Hongbo Jia, Nathalie L. Rochefort, Xiaowei Chen & Arthur Konnerth
In sensory cortex regions, neurons are tuned to specific stimulus features. For example, in the visual cortex, many neurons fire predominantly in response to moving objects of a preferred orientation. However, the characteristics of the synaptic input that cortical neurons receive to generate their output firing pattern remain unclear. Here we report a novel approach for the visualization and functional mapping of sensory inputs to the dendrites of cortical neurons in vivo. By combining high-speed two-photon imaging with electrophysiological recordings, we identify local subthreshold calcium signals that correspond to orientation-specific synaptic inputs. We find that even inputs that share the same orientation preference are widely distributed throughout the dendritic tree. At the same time, inputs of different orientation preference are interspersed, so that adjacent dendritic segments are tuned to distinct orientations. Thus, orientation-tuned neurons can compute their characteristic firing pattern by integrating spatially distributed synaptic inputs coding for multiple stimulus orientations.
This is a neat paper with a lovely result about the lack of dendritic organization of inputs of different tuning directions. The depth of data analysis, however, seems a bit weak. For example, do the hotspots that shared the recorded neuron’s tuning properties tend to occur closer to the soma than the non-co-tuned hotspots? Of course, physical distance is an imperfect proxy for electrotonic distance, but it’s an easy analysis to do with the dataset in hand, and maybe it would turn up something. Seems like there could have been more analysis of these types of questions.
Agree or Disagree
5 
1
Impressive paper. Using a combination of electrophysiology and Ca imaging in vivo, the authors are able to map hotspots of synaptic activity elicited by visual input (drifting gratings) onto the dendritic tree. They found that hotspots with the same orientation tuning were distributed across the dendritic arbor. This result was striking and surprising because many people expected that inputs with the same orientation tuning would cluster together, to better evoke APs, perhaps by generating dendritic spikes.
However, I am not convinced of one of the authors main conclusions: that there is no clustering of inputs. Sure they show that hotspots with the same orientation tuning are distributed. But what exactly is a “hotspot”? Elevated Ca as measured by a OGB-1, a high affinity Ca sensor. That does not mean that a hotspot = a single synapse. Ca-imaging does not have single synapse specificity. So how do we know that a single hotspot does not correspond to a cluster of multiple adjacent synapses with the same orientation tuning? We don’t. The authors provide some weak evidence that the Ca signal they see is similar to what can be produced from a single synapse, but this very important possibility is not directly tested.
So a very interesting paper but I do not feel they provide convincing evidence that there is no clustering of synaptic inputs with similar orientation tuning.
What do you think?
-Yurut
Agree or Disagree
7 
0
Other observations/questions about this paper:
In Fig3 they show hotspots mapped onto the dendritic tree of one cell. All the hotspots are within about 60uM of the soma, which was the imaging area. Are there hotspots in the distal dendrites?
I wonder about the distribution of hotspots with the same orientation selectivity. OK I believe them when they say they are not clustered but with the examples they give, they make it seem like they are evenly spread throughout the dendritic tree, that is spatially distributed, perhaps non-randomly. One would expect if they were placed randomly, some would wind up next to each other by chance. Seems like they could test that if their data set is large enough.
-Yurut
Agree or Disagree
2 
0
Does anyone know what computation occurs at L2/3 of the primary visual cortex? How does this finding fit in with what know about what L2/3 neurons are doing to visual information?
Agree or Disagree
0 
0
I agree with Dr. Lu: this is a beautiful paper with a surprising a finding. The recent description of local Hebbian mechanisms at approximately the spine level makes a strong prediction for clustering of functionally related inputs. The fact that they find distributed orientation tuning along a dendrite suggests that either their resolution is too coarse to detect functional clustering, or the prediction for clustering is wrong. The authors suggest that species with a more columnar organization of visual cortex might show clustering, I wonder what they would find in whisker cortex?
Agree or Disagree
0 
0
..they could have asked whether the average input a given neuron gets (as evaluated by average tuning of the hotspots) corresponds to the total neuron tuning. E.g. look at the example neuron in fig. 3. It has vertical tuning. Now look at the hotspots. 13 hotspots, none of which are for pure horizontal tuning. 6 are for vertical and 7 for diagonal. So maybe if you averaged them together, you’d get more or less the vertical tuning of the cell? It seems like another analysis that would be easy to run on all their neurons, really would like to see it.
Agree or Disagree
0 
0
I’ve only skimmed the paper and read the news & views. But, are p28 mice adults? I don’t know enough about the species to know. And, I’m unaware of how much is known about the physiology/anatomy of orientation selectivity in the mouse. There’s a big literature in the cat, and for the cat, this would be a surprising result indeed. But what does orientation selectivity look like in a mouse? How broad are the tuning widths?
I ask those questions, ’cause it makes me wonder about the generalization of the results they’re discussing. Is the apparent lack of “clustering” species specific, and does it result from the specific physiology of orientation tuning in the mouse?
And, for “curious” there’s no clear answer to what “computation occurs at L2/3 of primary visual cortex.” generally, and almost nothing in the mouse, to my awareness. Someone correct me if I’m wrong.
Agree or Disagree
0 
0
neurolover: I don’t know much about this paper in question, but I’m a mouse researcher. p28 is very young, between periadolescence and adolescence at most. To give you a point of reference, mice are weaned at p21, and young adulthood is often thought of as p60, though some believe adulthood to be later than that.
Agree or Disagree
2 
3
The comment by jhb has been removed at the request of the author.
Agree or Disagree
0 
9
This paper highlights one of the most significant challenges in our field (and others): publishing exciting and topical data in high impact journals necessarily results in the presentation of only a limited subset of data and experiments. For a study that is being employed to “answer” a debate within the dendritic integration community, it is quite sparse in terms of the amount of data shown. Given, the experiments are technically challenging, but as a number of commentators pointed out above, it certainly seems like there were analyses and/or further experiments that could have been conducted relatively easily. Unfortunately, readers are left with a plethora of unanswered questions (and not necessarily the good kind that a comprehensive study routinely provokes), complicating the interpretation of an already convoluted subject.
Dr. Lu raises a number of relevant technical concerns in his comments above, to which I would add two points that further problematize the presented data in light of the clustering/non-linear integration debate. First – these animals are anesthetized, which has been shown to affect a variety of conductances (Ih in particular) as well as inhibit GABAergic transmission, both of which are of course critical for processing in the cortical microcircuit. Within this paradigm, the investigators are presumably forced to use these 1-second presentations of visual stimulus in order to resolve their Ca2+-signals, which may be totally irrelevant in an awake animal with “intact” corticothalamic activity. Second – in order to isolate dendritic “hotspots”, the authors inject hyperpolarizing current through the soma to prevent axonal AP generation, which is going to significantly alter the balance of membrane conductances, limiting the initiation of any dendritic non-linearities.
Given these issues, as well as those raised previously in the comments, its hard to evaluate the claims in the paper (and the accompanying article by Ferster and Priebe) regarding clustering and non-linear integration. Instead, I think the most parsimonious interpretation comes from what is shown more conclusively in terms of “functional anatomy” – there are representations of different orientations along individual branches in L2 pyramids rather than a restricted whole-branch tuning schema. How the different tuning orientations interact along individual branches and with other branches at the axon in awake behaving animals remains to be seen.
Agree or Disagree
1 
0