A few days later, Henry Markram, leader of the Blue Brain Project at EPFL, sent off an e-mail to IBM CTO Bernard Meyerson harshly criticizing the IBM press release, and cc’ed several reporters. This brought a spate of shock media into the usually placid arena of computational neuroscience reporting, with headlines such as “IBM’s cat-brain sim a ‘scam,’ says Swiss boffin: Neuroscientist hairs on end”, and “Meow! IBM cat brain simulation dissed as ‘hoax’ by rival scientist”.  One reporter chose to highlight the rivalry as cat versus rat, using the different animal model choice of the two researchers as a theme.  Since then, additional criticisms from Markram have appeared online.

Find out more after the jump.

In the aftermath, IBM has stood behind the announcement, citing for Network World their team’s involvement with “Stanford University, University of Wisconsin-Madison, Cornell University, Columbia University Medical Center, University of California-Merced and Lawrence Berkeley National Laboratory” as defense.  Who are the researchers they are standing behind?  According to Modha’s blog, they are:

• Stanford University: Brian A. Wandell (Prof of Psychology, Electrical Engineering), H.-S. Philip Wong (Prof of Electrical Engineering)
• Cornell University: Rajit Manohar (Prof of Electrical Engineering)
• Columbia University Medical Center: Stefano Fusi (Prof of Theoretical Neuroscience)
• University of Wisconsin-Madison: Giulio Tononi (Prof of Psychiatry)
• University of California-Merced: Christopher Kello (Prof of Cognitive Science)

For this neurodude, it is interesting how this disagreement may be symbolic of the gap that still remains between neuroscience and AI.  Markram is a neuroscientist turned technologist, while Modha is a computer engineer who wants to derive technological insight from biological  systems.  They are approaching the ideal of reverse engineering the brain from very different perspectives, and its only natural that they value different milestones.  The IBM team, even with the additional professors on their team, still lacks mainstream neuroscientists to help validate their claims.  That being said, the public realization of this could be a positive thing for both fields.  Although some frustration has resulted from this, this could be a great opportunity for the breakdown of walls between these fields.

In the end though, it does seem like Markram has a point.  The IBM press release clearly went too far.  Whether the angry public e-mail was the best strategic way to make the point remains to be seen.  It will be interesting to see what the next move from the IBM team will look like.

Watch it online at the TED website.

Two approaches

Broadly, there are two approaches to synaptic tagging: 1) mRNA is distributed widely and translated locally at tagged locations; 2) protein products are distributed widely in the bodies of dendrites but only contact/off-load at tagged synaptic specializations. This News & Views gives a nice overview of the two papers, which find approach 1) in Aplysia cultures with sensorin mRNA and approach 2) in rat hippocampal neurons with Vesl-1S/Homer-1a protein. It amazes me that both were found pretty much simultaneously, but that might have more to do with the use of the photoconvertible Dendra2 protein than anything else.

With both approaches, we still don’t know why mRNA/protein is directed to a certain location. That is, we can visualize synaptic tagging but we don’t know what is the tag, its ontogeny, or the mechanism of tagging. But that might not be so important to understanding more about neural function. These new tools might allow us to image plasticity at many synapses at once, perhaps even in vivo. But before that, more work is needed… does the optical signal (from the Dendra fusion protein) correlate with degree of potentiation? Can we detect plasticity in the opposite direction, ie. synaptic depression, through another tag?  (As a sidenote to approach 1), the use of 5′ and 3′ UTRs as a sort of molecular zipcode is also intriguing.)

Click on for some of my favorite passages from the Abbott piece.

Abbott uses the following problem of input decoding

Spike counts and neuronal firing rates are positive quantities. This simple fact has important implications for neural coding and neural circuits that provide a framework for thinking about a number of research directions taken over the past 20 years.

to highlight new work in synchrony, dendritic compartments, and balanced excitation-inhibition. This is probably the best part of the whole article. With some simple arithmetic, he motivates and explains solutions to the problem of correlating neural activity with real events.

The successes of circuit models (and principles of circuit models) in primary visual cortex:

We now have plausible mechanisms for how simple and complex cells obtain their basic response characteristics. Although no single consensus abouthow the circuits of primary visual cortex operate has arisen from this body of work, this may simply reflect the fact that multiple mechanisms contribute. In other words, many of these ideas are probably correct in one way or another, and the wealth of ideas in this field should be viewed as a success. Circuit-level modeling is now advancing beyond primary sensory areas (for example, Cadieu etal., 2007) and to the consideration of phenomena such as working memory through sustained activity (Amit and Brunel, 1997,Compte etal., 2000,Seung etal., 2000) and decision making (Wang, 2002,Machens etal., 2005).

