudimentary, and not very quantitative, so that it would be difficult to know what parameters to enter into your model.

That being said, I'm all for trying to develop a preliminary model that integrates the influence of glia into neuronal network signaling. In fact, we're working with some modelers on incorporating retrograde synaptic signaling into a bursting model of neuroendocrine cells.

Erick, yes. Now there is strong evidence for the role of astrocytes in the analysis of information in the brain. Let me pick up for you for a few articles in which the authors discuss the specific mechanisms for the participation of astrocytes in the transmission of information.

I think the problem is that most of the information is still at the 'glia can do this under these circumstances' and not 'under these circumstances glia will do this'. Thus, it is currently difficult to create a meaningful model - one that is rigorous, predictive and (most important) testable.

Jeffrey and Elise, do you mean that mechanisms underlying glial function are still uncovered? If this is so, we can formulate some hypothesis through a preliminar "glial-cell-model", such that we could eventually test them. After all, this is one of the purposes for which computational models are formulated.

Vladimir, thank you so much. Please, let me know when you get the articles.

Dear Eric! Let me send you two links of splendid articles that are available in PubMed free. Just a few references to the work of scientists in the field of interaction of astrocytes and neurons.

Habas A., Hahn J., Wang X., Margeta M.Neuronal activity regulates astrocytic Nrf2 signaling. Proc. Natl. Acad. Sci. U. S. A. 2013. Vol. 110 (45). 18291-6. doi: 10.1073/pnas.1208764110.

Bosier B., Bellocchio L., Metna-Laurent M., Soria-Gomez E., Matias I., Hebert-Chatelain E., Cannich A., Maitre M., Leste-Lasserre T., Cardinal P., Mendizabal-Zubiaga J., Canduela M.J., Reguero L., Hermans E., Grandes P., Cota D., Marsicano G. Astroglial CB1 cannabinoid receptors regulate leptin signaling in mouse brain astrocytes. Mol. Metab. 2013. Vol. 2 (4). P. 393-404. doi: 10.1016/j.molmet.2013.08.001. eCollection 2013.

And also, just a few references to the work of scientists in the field of interaction of astrocytes and neurons.

Cárdenas A., Kong M., Alvarez A., Maldonado H., Leyton L. Signaling pathways involved in neuron-astrocyte adhesion and migration. Curr. Mol. Med. 2014. Vol. 14 (2). P. 275-290.

Arizono M., Bannai H., Mikoshiba K. Imaging mGluR5 dynamics in astrocytes using quantum dots. Curr. Protoc. Neurosci. 2014. Vol. 66:Unit 2.21.. doi: 10.1002/0471142301.ns0221s66.

Di S., Popescu I.R., Tasker J.G. Glial control of endocannabinoid heterosynaptic modulation in hypothalamic magnocellular neuroendocrine cells. J. Neurosci. 2013. Vol. 33 (46):18331-42. doi: 10.1523/JNEUROSCI.2971-12.2013.

Dear Erick! Sorry for spelling your name

Vladimir, don't worry about the "misspelling"

Thank you so much for the links. Please, give me a couple of days to check them out..

Best regards!

The paper that I referred to earlier in which glia play an integral part in a retrograde circuit that signals to upstream neurons just came out. I apologize for the self promotion, but I think this form of trans-neuronal-glial communication will prove important when considering an integrated model of neuronal-glial signaling.

Haam, J. Halmos, K.C. and Tasker, J.G. (2014) Nutritional State-Dependent Ghrelin Activation of Vasopressin Neurons via Retrograde Trans-Neuronal-Glial Stimulation of Excitatory GABA Circuits. Journal of Neuroscience 34: 6201-6213; doi: 10.1523/JNEUROSCI.3178-13.2014.

That's great Jeffrey!

Is the paper available?

Erick, I don't think I'm allowed to post it for 6 months, but I can send it to you if you send me your email.

Let me offer a complementary perspective. The (spatial) buffering by astrocytes is not only a regulation mechanism of extracellular potassium concentration but can also lead to new type of excitability on very slow times scales (near DC in EEG terms), hence glial cells are an integral part of some new sort of function.

In my opinion, information processing in the brain is far too much attached to neurons firing action potential, that is, to electrical membrane potential and to the physical picture of equivalent electrical circuits. There glial cells do not find their proper place.

