- Email: [email protected]
Neuroscience: Figured Out by Feedback to the Thalamus
Current Biology
Dispatches neurons is important for the maintenance of body size memory. It is tempting to speculate about underlying neuronal computation for selfbody-size memory and what guides the initiation of gap crossing. Confronted with a cleft in the walking path, the perceived gap width must be compared with the body reach information. As the gap is an obstacle that may as default counteract locomotion, flies may keep walking only if the predicted body reach surpasses the perceived gap width (Figure 2). Intriguingly, a large-scale GAL4 screen by the same research group revealed that the C2/C3 neurons in the early visual system are critical in accurate estimation of gap size by controlling contrast gain [5]. This C2/C3 pathway may thus provide the comparator with the gap size information (Figure 2). At the same time, this gain control pathway might also be used for the formation of self-body-size memory by regulating the visual feedback signal (Figure 2). This could be tested in the future by transiently blocking the C2/C3 activity during training. The D7 neurons in the protocerebral bridge may be the site of body-sizememory formation, since the Rutabaga adenylate cyclase, a molecular coincidence detector for learning [12], functions in these cells (Figure 2).
Considering the post-training role of dCREB2-b in the D7 neurons, these cells could alternatively act as the comparator to determine gap crossing initiation (Figure 2). The protocerebral bridge was also shown to be important for targeting the landing site of the gap [13]. It will be an exciting endeavor to understand how the protocerebral bridge circuit integrates all relevant information and helps the fly to determine whether to attempt a crossing or take the view that discretion is the better part of valour, and to refrain. REFERENCES 1. Kravitz, E.A., and Huber, R. (2003). Aggression in invertebrates. Curr. Opin. Neurobiol. 13, 736–743. 2. Ben-Nun, A., Guershon, M., and Ayali, A. (2013). Self body-size perception in an insect. Naturwissenschaften 100, 479–484. 3. Krause, T., Spindler, L., Poeck, B., and Strauss, R. (2019). Drosophila acquires a longlasting body size memory from visual feedback. Curr. Biol. 29, 1833–1841.e3. 4. Pick, S., and Strauss, R. (2005). Goal-driven behavioral adaptations in gap-climbing Drosophila. Curr. Biol. 15, 1473–1478. 5. Triphan, T., Nern, A., Roberts, S.F., Korff, W., Naiman, D.Q., and Strauss, R. (2016). A screen for constituents of motor control and decision making in Drosophila reveals visual distanceestimation neurons. Sci. Rep. 6, 27000.
6. Heisenberg, M., and Wolf, R. (1984). Vision in Drosophila. Genetics of Microbehavior. Studies of Brain Function vol. 12 (New York, Heidelberg: Springer). 7. Davis, H.P., and Squire, L.R. (1984). Protein synthesis and memory: a review. Psycholog. Bull. 96, 518–559. 8. Wolff, T., Iyer, N.A., and Rubin, G.M. (2015). Neuroarchitecture and neuroanatomy of the Drosophila central complex: A GAL4-based dissection of protocerebral bridge neurons and circuits. J. Comp. Neurol. 523, 997–1037. 9. Lin, C.Y., Chuang, C.C., Hua, T.E., Chen, C.C., Dickson, B.J., Greenspan, R.J., and Chiang, A.S. (2013). A comprehensive wiring diagram of the protocerebral bridge for visual information processing in the Drosophila brain. Cell Rep. 3, 1739–1753. 10. Franconville, R., Beron, C., and Jayaraman, V. (2018). Building a functional connectome of the Drosophila central complex. eLife 7, e37017. 11. Perazzona, B., Isabel, G., Preat, T., and Davis, R.L. (2004). The role of cAMP response element-binding protein in Drosophila longterm memory. J. Neurosci. 24, 8823–8828. nio, P., and Preat, T. (2010). 12. Gervasi, N., Tche PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl cyclase and spatial regulation by dunce phosphodiesterase. Neuron 65, 516–529. 13. Triphan, T., Poeck, B., Neuser, K., and Strauss, R. (2010). Visual targeting of motor actions in climbing Drosophila. Curr. Biol. 20, 663–668.
Neuroscience: Figured Out by Feedback to the Thalamus Matthew W. Self1 and Pieter R. Roelfsema1,2,3 1Department
of Vision & Cognition, Netherlands Institute for Neuroscience, Meibergdreef 47, 1105 BA, Amsterdam, the Netherlands of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University, De Boelelaan 1085, 1081HV Amsterdam, The Netherlands 3Psychiatry department, Academic Medical Center, Postbus 22660, 1100 DD Amsterdam, The Netherlands Correspondence: [email protected] (M.W.S.), [email protected] (P.R.R.) https://doi.org/10.1016/j.cub.2019.05.022 2Department
The lateral geniculate nucleus of the thalamus (LGN) is a relay nucleus between the retina and the visual cortex. A new brain imaging study shows that LGN activity is modulated by figure–ground organization, even when the figure and ground are presented to different eyes: a hallmark of a cortical feedback effect. Our visual perception does not directly resemble the pattern of lightness recorded by the photoreceptors of the
retina: it is instead organized to enhance our perception of objects; regions of the scene belonging to the same object are
R574 Current Biology 29, R568–R591, June 17, 2019 ª 2019 Elsevier Ltd.
