Interneurons take charge

The brain's hippocampal region contains many classes of interneurons, which, it transpires, show different patterns of activity. They might contribute to memory by shaping the dynamics of neuronal networks.

Learning something new — and recalling what you've learned — involves large numbers of nerve cells, distributed throughout different parts of the brain. Neuronal assemblies in the hippocampus, for instance, are thought to be essential in encoding, consolidating and retrieving memories. These neuronal networks alternate between several functional states, each characterized by a temporally structured pattern of electrical activity that may facilitate a particular aspect of memory processing. On page 844 of this issue, Klausberger et al.1 present results that bring us closer to understanding the precise but varied ways in which distinct classes of hippocampal neurons impose such network states. Their study focuses on interneurons — small neurons that are involved in the local processing of nerve signals. They find that these cells, which generally have inhibitory activities, seem to work together in a strictly coordinated manner to control assemblies of excitatory pyramidal nerve cells.

In particular, the new results1 show that the dynamics of hippocampal networks are related to the diversity in the hippocampal interneuron population. Pyramidal cells are relatively uniform in their structure and behaviour. But interneurons form distinct classes, in terms of their shape, the inputs to which they respond, the other neuron populations they connect to, and the specific parts of neurons with which they make contact2,3,4. Klausberger et al. now show that morphologically distinct classes of hippocampal interneurons also contribute differently to network states.

The authors recorded the electrical impulses (spikes) generated by three types of hippocampal interneurons — basket cells, axo-axonic cells, and oriens–lacunosum-moleculare (O-LM) cells — during two network states in anaesthetized rats. One of these states is characterized by theta oscillations (which have a frequency of 4–10 Hz), the other by short ripple oscillations (120–200 Hz) on a background of more irregular activity. In conscious animals, theta oscillations are associated with movement and exploration, whereas ripples are associated with inactivity and slow-wave sleep5.

The three classes of interneurons exhibited distinct, state-dependent patterns of activity (Fig. 1). For instance, during theta oscillations, basket cells fired on the descending phase of each wave, axo-axonic cells just after the peak, and O-LM interneurons at the trough. During ripple oscillations, basket cells discharged one or more phase-locked spikes; axo-axonic cells fired at the beginning of each ripple sequence; and O-LM cells became totally silent. The implication is that the three classes of interneurons, through their characteristic patterns of activity, make specific contributions to the production of hippocampal network states.

Figure 1: Interneurons and electrical oscillations.

The figure shows the activity profiles of three types of hippocampal interneuron during two brain states, based on the findings of Klausberger et al.1. Colours indicate the probability that a given interneuron will fire (maximum red, minimum blue). During theta oscillations, basket cells fire on the descending phase of local theta waves, axo-axonic cells fire just after the peak, and O-LM cells fire at the trough. During ripple oscillations, basket cells discharge one or several phase-locked spikes, axo-axonic cells fire only at the beginning of the ripple sequence, and O-LM cells become silent. The variation within each group is small, suggesting that classes of interneurons exert precise control over distinct aspects of hippocampal network dynamics.

It remains to be seen, however, exactly what these contributions are. For example, what is the function of the sudden drop in activity of O-LM cells during ripple oscillations? As interneurons generally inhibit other nerve cells, this drop in activity means a drop in inhibition. Axons from O-LM cells target the outermost portion of pyramidal-cell dendrites — the area in which pyramidal cells receive excitatory inputs from another brain region, the entorhinal cortex, which mediates highly processed sensory information to the hippo-campus. So does the drop in inhibition amplify the cortical input? Does it enable the excitatory synapses (connections) between entorhinal cells and hippocampal cells to become modified, and do such changes occur specifically during ripple oscillations? During ripples, pyramidal cells tend to fire in patterns that are reminiscent of their firing during recent awake behaviour6. Such 'reactivation' may contribute to memory consolidation in the hippocampus and the neocortex6, where the permanent storage of memories is thought to take place7. Are O-LM cells involved in this process?

Moreover, what is the function of the precisely timed firing of basket cells and axo-axonic cells during ripple and theta oscillations? Does it synchronize the output from hippocampal pyramidal cells to the neocortex8? And, if so, how do the neocortical target neurons respond to such output?

