Imaging input and output dynamics of neocortical ... - Semantic Scholar
Amiram Grinvald* Department of Neurobiology, The Weizmann Institute of Science, 76100 Rehovot, Israel
M
any remarkable properties of neocortical networks are not detectable in the activities of single neurons. How the properties of single neurons and their synaptic connections combine to form networks that are capable of striking sensory processing and higher brain functions remains a major enigma. To understand how neuronal networks function, the activities of numerous single neurons must be studied simultaneously. The last few decades have seen the emergence of powerful functional imaging techniques that cover broad spatial and temporal scales (Fig. 1A), from single molecules to the intact human brain, which can be imaged noninvasively by techniques such as EEG, magnetoencephalography, and functional MRI. However, comprehensive understanding of neuronal computations requires spatial resolution at the level of single cells, and the speed of communication in these networks demands temporal resolution within milliseconds. As an additional imperative for understanding neuronal information processing, the inputs (synaptic potentials) must be distinguished from the outputs [action potentials (APs)]. Without such dissection, we cannot fully understand perception, higher brain functions, and behavior. Up to now, a comprehensive description of network input and output activity at the level of a single cortical neurons has not been possible. However, in a recent issue of PNAS, Kerr et al. (1) demonstrated that, by employing two-photon calcium imaging of bulk-labeled tissue, local input and output activities in the neocortex can be dissected in vivo. This approach should facilitate the exploration of basic mechanisms underlying neocortical development, function, and plasticity. Several types of techniques have been used to study cortical networks function. Modern extracellular recording methods (2) enable us to obtain simultaneous measurements from multiple cells, but they suffer from poorly defined cell identities and a severe sampling problem and are incapable of resolving nonactive neurons. Optical imaging of voltage-sensitive dye (VSD) signals (3) has revealed spatiotemporal dynamics on the scale of cortical columns. However, it still lacks single-cell resolution, although in vitro imaging of www.pnas.org!cgi!doi!10.1073!pnas.0506755102
individual processes within single cells has been demonstrated (4). Calcium imaging of electrical activity can be used to monitor activity in neuronal networks because APs and synaptic potentials promote calcium influx through voltage-dependent calcium channels, creating changes in intracellular calcium concentration in most neurons. Several research groups have made important contributions to the development of calcium imaging. First was the use of a naturally occurring substance, Acquarin, for optical imaging of calcium. Next, organic calcium indicators like Arsenazo (5) were injected into each neuron. A major breakthrough was providing an easy method for trapping probes in a large population of intact cells (e.g., ref. 6), which made it unnecessary to inject dye into each of the numerous single cells to be monitored. Instead, an ester derivative of the probe permeates the neuronal membrane, and the ester bond is cleaved by an intracellular enzyme; as a result, the charged probe is trapped and becomes an effective indicator. Next the heroic and successful efforts of Tsien (7) provided scientists with a large variety of such AM calcium probes. Another fundamental contribution was related to the resolution accomplished by optical imaging depending on the optics used. The laser scanning methodology coupled to two-photon imaging, developed by Denk et al. (8), provides superb spatial and temporal resolution in three-dimensional brain tissue. Yuste and Katz (9), using AM probes in young brain slices, pioneered imaging activities of many single cells. However, loading in adult slice and in vivo failed (10). With the aim of using AM probes also in adult neocortex in vivo, Stosiek et al. (11) recently developed the multicell bolus loading technique for calciumindicator loading of large cell populations in vivo. Ohki et al. (12) used this technique to provide a stunning view for the high precision of cortical maps for direction selectivity in cat visual cortex. Kerr et al. (1) add a significant building block to this line of work by focusing on the temporal domain and the dissection of input and output activities, in vivo. The membrane-permeable calcium indicator was pressure-ejected through a glass pipette into layers 2 and 3 of the neocortex in anesthetized rats. Approximately 1 h later, all cells within a radius of several hundred microns were labeled (Fig. 1 B