We recently suggest in the new manuscript on arxiv (http://arxiv.org/abs/1404.3031) that with regard to the new excitability another picture must be adopted, namely that of thermodynamic cycles in which the glial cells play an important functional role even though they are merely considered the systems, ie, neurons, surroundings.

See also:

http://www.scilogs.com/gray-matters/when-neurons-let-off-steam/

Thank you Markus. That's exactly what we need: several complementary approaches!

Jeffrey, thank you very much. You have my word: I won't reveal any detail about your article...

My e-mail:

[email protected]

NO, it is wrong

It is assumption for justifying large EU and US funding for connectomics and multineuron computational models. In reality neurons form the network were the excitation moved along pre-determined routes. Astrocytic network embedded into neuronal regulates signal propagation along those routes through release of gliotransmitters. Although actual neuronal network remains the same, spatio-temporal pattern of gliotransmitters (aka guiding template) creates large number of functionally different networks. These functionally different networks will process the signal differently. Thus astrocytes allow to increase computational capacity of the brain may fold. I would speculate this is the reason why astrocytes, but not neurons become more complex in humans.

Here is slightly primitive description of this idea

http://www.ncbi.nlm.nih.gov/pubmed/17881089

But is was published several years ago. Not we have a lot more supportive evidence.

Alexey: our recent description (detailed above and near the beginning of this question topic) of a 'Sandwich synapse', where the glial cell acts in series between two neurons shows that glial cells can, at least in principle, at as part of the signalling network (in series), and not just as a modulator (in parallel). If this is the case in the central nervous system (the work was done in the DRG) then we may have to reconsider the entire cellular basis of information coding and processing.

Sorry to intervene, this discussion is very interesting and meaningful, but also funny (especially the down vote that Alexey received is hilarious!). Glia cells, astrocytes, astroglia. Then there are the activated ones, and now we see them involved in synapse sandwiches. We are scraping the surface. It all reminds me of the history of "junk DNA". If we cannot explain it, we discard it as being trivial. Now we know better that Junk DNA does not exist (in biology we learn that redundancy is unsustainable, in other words, what is useless will be dropped). The same is true with all non-neuronal brain cells. We call them glia (keeping neurons together like glue) out of arrogance, just like we used to call non-coding DNA junk. I dare say that there are many different types of non-neuronal cells in the CNS, each type having a different function. Pretty much like we have different neuronal cells with different functions. It is way too early to build a computational model until we better understand the biology of the many types of non-neuronal cell in the CNS and their functions.

And do we have complete knowledge of all the different types and functions of neuronal cells yet?

I do not understand the down vote that Alexey receive too!

We need to take into account every angle, point of view, or aspect that we did not before. Different does not mean wrong...

[I just gave him a plus to counteract it, perhaps others could follow suit?]

Jan, it is not too early for a preliminary model, we just need an idea of how the glia might work. The whole idea of a model is to give us an idea to test, to see if we understand. This question suggests a model whereby glia process information as well as transfer it from one neuron to another. Whether that is a good model or not is the topic of Alexey's post. We have many models of neurons and neural networks, long before we have an idea of all the functions they can perform.

I know I am being provocative and my suggestions are open for criticism, but major point which I try to make is that our brain is far more complex than we think. A couple days ago I had an argument with my friend who is a PI in RIKEN BSI. He thinks that we are already know major principles of the brain operation and just need to put it all together. This is rather technical issue and neuroscience is no longer interesting.

In reality we know quite a bit about neuronal network, about integration of synaptic inputs in neurons and subcellular processing of these inputs. But we do not know much about operation of astrocytic network-syncytium, subcellular information processing in astrocytes. Neuronal and astrocytic networks co-exist in the brain and interact. This interaction is largely extrasynaptic and poorly understood. And there is nothing wrong in the term "modulation". Long-term changes in efficiency of extrasynaptic "modulation" may be equally good way for information/memory storage in the brain as synaptic plasticity (LTP or LTD).

We are trying to understand these phenomena.

Here is the paper where we show that signal processing in astrocytes fundamentally different from neuronal integration of synaptic inputs:

http://www.ncbi.nlm.nih.gov/pubmed/24484772

This suggests that interaction of two cellular networks with different computational properties can give very complex operational output.