grouped together in perception and segregated from the background. We have long known [1] that the neural
Current Biology
Dispatches representations of objects (or ‘figures’) are enhanced in the primary visual cortex (V1) and it is thought that loops of feedforward and feedback processing between V1 and higher cortical visual areas, so-called ‘recurrent processing’, are necessary to correctly segregate the visual scene [2]. The mechanisms of figure–ground segregation are also closely intertwined with the mechanisms of visual attention. Attention enhances the strength of figure–ground modulation signals in visual cortex [3], while at the same time it respects the boundaries of figures and spreads within perceptual objects [4]. Much of the research into figure–ground segregation has focused on recurrent processing between cortical visual areas, but cortical visual areas also send feedback projections to the lateral geniculate nucleus (LGN) via corticogeniculate projection neurons situated in layer 6 of visual cortex (reviewed in [5]). A recent study in monkeys [6] suggested that these feedback projections are able to modulate the activity of LGN relay neurons, enhancing their response to figures, but the properties of this feedback projection remain poorly understood and proof that the modulation is caused by cortical feedback to the LGN has been lacking. As reported in this issue of Current Biology, Poltoratski et al. [7] have studied the representation of figures and backgrounds in human LGN using functional magnetic brain imaging (fMRI). They have found evidence that cortex sends a feedback signal to the LGN that is enhanced on figures, even when attention is directed elsewhere, suggesting an automatic labelling of figure regions at a very early level of processing in the visual system. Participants in the study of Poltoratski et al. [7] fixated on a central cross and saw oriented textures that were regenerated every 200 ms. The textures contained two figure regions on opposite sides of fixation, one in which the orientation of the texture was orthogonal to the background (the ‘incongruent’ patch) producing a strong figure–ground percept, and one in which the texture had the same orientation as the background (the ‘congruent’ patch). Importantly, in both cases, the boundaries of the figure were masked out with a grey ring to remove any boundary-related effects. Attention was manipulated by instructing the
Visual areas RIGHT Eye image
V1 LGN
LEFT Eye image
Current Biology
Figure 1. Presenting figure and background to separate eyes reveals a cortical feedback effect in the LGN. Poltoratski et al. [7] presented participants with two patches; a congruent patch in which the orientation of the patch was the same as the background, and an incongruent patch in which the orientation was orthogonal and appeared as a perceptually segregated figure. In the critical experiment of the paper, the patches were presented to one eye and the backgrounds to the other eye using red/green anaglyph glasses. The participants performed a task at fixation, meaning that the patches were unattended. The projections of the retina to the LGN (only the right LGN is shown in detail here for simplicity) are shown coded by their eye of origin. The LGN is a six-layered structure with different layers responding exclusively to visual input from only one eye. LGN neurons could not therefore discriminate between the congruent and incongruent patch using the information they received from the retina. LGN neurons project to layer 4 of V1 in a segregated columnar fashion, forming the wellknown ocular-dominance columns. However, after this stage information from the two eyes is combined by cells in the superficial and deep layers of V1. It is thought that figure–ground segregation occurs through recurrent interactions with higher visual areas (shown by the yellow arrows). Cells in the deep layers of V1 and the higher visual areas send feedback projections back to the LGN (orange arrows) and are likely the source of the difference in BOLD signal magnitude between the congruent and incongruent patches observed in this study.
participants to attend to one patch and monitor the texture for a change in spatial frequency content and ignore changes at the opposite location. In this way, the authors could examine the responses of attended and unattended texture patches. As neurons in the LGN do not represent orientation, they should be blind to the differences between the two texture patches, which differed only in the relative orientation compared to the background. Furthermore, the receptive fields of LGN neurons are small and the difference between congruent and incongruent texture patches was determined by image regions outside the texture patch, far from the receptive field. Hence, the feedforward input from the retina to the LGN neurons was held constant.
