Astrocytes as aide-mémoires

Memory formation is known to occur at the level of synaptic contacts between neurons. It therefore comes as a surprise that another type of brain cell, the astrocyte, is also involved in establishing memory.

Memory is the result of long-lasting changes in synaptic activity usually involving the activation of NMDA receptors (NMDARs) — a special class of receptor for the excitatory neurotransmitter glutamate. Memory formation has always been thought to depend on events occurring exclusively in neurons. But the brain possesses another cell population, glial cells, which include the highly ramified, star-shaped astrocytes. Despite their abundance — they make up 90% of all human brain cells — astrocytes have been relatively overlooked in the search for mechanisms of memory formation because they lack electrical excitability and do not communicate like neurons do. But astrocytes are not silent; they display another type of excitability involving changes in the intracellular concentration of calcium ions (Ca2+). In this issue (page 232), Henneberger et al.1 show that one function of the increase in astrocyte Ca2+ is to trigger the release of molecules required for establishing synaptic memory.

Previous studies have observed2 that the processes extending from different astrocytes do not overlap; that is, each astrocyte occupies an exclusive territory in the brain. This observation forms the basis of the concept2,3 that each astrocyte territory represents an island made up of many thousands of synapses (about 140,000 in the hippocampal region of the brain, for instance) whose activity is controlled by that astrocyte. Henneberger et al.1 provide the first direct evidence for this proposal. The authors induce long-term potentiation (LTP) of excitatory synapses in the hippocampus using a high-frequency-stimulation protocol, which involves applying repetitive electrical stimuli to the presynaptic fibres. LTP is the sustained increase in synaptic strength associated with memory formation, and the authors monitored this synaptic potentiation locally, in domains roughly corresponding to the territories of individual astrocytes. They did this by recording the electrical signal generated by the ensemble of synapses in the territory, using an extracellular electrode or, alternatively, directly through the astrocyte.

Henneberger and colleagues cleverly use a pipette to record synaptic activity and simultaneously manipulate the cytosol of an individual astrocyte by introducing Ca2+-buffering agents to prevent (clamp) any increase in intracellular Ca2+. Surprisingly, this manoeuvre abolishes the induction of LTP at the surrounding synapses. Moreover, when the authors clamp only one of two neighbouring astrocytes, LTP induction is prevented exclusively at the synapses in the territory of the astrocyte whose rise in intracellular Ca2+ is blocked. It occurs normally at synapses in the territory of the non-clamped astrocyte, provided this cell is at least 200 μm away (Fig. 1, overleaf).

Figure 1: Astrocytes and synaptic memory.

Henneberger and colleagues' single-cell experiments1 demonstrate that a rise in astrocyte intracellular Ca2+ controls the induction of long-term potentiation (LTP) at nearby synapses. The authors induce LTP at hippocampal synapses using high-frequency stimulation (HFS). With a pipette, they load individual astrocytes with a control solution (top cell) or a buffer that prevents an increase in intracellular Ca2+ (bottom 'clamped' cell). After HFS, intracellular Ca2+ levels increase in the top astrocyte, which releases D-serine. D-Serine binds to NMDA receptors to promote LTP establishment when glutamate is released from the presynaptic terminal. Preventing the rise in intracellular Ca2+ in the bottom astrocyte abolishes D-serine release and prevents LTP establishment in its territory, but does not affect LTP in the top astrocyte's domain provided that the two cells are at least 200 μm apart.

LTP can be rescued in the clamped astrocyte territory by adding the amino acid D-serine. D-Serine interacts with the so-called glycine-binding site of the NMDAR, permitting its transmembrane channel to open when glutamate binds. Previous studies in astrocyte cultures and hypothalamic slices had suggested that astrocytes can release D-serine through Ca2+-dependent exocytosis4 (release of substances stored in vesicles), implicating these cells as a possible source of D-serine acting at synaptic NMDARs5.

To investigate whether D-serine has a similar effect in the hippocampus, Henneberger et al.1 first showed that clamping Ca2+ in an astrocyte produces a 25% reduction in the NMDAR current in the surrounding hippocampal synapses. This reduction is reversed by the addition of D-serine, which confirms a causal link between astrocyte Ca2+ elevation and occupancy of the NMDAR glycine-binding site.