and C). Notably, staining was observed not only in cell bodies but also in the neuropil. Because the loading technique was nonspecific, dissection of the various calcium-signal components was necessary. Using a red fluorescent dye (13) that identifies astrocytes in vivo to counterstain the astrocytic network (yellow cells in Fig. 1 B and C), they observed slow oscillations on a time scale of minutes in identified astrocytes. Neurons, in contrast, displayed spontaneous but infrequent calcium transients of short (!1 s) duration with rapid onset and exponential decay (Fig. 1F Right Inset) resembling APevoked calcium transients observed in vitro and in vivo (9, 10, 14). To determine the reliability of spike detection, Kerr et al. (1) carried out cell-attached recordings. Spontaneous APs were then recorded extracellularly, while somatic calcium transients in the same cell were simultaneously measured optically. Strikingly, 97% of single APs and 100% of bursts were detected. The authors concluded that AP activity is reliably resolved with both single-cell and single-AP resolution. Thus, it was possible to optically detect the spike trains, representing output activity, in local neuronal circuits, providing an optical analog of multiple single-unit recordings but with the added advantage that AP activity can be assigned to all identified neurons within a cortical volume. For neuroscientists, this is a dream come true. A significant contribution of Kerr et al. (1) was their exploration of the origin of the large fluorescence changes observed in the neuropil surrounding cell bodies. The time course of fluorescence intensity in regions not containing cell somata revealed large fluctuations, representing a bulk measurement of calcium signals in neuropil structures. To determine the origin of the neuropil signal, they combined in vivo two-photon calcium imaging with electrocorticograms (ECoG) and intracellular whole-cell recordings. In whole-cell recordings from individual cells, the membrane potential fluctuated between up and down states, as typically observed in neocortical neurons (15–17). This ongoing spontaneous activity was also apparent in See companion article on page 14063 in issue 39 of volume 102. *E-mail: [email protected]. © 2005 by The National Academy of Sciences of the USA
PNAS " October 4, 2005 " vol. 102 " no. 40 " 14125–14126
COMMENTARY
Imaging input and output dynamics of neocortical networks in vivo: Exciting times ahead
Fig. 1. The spatiotemporal capabilities of available tools for studying neocortical structure and function. (A) These are depicted by the colored rectangles. Optical imaging (red) covers almost the entire area, and Kerr et al. (1) established a corner stone here. Multiphoton calcium imaging currently occupies the lower two-thirds of the optical imaging ‘‘territory.’’(B) Side projection of a two-photon image of OGB-1 loaded cells in the neocortex showing neurons (green) and astrocytes (yellow). (C) Higher magnification area of stained cell. (D) Pseudocolored representation of single cells shown in B depicting the fraction of up-states in which single neurons were active during a 90-s period (color scale). (E) Ongoing calcium transient from the neuropil; the OEG fluctuations (red) correlate with electrical ECoG signals (black). (F) Simultaneous calcium transients from the neuropil; the input (red) and the output from an identified neuron (green) recorded over several minutes. (Right Inset) Ongoing neuropil input signal and an output on expanded time scale.
the ECoG, which correlated with the intracellular recording. Finally, the neuropil fluorescence signals were well synchronized with this electrical activity, indicating that they reflect ongoing up and down states (Fig. 1E). In view of the strong correlation with the ECoG, the authors termed the neuropil signal an ‘‘optical encephalogram’’ (OEG). Several neuropil structures could contribute to the OEG axons and presynaptic boutons, dendrites and dendritic spines, and glial processes. A contribution by astrocytes was ruled out because the calcium oscillations displayed by these cells operate on a much slower time scale. The authors then addressed the important question of whether the OEG represents mainly axonal (presynaptic) or mainly dendritic (postsynaptic) activity or both. By loading the calcium probe in different restricted depths and employing pharmacological manipulations, they concluded that OEG fluctuations originate predominantly from axonal structures, presumably reflecting the bulk average of AP-evoked calcium transients in presynaptic boutons and axons, rather than dendrites and spines. Thus, the OEG can be viewed as a measure for volume-averaged input activity to the particular local region. No other available technique offers imaging of both input and output (Fig. 1F) at this spectacular spatial and temporal resolution. Using these OEG and spike activity, Kerr et al. (1) characterized spontaneous activity during cortical up states (Fig. 1D) and found that the number of spiking neurons was directly dependent