The results, however, showed unambiguously that voxels in the LGN representing the incongruent patch had higher BOLD signals than those representing the congruent patch. Where does this signal come from? To test this, Poltoratski et al. [7] took advantage of a quirk of the anatomical organization of the visual system. Signals from the two eyes are processed by different layers within the LGN and hence the neural responses there are purely monocular. The authors presented the figures to one eye and the background to the other eye using anaglyph (red/green) glasses (Figure 1). This manipulation removes any possibility of interactions between the texture patches and their surround within the LGN itself. The results, however, remained remarkably similar to those of the first experiment: the incongruent
Current Biology 29, R568–R591, June 17, 2019 R575
Current Biology
Dispatches texture patch produced higher BOLD signals in the LGN compared to the congruent patch. In this version of the experiment, both patches were unattended as attention was directed to a central fixation task and the results were replicated for two different central tasks of different difficulty levels. The results provide strong evidence that the difference in signal between the congruent and incongruent patch must be due to feedback from a brain region which combines information across the two eyes. As binocular integration occurs first in the superficial layers of V1, this implicates cortex as the source of the signals. This feedback signal from the cortex to the LGN that enhances the figure representation even occurs when the figure is ignored. The paper of Poltoratski et al. [7] adds to a growing list of studies which have found modulations of activity in the LGN for processes that were previously thought to be exclusively cortical in nature. Neural activity in the LGN is modulated by spatial attention [8] and brain imaging studies have even found modulations of LGN BOLD signals related to orientation [9] and binocular rivalry [10,11]. One potential caveat concerns the use of fMRI to infer changes in neural activity in the LGN. The BOLD signal is particularly sensitive to synaptic inputs and, as corticogeniculate inputs to the LGN actually outnumber the synaptic inputs received from retinal ganglion cells [5], it is likely that the BOLD signal in the LGN is dominated by feedback signals. It is impossible to tell using non-invasive techniques whether these modulatory inputs actually give rise to changes in spiking activity in the LGN. Nevertheless, the new study [7] shows that figure– ground information is fed-back to the LGN and, given a previous monkey study [6] which did demonstrate an enhanced spiking response on figures, this suggests that LGN spiking activity is modulated by figure–ground structure in humans as well. The work of Poltoratski et al. [7] moves beyond previous studies by showing an enhanced representation of figures that fall outside the focus of attention. Typically in primate studies of figure–ground modulation, the figure is also the target for a saccadic eyemovement and is associated with reward during many months of behavioral training
(though see [3,12,13]). This has led to the view that figure–ground modulation can be seen as a form of object-based attentional selection. The new results [7] provide support for an alternative viewpoint in which potential figure-regions are enhanced in parallel across the visual scene in a relatively automatic fashion. These unattended figure representations have been referred to as ‘proto-objects’ [12,14] and we have recently demonstrated that the neural representations of unattended protoobjects are enhanced in the primary visual cortex in monkeys [12]. In proto-object theory, the visual system utilizes the statistics of objects in natural scenes (for example, the Gestalt grouping cues) to enhance the representations of visual features which are likely to belong to objects. The study of Poltoratski et al. [7] indicates that cortex enhances the representations of proto-objects through recurrent loops with the LGN. Objectbased attention can then act upon these proto-objects to select and group visual features into coherent object representations [4,15,16]. Poltoratski et al. [7] link the parallel enhancement of both attended and nonattended figures to the predictive-coding framework [17,18]. According to this view, the uniform orientation of the background elements provides a predictable spatial context, which is sent to early visual areas to ‘explain away’ the predictable background elements. The figure elements, having an unpredictable orientation, produce an error signal that is propagated through to higher areas. While the results are broadly in line with this theory, it should be noted that it is impossible to determine whether background activity is being suppressed using fMRI and previous studies have demonstrated that feedback connections also enhance representations that are predicted by the higher areas, which goes against the predictive coding framework [3,19]. Furthermore, the predictions of such a predictive coding depend on spatial scale. Poltoratski et al. [7] interpret the enhanced activity for the figure in the LGN as evidence that a relatively global contextual feedback signal is sent back to the thalamus, as the orientation of the figure elements is only unpredictable at large spatial scales. At a finer scale,
R576 Current Biology 29, R568–R591, June 17, 2019
however, the central figure elements can be predicted from nearby elements of the same patch, which should lead to a relative suppression of the center of objects. Feedback connections exert both excitatory and inhibitory effects on predicted information at lower processing levels [20], and it is likely that this topic will remain an active area of research in the coming years. Poltoratski et al. [7] have provided strong evidence that cortex does send feedback back to the LGN to help organize our perception, but leaves us with the tantalizing puzzle of what the function of this feedback ultimately is. Is it a predictive template that gates incoming information in the thalamus to improve coding efficiency or a signal that boosts potentially interesting regions of the visual scene to aid their selection by attentional mechanisms? Whatever the result, this study provides a fascinating addition to our knowledge of the mechanisms of perceptual organization. REFERENCES 1. Lamme, V.A. (1995). The neurophysiology of figure-ground segregation in primary visual cortex. J. Neurosci. 15, 1605–1615. 2. Lamme, V.A., and Roelfsema, P.R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends Neurosci. 23, 571–579. 3. Poort, J., Raudies, F., Wannig, A., Lamme, V.A., Neumann, H., and Roelfsema, P.R. (2012). The role of attention in figure-ground segregation in areas V1 and V4 of the visual cortex. Neuron 75, 143–156. 4. Jeurissen, D., Self, M.W., and Roelfsema, P.R. (2016). Serial grouping of 2D-image regions with object-based attention in humans. eLife 5, e14320. 5. Briggs, F., and Usrey, W.M. (2011). Corticogeniculate feedback and visual processing in the primate. J. Physiol. 589, 33–40. 6. Jones, H.E., Andolina, I.M., Shipp, S.D., Adams, D.L., Cudeiro, J., Salt, T.E., and Sillito, A.M. (2015). Figure-ground modulation in awake primate thalamus. Proc. Natl. Acad. Sci. USA 112, 7085–7090. 7. Poltoratski, S., Maier, A., Newton, A.T., and Tong, F. (2019). Figure-ground modulation in the human lateral geniculate nucleus is distinguishable from top-down attention. Curr. Biol. 29, 2051–2057. 8. McAlonan, K., Cavanaugh, J., and Wurtz, R.H. (2008). Guarding the gateway to cortex with attention in visual thalamus. Nature 456, 391–394.
