Each synapse to its own

Article metrics

A neuron can receive thousands of inputs that, together, tell it when to fire. New techniques can image the activity of many inputs, and shed light on how single neurons perform computations in response.

Synaptic communication between neurons is fundamental to how the brain processes and transforms information. Uncovering the neural circuitry has therefore been a major endeavour for neuroscientists, to understand the neural basis of perception and action. In this issue (page 1307), Jia et al.1 present one of the most detailed views of synaptic integration within the brain so far, addressing a classic and vigorously debated question in sensory system physiology: how do neurons in the primary visual cortex acquire orientation selectivity?

For the past two decades, mammalian brain slices have been an essential tool for studying neural circuitry. Slice experiments have revealed the functional diversity of synapses, synaptic integration within the neuronal branches that constitute a dendritic tree, how voltage-gated currents in the dendrites shape integration, and how activity-dependent changes in synaptic strength might contribute to memory formation and to development. But the physiological model that the brain slice provides comes at a cost, as natural neural connections are lost when the slice is prepared. It has therefore been difficult to determine which of the above features are crucial to the function of neurons in the intact brain, and how they would interact in response to naturally driven activity.

Intracellular electrode recordings from the intact brain have provided extensive information about the aggregate synaptic input that reaches a cortical neuron in response to visual stimulation. Jia et al.1, however, now use optical recordings to reveal the visual selectivity of single synaptic inputs to cortical neurons and — their most striking finding — the distribution within the dendritic tree of inputs of different orientation preferences.

To achieve this, the authors used an intracellular electrode to fill individual cortical neurons with a Ca2+-sensitive dye and then used two-photon microscopy to image the Ca2+ signals evoked by simple visual stimuli of different orientations. They observed rapid, localized signals throughout the dendritic tree that arise from Ca2+ entry through N-methyl-D-aspartate (NMDA)-type glutamate receptors. These 'hot spots' were sparsely distributed and probably represent the activity at single synapses.

Previous work in the cat visual cortex2 has shown that the firing rate of action potentials and the membrane-potential responses recorded in the cell body of a neuron may be strongly selective for a single orientation. That is, some neurons respond vigorously to vertical contours and not at all to horizontal contours; other neurons respond to horizontal stimuli; and still others to oblique stimuli.

Jia et al. find that the individual synaptic inputs that excite a neuron are also strongly orientation selective. But these inputs vary markedly in their orientation preference relative to one another and relative to the orientation of the cell as a whole. These diverse inputs sum linearly to create an aggregate input that is only weakly selective for orientation.

The preferred orientation depends on which synaptic inputs happen to predominate. This weak bias in the aggregate input is then filtered through the spike threshold: only the largest responses, those evoked by stimuli very near the preferred orientation, carry the membrane potential across the threshold and trigger the neuron to spike. Similar in principle to what has been found from intracellular recordings in the cat, the range of orientations that trigger the spike output is much narrower than the range of orientations that generates synaptic input3,4. The notable finding here, however, is that the broadly tuned total synaptic input seems to arise not from a set of broadly selective individual synaptic inputs with similar preferences, but from a set of sharply tuned inputs with diverse preferences (Fig. 1).

Figure 1: Synapses of an orientation-selective neuron.

Three models of dendritic organization can account for orientation-selective neuronal responses in the primary visual cortex. a, According to a weak-bias model, synaptic inputs along all dendrites are broadly tuned for orientation, but each provides the same bias in orientation preference. b, Synaptic inputs could also be narrowly tuned for a specific orientation along each dendrite, but vary in orientation preference from dendrite to dendrite. c, Jia et al.1 show that, although the synaptic inputs along the dendrites are narrowly tuned for orientation and vary from one to another in orientation preference, there is no organization in orientation preference along each dendrite.

Because the authors were imaging the entire dendritic tree and measuring the orientation selectivity of each individual input, they could observe directly whether the synaptic inputs were grouped anatomically within the dendritic tree according to their individual orientation preferences. One might expect that inputs with similar orientations would be co-localized onto the same dendrite (Fig. 1b). Indeed, this type of physical segregation would be useful if voltage-gated currents in the dendrites were to perform local, nonlinear computations on the inputs to a single dendrite, before passing on the results to the cell body5. Jia et al. found no evidence for such dendritic computational units. Instead, they report that the inputs of similar orientation preference are scattered throughout the dendritic tree, with no obvious organization (Fig. 1c). The 'computational unit' in these neurons seems to be the whole dendritic tree, rather than individual dendrites.

These results1 also speak to the rules governing cortical development. An elegant series of two-photon experiments6,7 has shown that, in rodents, neurons with different orientation preferences are randomly intermingled. The apparently random arrangement of inputs that Jia et al. describe could therefore arise if each cortical neuron sampled synaptic inputs randomly from its near neighbours, rather than picking and choosing among different inputs according to its ultimate orientation preference. If this were so, during development there would be no need for 'Hebbian plasticity' and instead neurons could wire together whether they fired together or not.

In contrast to the mouse, in the cat and in primates a simple, proximity-based connection scheme would have very different results. There, neurons with different orientation preferences are clustered into orientation columns. Except at 'pinwheels', where many columns of different orientations meet, neurons that connected indiscriminately to their near neighbours would receive inputs that were closely matched in orientation. And indeed, the aggregate synaptic input to cortical neurons in the cat is more sharply tuned for orientation than Jia et al. observe in the mouse. But it is worth noting that, at least in the cat, long-range connections (those that cross orientation columns) are arranged by orientation8. That local connections obey different rules would therefore be a surprising result.

All of Jia and colleagues' findings1 relate to questions that lie at the heart of cellular and systems neuroscience: what are the rules of neuronal connectivity, how do neurons become selective for stimulus features, and do dendrites function as discrete computational units? They will surely spark many additional studies. But whatever the outcome, these methods, and others like them, are taking the analysis of the brain to unprecedented levels.


  1. 1

    Jia, H., Rochefort, N. L., Chen, X. & Konnerth, A. Nature 464, 1307–1312 (2010).

  2. 2

    Creutzfeldt, O. D., Kuhnt, U. & Benevento, L. A. Exp. Brain Res. 21, 251–274 (1974).

  3. 3

    Priebe, N. J. & Ferster, D. Neuron 57, 482–497 (2008).

  4. 4

    Carandini, M. & Ferster, D. J. Neurosci. 20, 470–484 (2000).

  5. 5

    Archie, K. A. & Mel, B. W. Nature Neurosci. 3, 54–63 (2000).

  6. 6

    Sohya, K., Kameyama, K., Yanagawa, Y., Obata, K. & Tsumoto, T. J. Neurosci. 27, 2145–2149 (2007).

  7. 7

    Ohki, K., Chung, S., Ch'ng, Y. H., Kara, P. & Reid, R. C. Nature 433, 597–603 (2005).

  8. 8

    Gilbert, C. D. & Wiesel, T. N. Nature 280, 120–125 (1979).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.