on the OEG amplitude, revealing an input–output transfer function for the local network characterized by a threshold below which no APs were generated and above which the relationship was approximately linear. The stochastic behavior reported by Kerr et al. suggests that spikes do not explicitly depend on previous spiking times. This finding is in contrast to those of other studies, in which both spatial and temporal spiking structures in acute cortical slices were reported (18, 19). The in vitro preparation used in the previous studies might have contributed to the discrepancy, underscoring the importance of in vivo experiments for exploration of processing in neuronal circuits. The approach developed by Kerr et al. (1) is likely to be adopted in the future by many investigators who wish to understand fundamental principles of cortical processing at the single-cell level, taking into account both the local input and entire output of the network. However, some additional improvements are necessary. Because of the frame acquisition rate and the relatively slow (1 s) decay of the calcium signal, the temporal resolution of this approach is limited ("20 Hz), causing calcium transients to merge as the APs occur in rapid succession (9, 10). Considering the amplitude of AP-evoked calcium transients might improve the temporal accuracy of the spike pattern extraction, because the amplitude depends on the number of APs (10, 19) and their relative timing, as shown here. However, in the present study of spontaneous activity,
only a small fraction of events consisted of more than one AP. In future studies of evoked activity, the spike rate will be much higher. Thus, the design and synthesis of better organic probes offering faster rise and decay times is important, as is the engineering of probes selective for specific cell types, based on modern molecular biology approaches (20). Furthermore, there is often a need to explore larger areas so that several cortical columns can be analyzed simultaneously; this could be accomplished by multisite injections, novel loading approaches, or novel calcium probes. Because the voltage dictates the behavior of the neurons, an important caveat regarding calcium imaging is its inability to report voltage changes directly. A possible solution is to combine the experiment with VSD imaging. An additional nontrivial undertaking is the separation of the input regions from the output in large three-dimensional space, and the image processing tools that should be advanced. Finally, the same technology should be adopted to study higher brain functions in an animal model that is awake and behaving rather than anesthetized. This should be feasible, as demonstrated by using VSD imaging in behaving monkeys. Regardless of future improvement, it is already clear that many intriguing questions that could not previously be addressed can now be pursued, and new conceptual frameworks can be anticipated. There are exciting times ahead in the exploration of cortical networks and their remarkable functions.
1. Kerr, J. N. D., Greenberg, D. & Helmchen, F. (2005) Proc. Natl. Acad. Sci. USA 102, 14063–14068. 2. Buzsaki, G. (2004) Nat. Neurosci. 7, 446–451. 3. Grinvald, A. & Hildesheim, R. (2004) Nat. Rev. Neurosci. 5, 874–885. 4. Djurisic, M., Antic, S., Chen, W. R. & Zecevic, D. J. (2004) Neuroscience 24, 6703–6714. 5. Brown, J. E., Cohen, L. B., De Weer, P., Pinto, L. H., Ross, W. N. & Salzberg, B. M. (1975) Biophys. J. 15, 1155–1160. 6. Rotman, B. & Papermaster, B. W. (1966) Proc. Natl. Acad. Sci. USA 55, 134–141. 7. Tsien, R. Y. (1983) Annu. Rev. Biophys. Bioeng. 12, 91–116.
8. Denk, W., Strickler, J. H. & Webb, W. W. (1990) Science 248, 73–76. 9. Yuste, R. & Katz, L. C. (1991) Neuron 6, 333–344. 10. Mao, B. Q., Hamzei-Sichani, F., Aronov, D., Froemke, R. C. & Yuste, R. (2001) Neuron. 32, 883–898. 11. Stosiek, C., Garaschuk, O., Holthoff, K. & Konnerth, A. (2003) Proc. Natl. Acad. Sci. USA 100, 7319–7324. 12. Ohki, K., Chung, S., Ch’ng, Y. H., Kara, P. & Reid, R. C. (2005) Nature 433, 597–603. 13. Nimmerjahn, A., Kirchhoff, F., Kerr, J. N. & Helmchen, F. (2004) Nat. Methods 1, 31–37. 14. Helmchen, F., Imoto, K. & Sakmann, B. (1996) Biophys. J. 70, 1069–1081.
15. Cowan, R. L. & Wilson, C. J. (1994) J. Neurophysiol. 71, 17–32. 16. Lampl, I., Reichova, I. & Ferster, D. (1999) Neuron 22, 361–374. 17. Petersen, C. C., Hahn, T. T., Mehta, M., Grinvald, A. & Sakmann, B. (2003) Proc. Natl. Acad. Sci. USA 100, 13638–13643. 18. Cossart, R., Aronov, D. & Yuste, R. (2003) Nature 423, 283–288. 19. Ikegaya, Y., Aaron, G., Cossart, R., Aronov, D., Lampl, I., Ferster, D. & Yuste, R. (2004) Science 304, 559–564. 20. Miyawaki, A. Llopis, J., Heim, R., McCaffery, J. M., Adams, J. A., Ikura, M. & Tsien, R. Y. (1997) Nature 388, 882–887.
