Completed on 21 Mar 2016 by John Tuthill . Sourced from http://biorxiv.org/content/early/2016/02/29/041889.
Login to endorse this review.
Author response is in blue.
Thanks for this nice review of the current evidence for attention in the insect nervous system. I appreciate the effort to bridge the gap between the more cognitive aspects of attention, and circuit and cell-type specific mechanisms of neuromodulation. Those of us who study behavioral state-dependent modulation in fly sensory circuits often feel that we are studying a form of attention, but may be reluctant to use this term because of its cognitive, anthropocentric implications.
In your review, you present two competing “models” for selective attention in the insect brain. The first is that there exist centralized attention-controlling circuits, which monitor sensory inputs in order to select relevant stimuli upon which to “focus”. The second is that attentional circuits are more distributed, but there exists some mechanism by which local attention circuits are used to recruit sensory signals into coherent “functional ensembles”. As you state in the last paragraph of your paper, diverse attention-like signals clearly exist throughout the fly CNS, so the second model is at least partly true. What is less clear to me is the requirement for “functional ensembles” that exist across the fly brain.
I would propose a third model: that distributed sensory systems may be independently modulated in a state-dependent manner, and that this hierarchical modulation iterated over successive processing stages gives rise to the high-level phenemona we refer to as attention. This model does not exclude the existence of top-down attention signals, but rather proposes that there exist both bottom-up (local) and top-down (global) mechanisms of attention, which can operate independently. A relevant analogy is the ethological model of behavior, which proposes that seemingly complex behaviors can emerge from the permutation and modification of basic sensorimotor reflexes.
An example of how attentional modulation can act at multiple levels within a sensory circuit is the state-dependent habituation of the locust descending contralateral movement detectors (DCMD) to looming stimuli, (Rowell, 1971). Like the examples in the fly lobula plate that you cite, the attentional signal that relieves DCMD habituation is input from an octopaminergic neuron (Bacon et al, 1997). Another interesting aspect of this system is that the locust DCMD can habituate to repeated stimuli, while remaining attentive to novel stimuli (Gray et al, 2005). A possible explanation for this phenomenon is that there exist both local and global attentional signals, implemented at different levels of the visual processing hierarchy, possibly even by the same attentional neuromodulator (octopamine). This is consistent with the demonstration in the fly that octopamine modulates visual motion coding at both local (lamina; Tuthill et al, 2014) and global (lobula plate; Suver et al, 2012) scales.
Bacon JP, Thompson KSJ, Stern M. (1995) Identified octopaminergic neurons provide an arousal mechanism in the locust brain. J Neurophysiol 74: 2739–2743, 1995.
Gray JR (2005) Habituated visual neurons in locusts remain sensitive to novel looming objects. J Exp Biol 208: 2515–2532, 2005.
Rowell CHF. Variable responsiveness of a visual interneurone in the freely-moving locust, and its relation to behaviour and arousal. J Exp Biol 55: 727–747, 1971b.
MP Suver, A Mamiya, MH Dickinson Octopamine neurons mediate flight-induced modulation of visual processing in Drosophila. Current Biology 22 (24), 2294-2302, 2012.
JC Tuthill, A Nern, GM Rubin, MB Reiser. "Wide-field feedback neurons dynamically tune early visual processing." Neuron 82(4): 887-895, 2014.
Hi John. Thanks very much for the thoughtful comments. Bruno and I have discussed them, and here's our thinking.
First of all, the references you've provided are totally pertinent and we wish we'd included them. It's not much of an excuse, but we had already exceeded the editorial stipulations on word count and, knowing that we wanted to focus on the 3 very recent Ca++ imaging papers, we knew that we'd have to neglect some great papers. (I suppose this is an argument for journal-free publication methods.)
As for the science, we agree that the option #3 you've described is reasonable and a good hypothesis. Our aims in describing both region-specific and brain-wide scenarios was not to be exclusive, and that, as always, the reality is likely a mixture of both. And correspondingly, it seems like your idea, state-dependent sensory-inward filtering leading to brain-wide patterns could also be happening.
I don’t think we have a position on whether brain-wide assemblies only arise by the close-synaptic interaction of neurons brain-wide (i.e. coordination arising because neurons that happen to be far apart spatially being close to each other in a connectomic/network sense) or by inherited correlation from state-dependent peripheral independent correlated modulation. Both seem plausible, and the evidence for brain-wide coordination, from electrode arrays, doesn't discriminate these hypotheses.
That said, we feel that the lack of a hierarchical feed-forward organization in the central brain (e.g. the many excitatory loops in the central complex) means that models relying on strictly feed-forward coordination may not be particularly pertinent. Conversely, the non-hierarchical organization of the central circuits may mean that even if peripheral/sensory activity patterns are not correlated, correlation can be engendered centrally.