Current Biology
Dispatches 9. Ling, S., Pratte, M.S., and Tong, F. (2015). Attention alters orientation processing in the human lateral geniculate nucleus. Nat. Neurosci. 18, 496–498.
13. Marcus, D.S., and Van Essen, D.C. (2002). Scene segmentation and attention in primate cortical areas V1 and V2. J. Neurophysiol. 88, 2648–2658.
10. Haynes, J.-D., Deichmann, R., and Rees, G. (2005). Eye-specific effects of binocular rivalry in the human lateral geniculate nucleus. Nature 438, 496–499.
14. von der Heydt, R. (2015). Figure-ground organization and the emergence of protoobjects in the visual cortex. Front. Psychol. 6, 1–10.
11. Wunderlich, K., Schneider, K.A., and Kastner, S. (2005). Neural correlates of binocular rivalry in the human lateral geniculate nucleus. Nat. Neurosci. 8, 1595–1602.
15. Rensink, R.A. (2000). The dynamic representation of scenes. Vis. Cogn. 7, 17–42.
12. Self, M.W., Jeurissen, D., van Ham, A.F., van Vugt, B., Poort, J., and Roelfsema, P.R. (2019). The segmentation of proto-objects in the monkey primary visual cortex. Curr. Biol. 29, 1019–1029.
16. Roelfsema, P.R. (2006). Cortical algorithms for perceptual grouping. Annu. Rev. Neurosci. 29, 203–227. 17. Rao, R.P., and Ballard, D.H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical
receptive-field effects. Nat. Neurosci. 2, 79–87. 18. Lee, T.S., and Mumford, D. (2003). Hierarchical Bayesian inference in the visual cortex. J. Opt. Soc. Am. A. Opt. Image Sci. Vis. 20, 1434– 1448. 19. Chelazzi, L., Miller, E.K., Duncan, J., and Desimone, R. (2001). Responses of neurons in macaque area V4 during memory-guided visual search. Cereb. Cortex 11, 761–772. 20. Zhang, S., Xu, M., Kamigaki, T., Phong Hoang Do, J., Chang, W.-C., Jenvay, S., Miyamichi, K., Luo, L., and Dan, Y. (2014). Long-range and local circuits for top-down modulation of visual cortex processing. Science 345, 660–665.
Zoology: Worming into the Origin of Bilaterians taz Ferdinand Marle Molecular Genetics Unit, Okinawa Institute of Science & Technology, Onna-son, Japan Correspondence: [email protected] https://doi.org/10.1016/j.cub.2019.05.006
Xenacoelomorphs, a group of worms with simple body organization, have been proposed to represent the first offshoot of bilaterians. A new study shows that they might instead belong to the deuterostomes, just as echinoderms and vertebrates. The nature of the last common ancestor of all bilaterally symmetrical animals — the urbilaterian — is one of the key questions in zoology, partly because it reflects on the origin of the key organ systems that make us who we are. There are two opposing views of what this ancestor may have been like — either it was a rather simple organism or a fairly complex one. And, a lot of the debate surrounding this dichotomy revolves around the phylogenetic position of a handful of worms — the xenacoelomorphs. Two alternative positions were successively proposed over the years: one suggests that this group of morphologically simple worms is the sister-group to all other bilaterians. This branching would mean that the bilaterian ancestor was a rather unsophisticated organism. Alternatively, other researchers have proposed that these xenacoelomorphs instead are a derived offshoot of deuterostomes, one of the two main clades of bilaterians, which mean that their simplicity derives from secondary character loss. A new study
Philippe, Max Telford and by Herve colleagues [1] in Current Biology brings new elements to support this latter tree, in particular a thorough evaluation of methodological bias that can affect phylogenetics reconstruction — such as the famous long-branch attraction. They conclude that xenacoelomorphs are the sister-group of a clade called ‘Ambulacraria’, which includes sea stars, sea urchins and acorn worms. The idea that the urbilaterian was a simple organism without a body cavity (coelom) reminiscent of the planula larva of some cnidarians was mentioned a long time ago, for instance in the famous zoology textbook of Libbie Hyman [2]. At this time, morphologically simple animals, such as nematode roundworms and flatworms, were considered as early branches of the animal tree of life. Morphological cladistics and molecular phylogeny then helped redefine our understanding of animal evolution by splitting all bilaterian animals into two main clades: the protostomes (including
insects, molluscs, annelids, flatworms) and the deuterostomes (including vertebrates, tunicates, sea urchins and sea stars). This reclassification implied that seemingly simple lineages may have originated through simplification and secondary character losses from a more complex coelom-bearing urbilaterian ancestor [3]. Then, in the late nineties, a study [4] pointed out that a neglected lineage of flatworms — the acoels — might have represented a sister-group to both protostomes and deuterostomes. This reignited the debate concerning the nature of the bilaterian ancestor and propelled these overlooked animals to the front stage of animal evolution. Acoel flatworms indeed show a very simple planula-like organization and do not possess structures that are found in other bilaterians, such as a coelomic cavity, excretory organs, or nerve chords (Figure 1A). All bilaterians that are not acoels were dubbed ‘Nephrozoa’ (highlighting the presence of an excretory
Current Biology 29, R568–R591, June 17, 2019 ª 2019 Elsevier Ltd. R577
Dispatches neurons is important for the maintenance of body size memory. It is tempting to speculate about underlying neuronal computation for selfbody-size memory and what guides the initiation of gap crossing. Confronted with a cleft in the walking path, the perceived gap width must be compared with the body reach information. As the gap is an obstacle that may as default counteract locomotion, flies may keep walking only if the predicted body reach surpasses the perceived gap width (Figure 2). Intriguingly, a large-scale GAL4 screen by the same research group revealed that the C2/C3 neurons in the early visual system are critical in accurate estimation of gap size by controlling contrast gain [5]. This C2/C3 pathway may thus provide the comparator with the gap size information (Figure 2). At the same time, this gain control pathway might also be used for the formation of self-body-size memory by regulating the visual feedback signal (Figure 2). This could be tested in the future by transiently blocking the C2/C3 activity during training. The D7 neurons in the protocerebral bridge may be the site of body-sizememory formation, since the Rutabaga adenylate cyclase, a molecular coincidence detector for learning [12], functions in these cells (Figure 2).
