Closing the gap between mind and brain with the dynamic connectome

At the pinnacle of the 17th century scientific revolution, René Descartes, the father of modern philosophy, published his monumental Meditations on First Philosophy (1), in which he proposed a division between soul and body—mind and brain—with the former in charge of our thoughts and conscious decisions (res cogitans) and the latter executing them through mechanical acts (res extensa). Scientists nowadays have departed from Cartesian dualism and mainly embrace materialism, the notion that mind and brain are one and the same thing: The mind is not an intangible entity in charge of our thoughts, but, rather, the activity of neurons, and the way neurons connect and form circuits determine our behavior. In the late 1940s, Donald Hebb (2) described a natural mechanism by which the connection between neurons can be changed to encode new experiences, something that was, a few decades later, implemented in the very influential Hopfield neural networks (3), updating the weights between nodes to store different memories. So, it seems that the structural connectivity between neurons, what has been named the “connectome” (4), determines behavior. However, although structural connectivity should clearly constrain brain function, the vast richness of behavior cannot merely be reduced to physical connections, and the “functional connectivity” taking place at a particular time is context and state dependent (5). Furthermore, neuromodulators regulate neural rhythms and should play a key role in modulating functional connectivity (6). Within this framework, the study by Kringelbach et al. (7) in PNAS shows how to integrate the structural anatomical connectivity with the dynamics of neuromodulatory systems into a “dynamic connectome.”

If behavior were solely determined by the structural connectivity of the brain, we would always obtain exactly the same response to a given stimulus. However, this is far from being the case. In particular, ambiguous percepts provide an excellent example of how the brain can react differently to the same sensory inputs (8, 9). A classic paradigm to show this effect is binocular rivalry, where two distinct images, presented independently to each eye, compete with each other and give rise to a fluctuating perception of one or the other (10). Single-cell recordings along the ventral visual pathway in monkeys have shown an increase in the number of neurons responding according to the subjective perception of the animals in higher visual areas (8). These areas project to the medial temporal lobe (MTL) (11), where, in humans, strong single-neuron responses have been found in response to the presentation of pictures of particular persons (12), and using short presentation times together with backward masking, it was shown that these neurons fire only when the stimulus is recognized and remain at baseline firing levels when it is not, even if the stimulus presentation, at the threshold of conscious perception, was exactly the same in both cases (Fig. 1A) (13). Another study using morphings between the faces of two persons found that these neurons, initially responding to one of the persons, fired only when the subjects said that the ambiguous morphed picture corresponded to this person, and not when they said it corresponded to the other one (Fig. 1B) (14). In the study of Fig. 1A, the different responses between recognized and not recognized trials could be attributed to a higher degree of attention when the pictures were recognized. But, in the study of Fig. 1B, given that the subjects did not know beforehand to which face the neuron fired, both pictures were equally salient, and, therefore, the different responses cannot be attributed to attention modulations but to an internal varying state of the system. These studies clearly show that a static connectome, with all its weights fixed and reacting in the same way to identical sensory stimuli, cannot capture the variability of observed behavior. But how can dynamic modulations of the structural anatomical connectivity be implemented in practice?

Fig. 1.

(A) Raster plots and peristimulus time histograms (PSTHs) of a single neuron in the entorhinal cortex that fired selectively to a picture of the World Trade Center. The different presentation durations are shown at the left of each plot (66-ms duration being at the threshold of recognition in this case). Trials where the pictures were (were not) recognized are displayed with a blue (red) mark at time 0. Note the striking difference in the responses to presentations when the picture was recognized compared to when it was not. Reprinted from ref. 13. (B) Raster plots and PSTHs of a single neuron in the hippocampus that responded to a picture of actress Jodie Foster (Right), but not to one of Nicole Kidman (Left). The response to the ambiguous morphed image (Center Left and Center Right) was larger when the subject recognized it as Foster compared to when the subject recognized it as Kidman.

The main tenet of the article by Kringelbach et al. (7) is that neurotransmitters’ systems can modulate the connectome over time, thus enabling a plethora of behaviors with the same underlying structural connectivity. To this end, a modeling approach is presented, in which the structural connectivity is estimated through diffusion MRI, and is coupled to the neurotransmitter system, estimated from positron electron tomography data. Both systems are portrayed by a set of mutually coupled dynamic equations, which are used to fit the functional connectivity, obtained from functional MRI. This approach was tested by exploring the effects that a psychedelic drug (psilocybin) had on neuronal activity, showing that the dynamically coupled neuronal and neuromodulatory systems give a significantly better fit to the measured data, compared to alternative models in which both systems were uncoupled, or in which the neuromodulatory system, rather than being dynamically updated, was frozen in time.

Kringelbach et al. (7) provide an elegant framework for incorporating into the connectome the dynamic variations caused by neuromodulatory systems. It is foreseeable that this approach could be further extended to other neuromodulators that have a role in cognitive functions, as has, for example, been shown for acetylcholine, which contributes to attention modulation (15). Such attention modulations could indeed explain how visual recognition models, coupled to the acetylcholine (or other neurotransmitters) neuromodulator dynamics, give rise to the variable percepts and neuronal

Kringelbach et al. provide an elegant framework for incorporating into the connectome the dynamic variations caused by neuromodulatory systems.

responses for exactly the same visual inputs as shown in Fig. 1A. But, beyond modeling attention levels, future studies may also widen the approach by Kringelbach et al. to incorporate the dynamics of the state of the system, which gives rise to the variabilities shown in Fig. 1B. In this respect, it has been argued that, in the human MTL, context is given by the coactivation of cell assemblies (12). It is therefore plausible that different sustained activations triggered by one or another context may change the ongoing state of the system and provide a variety of possible responses to the structural connectome, besides the one given by the dynamics of neuromodulators.

Returning to Descartes, the notion of a mind separated from the body, although incorrect, has an undeniable appeal, as it reflects our experience of reacting in so many different ways to similar situations. An autonomous mind, exercising our will, can certainly account for such a plethora of reactions that are then executed mechanically by the body. Just the brain and its underlying connectome (without an independent mind) seem, in principle, too mechanical and constrained to go beyond stimulus−response associations and allow the vast richness of our behavior. In this respect, Kringelbach et al. (7) provide a framework with which to formulate a dynamic connectome that comes closer to the versatility of our mind, not as an ethereal mysterious entity, but as the dynamically changing activity of neurons.