And the dangers of an unhealthy obsession with connectomics:

What can we learn from the complete connectome or, indeed, a complete mathematical description of a complex artificial network model?

First, what can’t we learn? It is unlikely, for example, that we could deduce the task that the network was constructed to perform even if we were given the complete equations and connections of the model. If, along with this information, we were told what this task was, it is unlikely that we could figure out how the network performs it. If we somehow managed to make any progress along these lines, the people who constructed the network could probably provide us with another one that performs the same task but has a different connectome. In a similarway, biological systems may operate in a more variable manner than we have suspected, as has been stressed by Eve Marder (Marder etal., 2007). These issues are particularly true of a class of network models known as liquid state or echostate networks (Maass etal., 2002,Jaeger, 2003). In these models, the vast majority of interneuronal connections are not directly related to the task being performed (they are typically chosen randomly and left unchanged), the exceptions being synapses onto the output units of the network. Nevertheless, the tuned values of the synapses onto the output units can only be understood through their relationships to the random synapses. Such systems represent enormous challenges for conventional anatomical and physiological approaches.

The fact that the connectome of an artificial neural network does not typically tell us what the network does or how it does it should not be taken as an indication that this information is useless. Far from it. But we must be willing to be more abstract in our thinking. The important issue for an artificial network is not how it works but how it was constructed, which means what training procedures and modification rules were used to get it to perform a task. Although this information is not provided directly by the connectome, much can be inferred. For example, it is important to know whether the network has a feedforward architecture or has strong feedback loops. Other features of the network layout, whether it has hubs or bottlenecks, how many layers it contains, and its degree of heterogeneity, provide important clues as well. Obtaining a high-resolution connectome in neuroscience will be of great value, but artificial neural networks provide a cautionary tale that reminds us that scientific revolutions tend to render uninteresting as many questions as they answer. We will be fortunate if the connectome project does this for neuroscience, butas we launch ourselves into it we should appreciate that, as artificial neural networks appear to suggest, we may be asking the wrong questions.

Finally a major challenge for the future:

This is where I think the future lies in theoretical investigations of cognitive function. We must learn how to build models that construct hypotheses through their internally generated activity while remaining sensitive to the constraints provided by externally generated sensory evidence.

Halorhodopsin article from Deisseroth’s lab:
Multimodal fast optical interrogation of neural circuitry [News & Views]

Also, there is an intriguing article on both the general excitement in the neuroscience community with this new technology and a possible intellectual property dispute over it.

The paper can be found Here.

Very interesting work. Modulation of the somatic potential seems to influence the EPSP, as measured by paired patch recordings of two layer 5 cells in cortical slice. Somatic depolarization from resting potential to near threshold results in an increase in evoked EPSPs.

In synaptic physiology, we often make a point of distinguishing intrinsic changes (eg. membrane potential) from synaptic conductance changes. Now it looks like the line between those might be a bit blurry!

Here’s a N&V by Eve Marder too.

Some beautiful work [original article] by Oliet’s lab in a recent issue of Cell demonstrates the importance of glia in synaptic plasticity. The show a system where D-serine and not glycine controls the NMDA receptor in a coagonist role (or perhaps glutamate is really the coagonist…) and show how similar pairing protocols can have opposite effects (LTD vs. LTP) depending on D-serine modulation by astrocytes. Yet more hidden factors in plasticity are being revealed!

Here’s the key figure:

More details from the News & Views summary after the jump.

From the N&V summary cited above:

A growing body of work, including a beautiful study in this issue of Cell (Panatier et al., 2006), argues for a revision of our old textbook ideas about glial cells, NMDA receptors, and D-serine as a link between them. Fluorescence imaging studies from a number of groups indicate that astrocytes respond to neurotransmitter not with action potentials, like most neurons, but with propagating waves of intracellular calcium ions thought to elicit release of “gliotransmitters” such as ATP, glutamate, and D-serine (Haydon, 2001 and Mothet et al., 2005). In the meantime, mounting evidence indicates that the NMDA receptor’s glycine site often is not saturated under physiological conditions (Danysz and Parsons, 1998), and that, at some synapses, the endogenous ligand for the glycine site is actually D-serine (Schell et al., 1995). A report that astrocytes release D-serine when stimulated by glutamate (Mothet et al., 2005) lends support to the hypothesis that glia may actively modulate NMDA receptor-mediated transmission (Schell et al., 1995) and, consequently, synaptic plasticity.