Considering the post-training role of dCREB2-b in the D7 neurons, these cells could alternatively act as the comparator to determine gap crossing initiation (Figure 2). The protocerebral bridge was also shown to be important for targeting the landing site of the gap [13]. It will be an exciting endeavor to understand how the protocerebral bridge circuit integrates all relevant information and helps the fly to determine whether to attempt a crossing or take the view that discretion is the better part of valour, and to refrain. REFERENCES 1. Kravitz, E.A., and Huber, R. (2003). Aggression in invertebrates. Curr. Opin. Neurobiol. 13, 736–743. 2. Ben-Nun, A., Guershon, M., and Ayali, A. (2013). Self body-size perception in an insect. Naturwissenschaften 100, 479–484. 3. Krause, T., Spindler, L., Poeck, B., and Strauss, R. (2019). Drosophila acquires a longlasting body size memory from visual feedback. Curr. Biol. 29, 1833–1841.e3. 4. Pick, S., and Strauss, R. (2005). Goal-driven behavioral adaptations in gap-climbing Drosophila. Curr. Biol. 15, 1473–1478. 5. Triphan, T., Nern, A., Roberts, S.F., Korff, W., Naiman, D.Q., and Strauss, R. (2016). A screen for constituents of motor control and decision making in Drosophila reveals visual distanceestimation neurons. Sci. Rep. 6, 27000.
6. Heisenberg, M., and Wolf, R. (1984). Vision in Drosophila. Genetics of Microbehavior. Studies of Brain Function vol. 12 (New York, Heidelberg: Springer). 7. Davis, H.P., and Squire, L.R. (1984). Protein synthesis and memory: a review. Psycholog. Bull. 96, 518–559. 8. Wolff, T., Iyer, N.A., and Rubin, G.M. (2015). Neuroarchitecture and neuroanatomy of the Drosophila central complex: A GAL4-based dissection of protocerebral bridge neurons and circuits. J. Comp. Neurol. 523, 997–1037. 9. Lin, C.Y., Chuang, C.C., Hua, T.E., Chen, C.C., Dickson, B.J., Greenspan, R.J., and Chiang, A.S. (2013). A comprehensive wiring diagram of the protocerebral bridge for visual information processing in the Drosophila brain. Cell Rep. 3, 1739–1753. 10. Franconville, R., Beron, C., and Jayaraman, V. (2018). Building a functional connectome of the Drosophila central complex. eLife 7, e37017. 11. Perazzona, B., Isabel, G., Preat, T., and Davis, R.L. (2004). The role of cAMP response element-binding protein in Drosophila longterm memory. J. Neurosci. 24, 8823–8828. nio, P., and Preat, T. (2010). 12. Gervasi, N., Tche PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl cyclase and spatial regulation by dunce phosphodiesterase. Neuron 65, 516–529. 13. Triphan, T., Poeck, B., Neuser, K., and Strauss, R. (2010). Visual targeting of motor actions in climbing Drosophila. Curr. Biol. 20, 663–668.
Neuroscience: Figured Out by Feedback to the Thalamus Matthew W. Self1 and Pieter R. Roelfsema1,2,3 1Department
of Vision & Cognition, Netherlands Institute for Neuroscience, Meibergdreef 47, 1105 BA, Amsterdam, the Netherlands of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University, De Boelelaan 1085, 1081HV Amsterdam, The Netherlands 3Psychiatry department, Academic Medical Center, Postbus 22660, 1100 DD Amsterdam, The Netherlands Correspondence: [email protected] (M.W.S.), [email protected] (P.R.R.) https://doi.org/10.1016/j.cub.2019.05.022 2Department
The lateral geniculate nucleus of the thalamus (LGN) is a relay nucleus between the retina and the visual cortex. A new brain imaging study shows that LGN activity is modulated by figure–ground organization, even when the figure and ground are presented to different eyes: a hallmark of a cortical feedback effect. Our visual perception does not directly resemble the pattern of lightness recorded by the photoreceptors of the
retina: it is instead organized to enhance our perception of objects; regions of the scene belonging to the same object are
R574 Current Biology 29, R568–R591, June 17, 2019 ª 2019 Elsevier Ltd.