14126 " www.pnas.org!cgi!doi!10.1073!pnas.0506755102
Grinvald
M
any remarkable properties of neocortical networks are not detectable in the activities of single neurons. How the properties of single neurons and their synaptic connections combine to form networks that are capable of striking sensory processing and higher brain functions remains a major enigma. To understand how neuronal networks function, the activities of numerous single neurons must be studied simultaneously. The last few decades have seen the emergence of powerful functional imaging techniques that cover broad spatial and temporal scales (Fig. 1A), from single molecules to the intact human brain, which can be imaged noninvasively by techniques such as EEG, magnetoencephalography, and functional MRI. However, comprehensive understanding of neuronal computations requires spatial resolution at the level of single cells, and the speed of communication in these networks demands temporal resolution within milliseconds. As an additional imperative for understanding neuronal information processing, the inputs (synaptic potentials) must be distinguished from the outputs [action potentials (APs)]. Without such dissection, we cannot fully understand perception, higher brain functions, and behavior. Up to now, a comprehensive description of network input and output activity at the level of a single cortical neurons has not been possible. However, in a recent issue of PNAS, Kerr et al. (1) demonstrated that, by employing two-photon calcium imaging of bulk-labeled tissue, local input and output activities in the neocortex can be dissected in vivo. This approach should facilitate the exploration of basic mechanisms underlying neocortical development, function, and plasticity. Several types of techniques have been used to study cortical networks function. Modern extracellular recording methods (2) enable us to obtain simultaneous measurements from multiple cells, but they suffer from poorly defined cell identities and a severe sampling problem and are incapable of resolving nonactive neurons. Optical imaging of voltage-sensitive dye (VSD) signals (3) has revealed spatiotemporal dynamics on the scale of cortical columns. However, it still lacks single-cell resolution, although in vitro imaging of www.pnas.org!cgi!doi!10.1073!pnas.0506755102
individual processes within single cells has been demonstrated (4). Calcium imaging of electrical activity can be used to monitor activity in neuronal networks because APs and synaptic potentials promote calcium influx through voltage-dependent calcium channels, creating changes in intracellular calcium concentration in most neurons. Several research groups have made important contributions to the development of calcium imaging. First was the use of a naturally occurring substance, Acquarin, for optical imaging of calcium. Next, organic calcium indicators like Arsenazo (5) were injected into each neuron. A major breakthrough was providing an easy method for trapping probes in a large population of intact cells (e.g., ref. 6), which made it unnecessary to inject dye into each of the numerous single cells to be monitored. Instead, an ester derivative of the probe permeates the neuronal membrane, and the ester bond is cleaved by an intracellular enzyme; as a result, the charged probe is trapped and becomes an effective indicator. Next the heroic and successful efforts of Tsien (7) provided scientists with a large variety of such AM calcium probes. Another fundamental contribution was related to the resolution accomplished by optical imaging depending on the optics used. The laser scanning methodology coupled to two-photon imaging, developed by Denk et al. (8), provides superb spatial and temporal resolution in three-dimensional brain tissue. Yuste and Katz (9), using AM probes in young brain slices, pioneered imaging activities of many single cells. However, loading in adult slice and in vivo failed (10). With the aim of using AM probes also in adult neocortex in vivo, Stosiek et al. (11) recently developed the multicell bolus loading technique for calciumindicator loading of large cell populations in vivo. Ohki et al. (12) used this technique to provide a stunning view for the high precision of cortical maps for direction selectivity in cat visual cortex. Kerr et al. (1) add a significant building block to this line of work by focusing on the temporal domain and the dissection of input and output activities, in vivo. The membrane-permeable calcium indicator was pressure-ejected through a glass pipette into layers 2 and 3 of the neocortex in anesthetized rats. Approximately 1 h later, all cells within a radius of several hundred microns were labeled (Fig. 1 B