In their new study, Panatier et al. (2006) address this idea in brain slices with electrophysiological recordings from neurons in the supraoptic nucleus (SON) of the rat hypothalamus. This neuroendocrine region controls numerous homeostatic systems throughout the body. When the hormone oxytocin is released in the SON during lactation, there is a reduction in the extent to which astrocytic processes surround excitatory synapses (Oliet et al., 2001). In earlier studies, Oliet and colleagues examined this system from a more neurocentric perspective. They showed that reduced glial coverage during lactation permits greater neurotransmitter diffusion beyond the immediate vicinity of the synapse, enhancing glutamate crosstalk between synapses and modulation of transmission by neuronal metabotropic glutamate receptors (Oliet et al., 2001 and Piet et al., 2004). In the present study, they show that glia do more than simply provide background for the neuronal melody—astrocytes actually conduct the orchestra, using D-serine as a baton.

For those interested in neuromodulators:

Treatment of striatal neurons with a D1 receptor agonist led to an increase in the dendritic staining intensity of NMDA receptor NR2B subunits. There was also an increase in the association of NR2B subunits with PSD-95 — a scaffold protein required for the assembly of NMDA receptors — and in the surface localization of NR2B-containing receptors.

Original article in J. Neurosci. from Dunah and colleagues. An excerpt from the original aricle of a neat application of FRET continues after the jump.

FRET was used to detect where the phosphorylated NR2B was in pervanadate-treated (pervanadate is a non-specific tyrosine phophatase inhibitor… ie. it prevents dephosphorylation) and normal striatal culture:

To confirm the localization of tyrosine-phosphorylated NR2B subunit in intact cells, we applied FLIM, a fluorescence resonance energy transfer (FRET)-based approach to detect protein–protein interactions (Fig. 8A, quantified in B). Untreated and pervanadate-treated striatal neurons at DIV21 were double immunostained under permeabilizing conditions for the NR2B subunit using an antibody to the C terminus (Wang et al., 1995Go) and for phosphotyrosine proteins with the anti-phosphotyrosine antibody; the donor fluorophore (NR2B) was labeled with Alexa Fluor 488 and the acceptor fluorophore (phosphotyrosine proteins) with Cy3. Proximity of the two fluorophores results in resonance energy transfer and shortens the fluorescence lifetime of the donor. We measured the donor lifetime in each image pixel using a pulsed laser and dual-photon confocal system. For negative controls, cells stained with only the donor fluorophore (Alexa Fluor 488) were used.

Using this approach, we detected robust staining for NR2B in both the soma and dendrites of neurons. In the untreated cells, only a small proportion of pixels exhibited a shortened donor lifetime, and these were concentrated along the dendritic regions (Fig. 8Ad,B), indicating the presence of tyrosine-phosphorylated NR2B. Treatment of striatal neurons with the phosphatase inhibitor pervanadate markedly enhanced the number of pixels with a short fluorescent lifetime, indicating a large increase in the extent of tyrosine phosphorylation of the carboxy tail of NR2B. Quantitative analysis of the signals in both the soma and the dendrites revealed that tyrosine-phosphorylated NR2B was increased in both compartments, but the effect was more than six times greater in the dendrites than in the soma (Fig. 8Af,B). These data demonstrate a strong association between tyrosine phosphorylation of NR2B and dendritic localization of the receptor protein.

Dan Feldman’s group at UCSD has found that different “sides” of STDP (ie. LTP vs. LTD) at cortical synapses might be mediated through distinct signalling pathways. The major finding was that LTD was induced independent of NMDA receptors. Rather, LTD required mGluRs and VGCCs.

There are many questions here. The most interesting to think about is, Are we going to find different STDP rules all over the brain? And, if so, what will be the commond ground between them?

Here’s the abstract:

Many cortical synapses exhibit spike timing-dependent plasticity (STDP) in which the precise timing of presynaptic and postsynaptic spikes induces synaptic strengthening [long-term potentiation (LTP)] or weakening [long-term depression (LTD)]. Standard models posit a single, postsynaptic, NMDA receptor-based coincidence detector for LTP and LTD components of STDP. We show instead that STDP at layer 4 to layer 2/3 synapses in somatosensory (S1) cortex involves separate calcium sources and coincidence detection mechanisms for LTP and LTD. LTP showed classical NMDA receptor dependence. LTD was independent of postsynaptic NMDA receptors and instead required group I metabotropic glutamate receptors and calcium from voltage-sensitive channels and IP3 receptor-gated stores. Downstream of postsynaptic calcium, LTD required retrograde endocannabinoid signaling, leading to presynaptic LTD expression, and also required activation of apparently presynaptic NMDA receptors. These LTP and LTD mechanisms detected firing coincidence on ~25 and ~125 ms time scales, respectively, and combined to implement the overall STDP rule. These findings indicate that STDP is not a unitary process and suggest that endocannabinoid-dependent LTD may be relevant to cortical map plasticity.