grouped together in perception and segregated from the background. We have long known [1] that the neural
Current Biology
Dispatches representations of objects (or ‘figures’) are enhanced in the primary visual cortex (V1) and it is thought that loops of feedforward and feedback processing between V1 and higher cortical visual areas, so-called ‘recurrent processing’, are necessary to correctly segregate the visual scene [2]. The mechanisms of figure–ground segregation are also closely intertwined with the mechanisms of visual attention. Attention enhances the strength of figure–ground modulation signals in visual cortex [3], while at the same time it respects the boundaries of figures and spreads within perceptual objects [4]. Much of the research into figure–ground segregation has focused on recurrent processing between cortical visual areas, but cortical visual areas also send feedback projections to the lateral geniculate nucleus (LGN) via corticogeniculate projection neurons situated in layer 6 of visual cortex (reviewed in [5]). A recent study in monkeys [6] suggested that these feedback projections are able to modulate the activity of LGN relay neurons, enhancing their response to figures, but the properties of this feedback projection remain poorly understood and proof that the modulation is caused by cortical feedback to the LGN has been lacking. As reported in this issue of Current Biology, Poltoratski et al. [7] have studied the representation of figures and backgrounds in human LGN using functional magnetic brain imaging (fMRI). They have found evidence that cortex sends a feedback signal to the LGN that is enhanced on figures, even when attention is directed elsewhere, suggesting an automatic labelling of figure regions at a very early level of processing in the visual system. Participants in the study of Poltoratski et al. [7] fixated on a central cross and saw oriented textures that were regenerated every 200 ms. The textures contained two figure regions on opposite sides of fixation, one in which the orientation of the texture was orthogonal to the background (the ‘incongruent’ patch) producing a strong figure–ground percept, and one in which the texture had the same orientation as the background (the ‘congruent’ patch). Importantly, in both cases, the boundaries of the figure were masked out with a grey ring to remove any boundary-related effects. Attention was manipulated by instructing the
Visual areas RIGHT Eye image
V1 LGN
LEFT Eye image
Current Biology
Figure 1. Presenting figure and background to separate eyes reveals a cortical feedback effect in the LGN. Poltoratski et al. [7] presented participants with two patches; a congruent patch in which the orientation of the patch was the same as the background, and an incongruent patch in which the orientation was orthogonal and appeared as a perceptually segregated figure. In the critical experiment of the paper, the patches were presented to one eye and the backgrounds to the other eye using red/green anaglyph glasses. The participants performed a task at fixation, meaning that the patches were unattended. The projections of the retina to the LGN (only the right LGN is shown in detail here for simplicity) are shown coded by their eye of origin. The LGN is a six-layered structure with different layers responding exclusively to visual input from only one eye. LGN neurons could not therefore discriminate between the congruent and incongruent patch using the information they received from the retina. LGN neurons project to layer 4 of V1 in a segregated columnar fashion, forming the wellknown ocular-dominance columns. However, after this stage information from the two eyes is combined by cells in the superficial and deep layers of V1. It is thought that figure–ground segregation occurs through recurrent interactions with higher visual areas (shown by the yellow arrows). Cells in the deep layers of V1 and the higher visual areas send feedback projections back to the LGN (orange arrows) and are likely the source of the difference in BOLD signal magnitude between the congruent and incongruent patches observed in this study.
participants to attend to one patch and monitor the texture for a change in spatial frequency content and ignore changes at the opposite location. In this way, the authors could examine the responses of attended and unattended texture patches. As neurons in the LGN do not represent orientation, they should be blind to the differences between the two texture patches, which differed only in the relative orientation compared to the background. Furthermore, the receptive fields of LGN neurons are small and the difference between congruent and incongruent texture patches was determined by image regions outside the texture patch, far from the receptive field. Hence, the feedforward input from the retina to the LGN neurons was held constant.