and C). Notably, staining was observed not only in cell bodies but also in the neuropil. Because the loading technique was nonspecific, dissection of the various calcium-signal components was necessary. Using a red fluorescent dye (13) that identifies astrocytes in vivo to counterstain the astrocytic network (yellow cells in Fig. 1 B and C), they observed slow oscillations on a time scale of minutes in identified astrocytes. Neurons, in contrast, displayed spontaneous but infrequent calcium transients of short (!1 s) duration with rapid onset and exponential decay (Fig. 1F Right Inset) resembling APevoked calcium transients observed in vitro and in vivo (9, 10, 14). To determine the reliability of spike detection, Kerr et al. (1) carried out cell-attached recordings. Spontaneous APs were then recorded extracellularly, while somatic calcium transients in the same cell were simultaneously measured optically. Strikingly, 97% of single APs and 100% of bursts were detected. The authors concluded that AP activity is reliably resolved with both single-cell and single-AP resolution. Thus, it was possible to optically detect the spike trains, representing output activity, in local neuronal circuits, providing an optical analog of multiple single-unit recordings but with the added advantage that AP activity can be assigned to all identified neurons within a cortical volume. For neuroscientists, this is a dream come true. A significant contribution of Kerr et al. (1) was their exploration of the origin of the large fluorescence changes observed in the neuropil surrounding cell bodies. The time course of fluorescence intensity in regions not containing cell somata revealed large fluctuations, representing a bulk measurement of calcium signals in neuropil structures. To determine the origin of the neuropil signal, they combined in vivo two-photon calcium imaging with electrocorticograms (ECoG) and intracellular whole-cell recordings. In whole-cell recordings from individual cells, the membrane potential fluctuated between up and down states, as typically observed in neocortical neurons (15–17). This ongoing spontaneous activity was also apparent in See companion article on page 14063 in issue 39 of volume 102. *E-mail: [email protected]. © 2005 by The National Academy of Sciences of the USA
PNAS " October 4, 2005 " vol. 102 " no. 40 " 14125–14126
COMMENTARY
Imaging input and output dynamics of neocortical networks in vivo: Exciting times ahead
Fig. 1. The spatiotemporal capabilities of available tools for studying neocortical structure and function. (A) These are depicted by the colored rectangles. Optical imaging (red) covers almost the entire area, and Kerr et al. (1) established a corner stone here. Multiphoton calcium imaging currently occupies the lower two-thirds of the optical imaging ‘‘territory.’’(B) Side projection of a two-photon image of OGB-1 loaded cells in the neocortex showing neurons (green) and astrocytes (yellow). (C) Higher magnification area of stained cell. (D) Pseudocolored representation of single cells shown in B depicting the fraction of up-states in which single neurons were active during a 90-s period (color scale). (E) Ongoing calcium transient from the neuropil; the OEG fluctuations (red) correlate with electrical ECoG signals (black). (F) Simultaneous calcium transients from the neuropil; the input (red) and the output from an identified neuron (green) recorded over several minutes. (Right Inset) Ongoing neuropil input signal and an output on expanded time scale.
the ECoG, which correlated with the intracellular recording. Finally, the neuropil fluorescence signals were well synchronized with this electrical activity, indicating that they reflect ongoing up and down states (Fig. 1E). In view of the strong correlation with the ECoG, the authors termed the neuropil signal an ‘‘optical encephalogram’’ (OEG). Several neuropil structures could contribute to the OEG axons and presynaptic boutons, dendrites and dendritic spines, and glial processes. A contribution by astrocytes was ruled out because the calcium oscillations displayed by these cells operate on a much slower time scale. The authors then addressed the important question of whether the OEG represents mainly axonal (presynaptic) or mainly dendritic (postsynaptic) activity or both. By loading the calcium probe in different restricted depths and employing pharmacological manipulations, they concluded that OEG fluctuations originate predominantly from axonal structures, presumably reflecting the bulk average of AP-evoked calcium transients in presynaptic boutons and axons, rather than dendrites and spines. Thus, the OEG can be viewed as a measure for volume-averaged input activity to the particular local region. No other available technique offers imaging of both input and output (Fig. 1F) at this spectacular spatial and temporal resolution. Using these OEG and spike activity, Kerr et al. (1) characterized spontaneous activity during cortical up states (Fig. 1D) and found that the number of spiking neurons was directly dependent