Here is another paper where we show that extrasynaptic signalling/modulation is also plastic process:

http://www.ncbi.nlm.nih.gov/pubmed/22832274

What if it is another cellular mechanism for learning and memory.

So I believe that we are just at the beginning of our journey, and neuroscience is still interesting. But we have to be prepared for the shift of paradigms.

Dear Alexey! It is true. Scientists discover new functions of glial cells. For example, it is clear that glial cells play an important role in the plasticity of the nervous tissue. And, therefore, glia needs to transmit information in the brain, memory processes. Number of glial cells in the brain exceeds the number of neurons. But if we imagine the population of all cells in the brain in the form of a symphony orchestra, the neuron acts as a soloist. Other musicians (glia) play along with the soloist. Each has its own role. Ultimately, it turns harmonious melody. So it is in the brain. So it is in life.

Dear Erick,

My answer will attempt to address some of the multiple computational facets of your question.

The very notion of "processing" is connoted and might convey the wrong ideas about what neurones, axons, and glial cells actually do in the big picture that slowly emerges from all the detailed studies of specific areas of the nervous systems of many different species over the last few decades.

I believe there has been a big hype about neurones that started 75 years ago, that has slowly shifted toward the synapses and the neurotransmitters, and marginally the axons. (Don't get me wrong: I do not discount Cujal's, Helmholtz's, and earlier scientists work, often carried out in the context of negative social pressures! This arbitrary point in History will be made clearer further.) As you noted, there is a renewed interest for the glial cells (over the last decade, while they had been treated like junk before - as Jan noted), especially the astrocytes, and more specifically for the role they play in synapse plasticity in mature organisms. Another recent paper that was forwarded to me shows that the "good old" action potential may not be the only kind of regenerative event that exists in neurones.

My reading of the related articles make me think there are currently three main computational approaches to understanding animal nervous systems, and brains.

On one side, the modelling driven by biological observations, that may be broadly classified as "computational neuroscience". There are PhD theses in this domain that have attempted to tackle other elements along with the neurones. They show that our mathematical tools and computing resources are challenged by demands of these models.

The second approach tries to abstract the details and make the computational pressure more tractable. The main objection is a question: what will we learn from models that may be "non biologically plausible"?

The third approach is the 75y old "AI" dream. I mention it because I consider that the seminal paper that McCulloch and Pitts published in 1943 is the source of many issues, including that - to this day - all of AI's neural nets produced so far are based on the premises that the propagation between the neurones of different layers is instantaneous, ditto 1943. I think this is a reasonable argument to explain why cognitive science, computational neurosciences, ethology, etc. researchers show some level of reservations when comp science/AI people stick their neck out and attempt to contribute.

I can see the merits in the first two approaches, however I think these are not the only avenues that could be explored today.

There's definitely a colossal body of knowledge based on observations of how bits of various species' nervous system behaved when stimulated. At the same time, it seems we are currently stuck in an equation that looks like: knowledge of the connectome x knowledge of the individual characteristics of the components = epsilon. I doubt we're yet to the level where Heisenberg's uncertainty principle applies, therefore the situation shall improve steadily. To me it looks like the pieces of a N-dimensional puzzle, with N unknown, with a lot of missing pieces indeed, and with very few clues about how everything fits together. This is a hell of a challenge! Therefore wherever answers come from, they surely will be welcome.

I am of the view that the vast collection of observations collected about the nervous system of various species can be used to attempt to reverse-engineer the underlying macro-mechanisms of specific behavioural responses, typically those that are shared amongst animal species. Such approach could pinpoint to regions of interest in the familiar neural pathways extensively analysed in the existing literature. I also expect that these predictions will be easy to confirm/ reject/ refine with in-vivo experiments. That is my answer to the "non biologically plausible" objection: if a remotely biologically plausible model can predict something that was overlooked before, it must have some benefits. Hopefully, I shall soon be able to start a PhD along this direction and attempt to shed some light on short-term memory... and adapt my beliefs as I make progress. One has to start somewhere.

Regards

Herve

First we need to address a better understanding of neuro-glia interactions in the particular brain area/model. After experimental identification of the major players shaping tripartit synapse in the area selected, we may want to model it.

Very interesting work about involvement of glia cells in neuronal transmission is done here in Dunedin.

http://www.ncbi.nlm.nih.gov/pubmed/24298150