The results, however, showed unambiguously that voxels in the LGN representing the incongruent patch had higher BOLD signals than those representing the congruent patch. Where does this signal come from? To test this, Poltoratski et al. [7] took advantage of a quirk of the anatomical organization of the visual system. Signals from the two eyes are processed by different layers within the LGN and hence the neural responses there are purely monocular. The authors presented the figures to one eye and the background to the other eye using anaglyph (red/green) glasses (Figure 1). This manipulation removes any possibility of interactions between the texture patches and their surround within the LGN itself. The results, however, remained remarkably similar to those of the first experiment: the incongruent
Current Biology 29, R568–R591, June 17, 2019 R575
Current Biology
Dispatches texture patch produced higher BOLD signals in the LGN compared to the congruent patch. In this version of the experiment, both patches were unattended as attention was directed to a central fixation task and the results were replicated for two different central tasks of different difficulty levels. The results provide strong evidence that the difference in signal between the congruent and incongruent patch must be due to feedback from a brain region which combines information across the two eyes. As binocular integration occurs first in the superficial layers of V1, this implicates cortex as the source of the signals. This feedback signal from the cortex to the LGN that enhances the figure representation even occurs when the figure is ignored. The paper of Poltoratski et al. [7] adds to a growing list of studies which have found modulations of activity in the LGN for processes that were previously thought to be exclusively cortical in nature. Neural activity in the LGN is modulated by spatial attention [8] and brain imaging studies have even found modulations of LGN BOLD signals related to orientation [9] and binocular rivalry [10,11]. One potential caveat concerns the use of fMRI to infer changes in neural activity in the LGN. The BOLD signal is particularly sensitive to synaptic inputs and, as corticogeniculate inputs to the LGN actually outnumber the synaptic inputs received from retinal ganglion cells [5], it is likely that the BOLD signal in the LGN is dominated by feedback signals. It is impossible to tell using non-invasive techniques whether these modulatory inputs actually give rise to changes in spiking activity in the LGN. Nevertheless, the new study [7] shows that figure– ground information is fed-back to the LGN and, given a previous monkey study [6] which did demonstrate an enhanced spiking response on figures, this suggests that LGN spiking activity is modulated by figure–ground structure in humans as well. The work of Poltoratski et al. [7] moves beyond previous studies by showing an enhanced representation of figures that fall outside the focus of attention. Typically in primate studies of figure–ground modulation, the figure is also the target for a saccadic eyemovement and is associated with reward during many months of behavioral training
(though see [3,12,13]). This has led to the view that figure–ground modulation can be seen as a form of object-based attentional selection. The new results [7] provide support for an alternative viewpoint in which potential figure-regions are enhanced in parallel across the visual scene in a relatively automatic fashion. These unattended figure representations have been referred to as ‘proto-objects’ [12,14] and we have recently demonstrated that the neural representations of unattended protoobjects are enhanced in the primary visual cortex in monkeys [12]. In proto-object theory, the visual system utilizes the statistics of objects in natural scenes (for example, the Gestalt grouping cues) to enhance the representations of visual features which are likely to belong to objects. The study of Poltoratski et al. [7] indicates that cortex enhances the representations of proto-objects through recurrent loops with the LGN. Objectbased attention can then act upon these proto-objects to select and group visual features into coherent object representations [4,15,16]. Poltoratski et al. [7] link the parallel enhancement of both attended and nonattended figures to the predictive-coding framework [17,18]. According to this view, the uniform orientation of the background elements provides a predictable spatial context, which is sent to early visual areas to ‘explain away’ the predictable background elements. The figure elements, having an unpredictable orientation, produce an error signal that is propagated through to higher areas. While the results are broadly in line with this theory, it should be noted that it is impossible to determine whether background activity is being suppressed using fMRI and previous studies have demonstrated that feedback connections also enhance representations that are predicted by the higher areas, which goes against the predictive coding framework [3,19]. Furthermore, the predictions of such a predictive coding depend on spatial scale. Poltoratski et al. [7] interpret the enhanced activity for the figure in the LGN as evidence that a relatively global contextual feedback signal is sent back to the thalamus, as the orientation of the figure elements is only unpredictable at large spatial scales. At a finer scale,
R576 Current Biology 29, R568–R591, June 17, 2019
however, the central figure elements can be predicted from nearby elements of the same patch, which should lead to a relative suppression of the center of objects. Feedback connections exert both excitatory and inhibitory effects on predicted information at lower processing levels [20], and it is likely that this topic will remain an active area of research in the coming years. Poltoratski et al. [7] have provided strong evidence that cortex does send feedback back to the LGN to help organize our perception, but leaves us with the tantalizing puzzle of what the function of this feedback ultimately is. Is it a predictive template that gates incoming information in the thalamus to improve coding efficiency or a signal that boosts potentially interesting regions of the visual scene to aid their selection by attentional mechanisms? Whatever the result, this study provides a fascinating addition to our knowledge of the mechanisms of perceptual organization. REFERENCES 1. Lamme, V.A. (1995). The neurophysiology of figure-ground segregation in primary visual cortex. J. Neurosci. 15, 1605–1615. 2. Lamme, V.A., and Roelfsema, P.R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends Neurosci. 23, 571–579. 3. Poort, J., Raudies, F., Wannig, A., Lamme, V.A., Neumann, H., and Roelfsema, P.R. (2012). The role of attention in figure-ground segregation in areas V1 and V4 of the visual cortex. Neuron 75, 143–156. 4. Jeurissen, D., Self, M.W., and Roelfsema, P.R. (2016). Serial grouping of 2D-image regions with object-based attention in humans. eLife 5, e14320. 5. Briggs, F., and Usrey, W.M. (2011). Corticogeniculate feedback and visual processing in the primate. J. Physiol. 589, 33–40. 6. Jones, H.E., Andolina, I.M., Shipp, S.D., Adams, D.L., Cudeiro, J., Salt, T.E., and Sillito, A.M. (2015). Figure-ground modulation in awake primate thalamus. Proc. Natl. Acad. Sci. USA 112, 7085–7090. 7. Poltoratski, S., Maier, A., Newton, A.T., and Tong, F. (2019). Figure-ground modulation in the human lateral geniculate nucleus is distinguishable from top-down attention. Curr. Biol. 29, 2051–2057. 8. McAlonan, K., Cavanaugh, J., and Wurtz, R.H. (2008). Guarding the gateway to cortex with attention in visual thalamus. Nature 456, 391–394.