Intriguingly, if D-serine is added without clamping astrocyte Ca2+, the NMDAR current is enhanced, but if LTP is induced without clamping astrocyte Ca2+, supplying D-serine does not increase synaptic potentiation. The authors propose that these results can be explained by differences in the level of astrocyte activation in the two cases. Indeed, the high-frequency stimulation that induces LTP is a strong synaptic stimulus that also vigorously stimulates a rise in astrocyte Ca2+, leading to massive D-serine release and transient saturation of the NMDAR glycine-binding site. By contrast, the synaptic stimulation used to study NMDAR currents is much milder, and in this case astrocytes show only infrequent oscillations in intracellular Ca2+. These oscillations are probably associated with the release of smaller amounts of D-serine, which only partially saturate the NMDAR glycine-binding site.

Do astrocytes release D-serine directly or do they secrete other substances that trigger D-serine release from neurons6? And what is the mechanism of substance release? Henneberger et al.1 address the first question by introducing a blocker of serine racemase, the enzyme responsible for D-serine synthesis, into the astrocyte. They find that this agent inhibits LTP, implying a causal link between astrocyte D-serine synthesis and local LTP induction. However, the serine racemase inhibitor also affects pyruvate production, which can perturb intra-astrocytic glutamate levels. Therefore, we are left with the possibility that glutamate release from astrocytes (which can occur in response to Ca2+ elevation7) is also needed for synaptic LTP.

The authors addressed the second question by introducing tetanus neurotoxin into the astrocyte to specifically inhibit vesicular exocytosis. This treatment also blocks LTP selectively around the domain of the manipulated astrocyte, suggesting that vesicular exocytosis is the mechanism underlying release of D-serine4,8 and/or of another astrocytic transmitter.

Henneberger and colleagues' elegant single-cell manipulations1 enable the study of astrocytes as individual functional units. But the size of the domain that falls under the synaptic influence of a single astrocyte is not clear. Indeed, astrocytes are connected to each other by gap-junction channels, and Ca2+-buffering agents introduced into one astrocyte can pass through these intercellular connections and affect Ca2+ levels in neighbouring astrocytes. By contrast, the tetanus neurotoxin is confined to the astrocyte into which it is introduced, and yet it blocks LTP in a larger territory. This suggests either that D-serine diffuses beyond the domain of the releasing astrocyte, or that another transmitter released via Ca2+-dependent exocytosis amplifies the spatial influence of a single astrocyte.

Henneberger and colleagues' work raises many new questions about the role of astrocytes in synaptic transmission. Apart from D-serine, these cells release glutamate and other transmitters9. One may wonder if these different molecules are co-released to act together, or are secreted discretely in response to different stimuli and at different cellular locations, each with distinct functional roles. Astrocyte-released glutamate has been shown to activate extrasynaptic NMDARs, in either dendrites or presynaptic terminals10. Does D-serine release participate in this extrasynaptic activation? Or is its action confined to synaptic NMDARs? Future definition of the plasma-membrane sites from which D-serine is released and at which it is taken up, compared with those of glutamate, will help to clarify these issues. Ultimately though, Henneberger and colleagues' work1 conveys an important message — the contribution of glial cells to synaptic functions cannot be overlooked, and any study of synaptic physiology will need to consider glial biology if scientists hope to achieve a comprehensive understanding of brain function.


  1. 1

    Henneberger, C., Papouin, T., Oliet, S. H. R. & Rusakov, D. A. Nature 463, 232–236 (2010).

  2. 2

    Bushong, E. A., Martone, M. E., Jones, Y. Z. & Ellisman, M. H. J. Neurosci. 22, 183–192 (2002).

  3. 3

    Halassa, M. M. et al. J. Neurosci. 27, 6473–6477 (2007).

  4. 4

    Mothet, J. P. et al. Proc. Natl Acad. Sci. USA 102, 5606–5611 (2005).

  5. 5

    Panatier, A. et al. Cell 125, 775–784 (2006).

  6. 6

    Kartvelishvily, E., Shleper, M., Balan, L., Dumin, E. & Wolosker, H. J. Biol. Chem. 281, 14151–14162 (2006).

  7. 7

    Bezzi, P. et al. Nature 391, 281–285 (1998).

  8. 8

    Fellin, T. et al. Proc. Natl Acad. Sci. USA 106, 15037–15042 (2009).

  9. 9

    Volterra, A. & Meldolesi, J. Nature Rev. Neurosci. 6, 626–640 (2005).

  10. 10

    Perea, G., Navarrete, M. & Araque, A. Trends Neurosci. 32, 421–431 (2009).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Santello, M., Volterra, A. Astrocytes as aide-mémoires. Nature 463, 169–170 (2010).

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.