Submited on: 25 Mar 2012 11:14:56 PM GMT
Published on: 26 Mar 2012 12:36:34 PM GMT
update June 15th 2012
Posted by Dr. John Smythies on 15 Jun 2012 05:31:33 PM GMT

The authors of this paper have recently signed a contract with the book publishers Elsevier to produce, edit and contribute to a book on the claustrum. This is the first book ever devoted to this subject. Its title will be "The Functional Neuroanatomy of the Claustrum". Included will be a chapter on hypotheses to explain the mechanism of action of the claustrum. This is devoted mainly to a new hypothesis that we have developed. A previous version of this hypothesis was presented in our WMC paper. The other chapters are written by leading experts in the field of claustral neuroscience.

Readers are advised that the scientific material in our WMC paper is now superceeded by our new hypothesis, but it retains historical value. They are advised to read the new book, when it comes out, to keep in touch with the latest developments in this fascinating field.

Update May 22, 2012
Posted by Dr. John Smythies on 22 May 2012 08:41:36 PM GMT


The publication of this paper two months ago started a vigorous debate in the claustrum research community. This has led us to update and refine our hypothesis. We summarize these new developments as follows:


When an incoming sensory volley reaches the brain, there are two immediate tasks the brain must perform. The first (A) is to determine whether the information contained in the volley matches the brain’s expectation of what that information should be. The second (B) is to find out if the stimulus signals a state of affairs that is potentially threatening or rewarding. Task (A) needs to be carried-out first because a mismatch may necessitate a rewrite of the software that governs (B). Grossberg and Versace (2008) have developed a detailed model (SMART) for (A) that runs as follows (for example, in vision). The match/mismatch computations in the LGN compare the bottom-up input from the retina (that reports what actually is outside) and the top-down input from layer 6 neurons in V1 (which reports that what the brain computes should most probably be outside). The matching process is carried out by a modulatory on-center / off-surround circuit that selects a critical attended-to feature.

A match results in increased synchrony in cerebral gamma oscillations, which promotes spike-timing-dependent plasticity (STDP), and is related to the continued use by the brain of the software it is currently using. Further, these authors suggest that this increased synchrony is the result of a “persistent resonant state” in feedback loops. If a mismatch occurs, the LGN activates the midline and intralaminar nuclei of the thalamus (MILN). This, in turn, activates layer 5 pyramidal neurons on their apical dendrites via synapses in layer 1, which then activate layer 6 neurons.  Subsequently, this activation cascade sets in motion a process, which promotes beta synchronization, inhibits STDP and generates modulation of software operations such as giving ‘reset’ and hypothesis-testing instructions. A mismatch also triggers activity in the basal nucleus of Meynert and widespread release of acetylcholine in the cortex. This cholinergic mechanism alters the degree of match required for bottom-up and top-down input to prevent reset.

         Grossberg and Versace’s SMART model then allots subsidiary functions in this process to circuits based on the cingulate gyrus and the hippocampus. However, the authors do not mention the claustrum. We will therefore posit how the claustrum may be involved in this process - by a role in processing synchronized oscillations. The model runs as follows, again, using vision as an example. If processing in the LGN as described above results in a “match” verdict, a retinotopic signal is sent to the claustrum’s visual map, which promptly relays it to the appropriate part of the cortex wherein it sets up a short-lived excitation. On the principle that cells that work together oscillate together, the activated cells will start to synchronize their oscillations. With a dynamic and ever-changing retinal input, these small groups of synchronized cells continually form and reform in competition. Their fate depends on the degree of subsequent recruitment and reinforcement dictated by the nature and details of the task being performed. On top of the general facilitation of gamma oscillations promoted by a ‘match’ verdict, the claustrum may add a ‘local’ perspective. Take, for example, a condition in which the subject is engaged in a complex task that involves coordination between sight and touch.  Frequent activations of selected visual and tactual cortical neurons will result. Both sets of cells will generate axon spikes that carry the frequency of these oscillations. If these neurons were each initially oscillating at their own frequencies, this concomitant activity will tend to bind their frequencies of oscillation, and of spike timing patterns, into line. These cortical neurons project to the claustrum via specific layer 6 pyramidal cells. The axons of these layer 6 cells synapse on the cell bodies and dendrites of claustral pyramidal (P) cells and will induce in them oscillations at the same frequency. The claustral P cells project back to the cortex and so can set up reverberating cortico-claustro-cortical loops.