Current Biology
Dispatches 9. Ling, S., Pratte, M.S., and Tong, F. (2015). Attention alters orientation processing in the human lateral geniculate nucleus. Nat. Neurosci. 18, 496–498.
13. Marcus, D.S., and Van Essen, D.C. (2002). Scene segmentation and attention in primate cortical areas V1 and V2. J. Neurophysiol. 88, 2648–2658.
10. Haynes, J.-D., Deichmann, R., and Rees, G. (2005). Eye-specific effects of binocular rivalry in the human lateral geniculate nucleus. Nature 438, 496–499.
14. von der Heydt, R. (2015). Figure-ground organization and the emergence of protoobjects in the visual cortex. Front. Psychol. 6, 1–10.
11. Wunderlich, K., Schneider, K.A., and Kastner, S. (2005). Neural correlates of binocular rivalry in the human lateral geniculate nucleus. Nat. Neurosci. 8, 1595–1602.
15. Rensink, R.A. (2000). The dynamic representation of scenes. Vis. Cogn. 7, 17–42.
12. Self, M.W., Jeurissen, D., van Ham, A.F., van Vugt, B., Poort, J., and Roelfsema, P.R. (2019). The segmentation of proto-objects in the monkey primary visual cortex. Curr. Biol. 29, 1019–1029.
16. Roelfsema, P.R. (2006). Cortical algorithms for perceptual grouping. Annu. Rev. Neurosci. 29, 203–227. 17. Rao, R.P., and Ballard, D.H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical
receptive-field effects. Nat. Neurosci. 2, 79–87. 18. Lee, T.S., and Mumford, D. (2003). Hierarchical Bayesian inference in the visual cortex. J. Opt. Soc. Am. A. Opt. Image Sci. Vis. 20, 1434– 1448. 19. Chelazzi, L., Miller, E.K., Duncan, J., and Desimone, R. (2001). Responses of neurons in macaque area V4 during memory-guided visual search. Cereb. Cortex 11, 761–772. 20. Zhang, S., Xu, M., Kamigaki, T., Phong Hoang Do, J., Chang, W.-C., Jenvay, S., Miyamichi, K., Luo, L., and Dan, Y. (2014). Long-range and local circuits for top-down modulation of visual cortex processing. Science 345, 660–665.
Zoology: Worming into the Origin of Bilaterians taz Ferdinand Marle Molecular Genetics Unit, Okinawa Institute of Science & Technology, Onna-son, Japan Correspondence: [email protected] https://doi.org/10.1016/j.cub.2019.05.006
Xenacoelomorphs, a group of worms with simple body organization, have been proposed to represent the first offshoot of bilaterians. A new study shows that they might instead belong to the deuterostomes, just as echinoderms and vertebrates. The nature of the last common ancestor of all bilaterally symmetrical animals — the urbilaterian — is one of the key questions in zoology, partly because it reflects on the origin of the key organ systems that make us who we are. There are two opposing views of what this ancestor may have been like — either it was a rather simple organism or a fairly complex one. And, a lot of the debate surrounding this dichotomy revolves around the phylogenetic position of a handful of worms — the xenacoelomorphs. Two alternative positions were successively proposed over the years: one suggests that this group of morphologically simple worms is the sister-group to all other bilaterians. This branching would mean that the bilaterian ancestor was a rather unsophisticated organism. Alternatively, other researchers have proposed that these xenacoelomorphs instead are a derived offshoot of deuterostomes, one of the two main clades of bilaterians, which mean that their simplicity derives from secondary character loss. A new study
Philippe, Max Telford and by Herve colleagues [1] in Current Biology brings new elements to support this latter tree, in particular a thorough evaluation of methodological bias that can affect phylogenetics reconstruction — such as the famous long-branch attraction. They conclude that xenacoelomorphs are the sister-group of a clade called ‘Ambulacraria’, which includes sea stars, sea urchins and acorn worms. The idea that the urbilaterian was a simple organism without a body cavity (coelom) reminiscent of the planula larva of some cnidarians was mentioned a long time ago, for instance in the famous zoology textbook of Libbie Hyman [2]. At this time, morphologically simple animals, such as nematode roundworms and flatworms, were considered as early branches of the animal tree of life. Morphological cladistics and molecular phylogeny then helped redefine our understanding of animal evolution by splitting all bilaterian animals into two main clades: the protostomes (including
insects, molluscs, annelids, flatworms) and the deuterostomes (including vertebrates, tunicates, sea urchins and sea stars). This reclassification implied that seemingly simple lineages may have originated through simplification and secondary character losses from a more complex coelom-bearing urbilaterian ancestor [3]. Then, in the late nineties, a study [4] pointed out that a neglected lineage of flatworms — the acoels — might have represented a sister-group to both protostomes and deuterostomes. This reignited the debate concerning the nature of the bilaterian ancestor and propelled these overlooked animals to the front stage of animal evolution. Acoel flatworms indeed show a very simple planula-like organization and do not possess structures that are found in other bilaterians, such as a coelomic cavity, excretory organs, or nerve chords (Figure 1A). All bilaterians that are not acoels were dubbed ‘Nephrozoa’ (highlighting the presence of an excretory
Current Biology 29, R568–R591, June 17, 2019 ª 2019 Elsevier Ltd. R577