To determine how the interneurons' precise control of network dynamics contributes to hippocampal functions such as memory processing9, the analysis must be extended to behavioural studies in conscious animals. Functional diversity in the hippocampal networks of conscious rats was first studied 30 years ago10, when nerve impulses recorded in the hippocampus were observed to originate from either 'complex-spike cells' or 'theta cells'. The complex-spike cells discharged at low rates but in bursts; theta cells were more active but their spikes were also more dispersed. These two types of neurons had the properties expected of pyramidal cells and interneurons, respectively. Their identity was eventually verified by staining the recorded neurons in anaesthetized rats11; this confirmed that spike parameters and spike patterns can indeed predict the morphology of hippocampal neurons.

The results of Klausberger et al.1 show that a similar approach can be used to distinguish between classes of interneurons. The spike activity generated by interneurons in a given class varies little, suggesting that state-dependent activity patterns provide valid signatures of basket cells, axo-axonic cells and O-LM cells in anaesthetized rats. Can these differences be extrapolated to awake animals? Similar profiles of activity have been observed during corresponding network states in anatomically unidentified hippocampal interneurons in conscious rats12. Obviously, there are differences between the awake and anaesthetized conditions13, but the new results represent a first step towards a physiological classification scheme that could be used to relate variations in activity profiles to particular interneurons and to specific memory operations14.

For these results to be applied to behavioural studies, other types of hippocampal interneurons must be analysed in a similar way. The criteria that distinguish between basket cells, axo-axonic cells and O-LM cells in the study by Klausberger et al. might be less useful in samples that also contain other interneuron classes. It may be necessary to explore additional parameters, such as waveform shape — which is known to distinguish pyramidal cells from some interneurons10,11 — as well as other network states and the cells' responses to specific drugs.

A physiologically based classification system for analysing behaviour represents one of two developments that should advance our understanding of hippocampal interneurons. The other is the genetic manipulation of mice, which may soon allow scientists to examine neuronal networks and memory processes when the function of particular interneurons has been genetically altered. Together, these new tools have the power to uncover the most fundamental principles of memory formation in the neuronal assemblies of the hippocampus.


  1. 1

    Klausberger, T. et al. Nature 421, 844–848 (2003).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Buhl, E. H., Halasy, K. & Somogyi, P. Nature 368, 823–828 (1994).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Freund, T. F. & Buzsáki, G. Hippocampus 6, 347–470 (1996).

    CAS  Article  Google Scholar 

  4. 4

    McBain, C. J. & Fisahn, A. Nature Rev. Neurosci. 2, 11–23 (2001).

    CAS  Article  Google Scholar 

  5. 5

    Buzsáki, G. Neuroscience 31, 551–570 (1989).

    Article  Google Scholar 

  6. 6

    Wilson, M. A. & McNaughton, B. L. Science 265, 676–679 (1994).

    ADS  CAS  Article  Google Scholar 

  7. 7

    Squire, L. R. & Alvarez, P. Curr. Opin. Neurobiol. 5, 169–177 (1995).

    CAS  Article  Google Scholar 

  8. 8

    Cobb, S. R., Buhl, E. H., Halasy, K., Paulsen, O. & Somogyi, P. Nature 378, 75–78 (1995).

    ADS  CAS  Article  Google Scholar 

  9. 9

    Paulsen, O. & Moser, E. I. Trends Neurosci. 21, 273–278 (1998).

    CAS  Article  Google Scholar 

  10. 10

    Ranck, J. B. Jr Exp. Neurol. 41, 461–531 (1973).

    Article  Google Scholar 

  11. 11

    Henze, D. A. et al. J. Neurophysiol. 84, 390–400 (2000).

    CAS  Article  Google Scholar 

  12. 12

    Csicsvari, J., Hirase, H., Czurko, A., Mamiya, A. & Buzsáki, G. J. Neurosci. 19, 274–287 (1999).

    CAS  Article  Google Scholar 

  13. 13

    Fox, S. E. & Ranck, J. B. Jr Exp. Brain Res. 62, 495–508 (1986).

    CAS  Article  Google Scholar 

  14. 14

    Fyhn, M., Molden, S., Hollup, S., Moser, M. B. & Moser, E. I. Neuron 35, 555–566 (2002).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Edvard I. Moser.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Moser, E. Interneurons take charge. Nature 421, 797–799 (2003).

Download citation

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.