The layer 6 cells in the cortex of the two groups (hearing and vision) are far apart, joined only by corticocortical connections. The claustral cell bodies are close proximity to one another, connected by short axon collaterals, and embedded in a GABAergic syncytium, which allows for very fast communication.  Activation of such GABAergic systems strongly promotes gamma synchronization (Vervaeke et al. 2010). As a result, co-temporal activity in the auditory and visual cortex in this case may result in a common frequency of oscillation under the influence of weak cortico-cortical interactions but strong intraclaustral interactions. This particular link-up may persist while this particular audio-visual task is being conducted. When the task is completed, the conjoined activity in the cortico-claustro-cortical system will begin to subside and eventually replaced when a new stimulus is received, after which the entire process will be repeated elsewhere.

Thus, this model allows the claustrum to promote synchronized intermodal gamma oscillations in widely separated parts of the cortex. In previous models it was thought that, in order to accomplish this, the claustrum needed direct input to individual claustral cells via axons emanating from the two disparate cortices. The data recently generated by Remedios et al. (2010) suggest that these direct connections do not exist. However, in our present model, such direct connections are unnecessary, since the majority of the binding is done inside the claustrum. Furthermore, complex spike codes are not involved (as in our previous model). In non-sensory “higher” cognitive processing operations, which require interaction between any higher cortical areas, the same mechanism may be involved. Weak cortico-cortical promotion of a common oscillation frequency is promoted by a strong intraclaustral facilitation. The P cells in the claustrum, embedded within a GABAergic syncytium, may generate a complex, fluctuating, competitive and dynamic domain of shared and disparate oscillations. This is modulated by the ever-changing pattern of afferent spikes from the cortex, as well as by chemical neuromodulators such as dopamine and input from subcortical structures. For example, positive reward leads to a widespread release of dopamine in the brain. Dopamine promotes gamma synchrony. Remedios et al. (2010) suggest that the short, sharp initial claustral signal from the sense organ to the cortex involves salience and not “binding.” We suggest the hypothesis that this short, sharp initial ‘priming’ signal only switches on the cortico-claustral mechanism which does, given favorable circumstances, effect “binding.” During a mismatch period, signals from the MILN may reset not only cortical software but claustral software as well. Grossberg and Versache (2008) give examples of several such priming operations in the thalamocortical system

This new formulation is very much in line with what Crick and Koch proposed in their hypothesis, namely, “…widespread intra-claustral interactions.” These, they believed, may be in the form of “…waves of information [that] can travel within the claustrum.” They further suggest that this may involve dendrodendritic synapses and networks of gap-junction-linked neurons. Additionally, they posit that claustral neurons “…could be especially sensitive to the timing of the inputs,” but they do not suggest any precise mechanism to perform these functions. Rahman and Baizer (2007) also suggest that the claustrum mediates “…integration across compartments mediated by inhibitory interneurons.” All we have added is the exact form of these interactions, namely, multiple synchronized oscillations.

This new hypothesis makes a prediction. The axons from the LGN, that carry the rapid ‘priming’ message, may synapse on neurons within the claustrum’s retinotropic map.  Whereas the axons from the cortex that engineer the secondary synchronized oscillations may synapse throughout the body of the claustrum.



New Reference


Grossberg S, Versace M. 2008 Spikes, synchrony and attentive learning by laminar thalamocortical circuits. Brain Res. 1218: 278-313.