on the OEG amplitude, revealing an input–output transfer function for the local network characterized by a threshold below which no APs were generated and above which the relationship was approximately linear. The stochastic behavior reported by Kerr et al. suggests that spikes do not explicitly depend on previous spiking times. This finding is in contrast to those of other studies, in which both spatial and temporal spiking structures in acute cortical slices were reported (18, 19). The in vitro preparation used in the previous studies might have contributed to the discrepancy, underscoring the importance of in vivo experiments for exploration of processing in neuronal circuits. The approach developed by Kerr et al. (1) is likely to be adopted in the future by many investigators who wish to understand fundamental principles of cortical processing at the single-cell level, taking into account both the local input and entire output of the network. However, some additional improvements are necessary. Because of the frame acquisition rate and the relatively slow (1 s) decay of the calcium signal, the temporal resolution of this approach is limited ("20 Hz), causing calcium transients to merge as the APs occur in rapid succession (9, 10). Considering the amplitude of AP-evoked calcium transients might improve the temporal accuracy of the spike pattern extraction, because the amplitude depends on the number of APs (10, 19) and their relative timing, as shown here. However, in the present study of spontaneous activity,
only a small fraction of events consisted of more than one AP. In future studies of evoked activity, the spike rate will be much higher. Thus, the design and synthesis of better organic probes offering faster rise and decay times is important, as is the engineering of probes selective for specific cell types, based on modern molecular biology approaches (20). Furthermore, there is often a need to explore larger areas so that several cortical columns can be analyzed simultaneously; this could be accomplished by multisite injections, novel loading approaches, or novel calcium probes. Because the voltage dictates the behavior of the neurons, an important caveat regarding calcium imaging is its inability to report voltage changes directly. A possible solution is to combine the experiment with VSD imaging. An additional nontrivial undertaking is the separation of the input regions from the output in large three-dimensional space, and the image processing tools that should be advanced. Finally, the same technology should be adopted to study higher brain functions in an animal model that is awake and behaving rather than anesthetized. This should be feasible, as demonstrated by using VSD imaging in behaving monkeys. Regardless of future improvement, it is already clear that many intriguing questions that could not previously be addressed can now be pursued, and new conceptual frameworks can be anticipated. There are exciting times ahead in the exploration of cortical networks and their remarkable functions.
1. Kerr, J. N. D., Greenberg, D. & Helmchen, F. (2005) Proc. Natl. Acad. Sci. USA 102, 14063–14068. 2. Buzsaki, G. (2004) Nat. Neurosci. 7, 446–451. 3. Grinvald, A. & Hildesheim, R. (2004) Nat. Rev. Neurosci. 5, 874–885. 4. Djurisic, M., Antic, S., Chen, W. R. & Zecevic, D. J. (2004) Neuroscience 24, 6703–6714. 5. Brown, J. E., Cohen, L. B., De Weer, P., Pinto, L. H., Ross, W. N. & Salzberg, B. M. (1975) Biophys. J. 15, 1155–1160. 6. Rotman, B. & Papermaster, B. W. (1966) Proc. Natl. Acad. Sci. USA 55, 134–141. 7. Tsien, R. Y. (1983) Annu. Rev. Biophys. Bioeng. 12, 91–116.
8. Denk, W., Strickler, J. H. & Webb, W. W. (1990) Science 248, 73–76. 9. Yuste, R. & Katz, L. C. (1991) Neuron 6, 333–344. 10. Mao, B. Q., Hamzei-Sichani, F., Aronov, D., Froemke, R. C. & Yuste, R. (2001) Neuron. 32, 883–898. 11. Stosiek, C., Garaschuk, O., Holthoff, K. & Konnerth, A. (2003) Proc. Natl. Acad. Sci. USA 100, 7319–7324. 12. Ohki, K., Chung, S., Ch’ng, Y. H., Kara, P. & Reid, R. C. (2005) Nature 433, 597–603. 13. Nimmerjahn, A., Kirchhoff, F., Kerr, J. N. & Helmchen, F. (2004) Nat. Methods 1, 31–37. 14. Helmchen, F., Imoto, K. & Sakmann, B. (1996) Biophys. J. 70, 1069–1081.
15. Cowan, R. L. & Wilson, C. J. (1994) J. Neurophysiol. 71, 17–32. 16. Lampl, I., Reichova, I. & Ferster, D. (1999) Neuron 22, 361–374. 17. Petersen, C. C., Hahn, T. T., Mehta, M., Grinvald, A. & Sakmann, B. (2003) Proc. Natl. Acad. Sci. USA 100, 13638–13643. 18. Cossart, R., Aronov, D. & Yuste, R. (2003) Nature 423, 283–288. 19. Ikegaya, Y., Aaron, G., Cossart, R., Aronov, D., Lampl, I., Ferster, D. & Yuste, R. (2004) Science 304, 559–564. 20. Miyawaki, A. Llopis, J., Heim, R., McCaffery, J. M., Adams, J. A., Ikura, M. & Tsien, R. Y. (1997) Nature 388, 882–887.
14126 " www.pnas.org!cgi!doi!10.1073!pnas.0506755102
Grinvald
