Chasing the Cortical Assembly

Damian J. Wallace, PhD, and Jason N. D. Kerr, PhD

Network Imaging Group Max Planck Institute for Biological Cybernetics Tübingen, Germany

© 2012 Kerr

Chasing the Cortical Assembly 43

Introduction NOTES Why is the cortex so difficult to understand? Although involves populations of that are thought to we know enormous amounts of detailed information form a percept of a stimulus. about the neurons that make up the cortex, placing this information back into context of the behaving The most influential theory regarding how activity animal is a serious challenge. In this chapter, we aim in individual neurons may translate into percept to outline some recent technical advances that may formation, the cell assembly hypothesis, was light the way toward the chase for the functional originally conceived by D. O. Hebb in 1949 (Hebb, ensemble. We summarize the progress that has been 1949). Hebb’s functional cell assembly hypothesis made using optical recording approaches with a view aimed to provide a mechanistic and anatomically to what can be expected in the near future, given relevant explanation of how groups of neurons, the recent technological advances. The modeling acting together, may form a percept. Through their and theoretical arguments surrounding neuronal multiple connections, Hebb proposed, neurons ensembles have been described in great detail form cell assemblies that are collectively activated previously (Palm, 1982; Braitenberg, 1978; Gerstein by sensory input and form a brief closed system et al., 1989; Harris, 2005; Mountcastle, 1997, 2003; after stimulation has ceased. Activity from each Wickens and Miller, 1997), so we will not review cell assembly can propagate and activate additional them here. connected cell assemblies in sequence, which he termed a “phase sequence,” and this was proposed to Testing Cortical Hypotheses be the core of neuronal representation of a stimulus- Both anatomically (Douglas and Martin, 2004, 2007) based percept (Hebb, 1949). As individual neurons and electrophysiologically (Spruston, 2008), the can leave and join cell assemblies using activity- properties of individual cells that make up the cortex based synaptic plasticity rules, and therefore can be are very well described, albeit mainly from single-cell members of multiple assemblies, the cell assembly recordings or recordings made from cells in isolation. can be dispersed throughout a cortical population Numerous theories about cortical function date back but linked through potentiated synaptic connections to the early part of the 20th century (von Economo (Gerstein et al., 1989; Gerstner et al., 1993). This and Koskinas, 1925; Hebb, 1949; Lorente de No, is where the challenge of testing the cell-assembly 1949; Mountcastle, 1978). Of these theories, very hypothesis lies: locating the neurons involved in few, if any, have been experimentally tested. Why is forming a cell assembly. Testing this hypothesis this? It is at least in part due to the vast scale of the has been exceedingly difficult owing to the vast problem. More importantly, it is not clear whether numbers of neurons potentially involved, as neither the proposed theories are able to generate hypotheses the number nor locations of ensemble members are that are testable using the available methods, or known. alternatively whether these theories are too general to generate testable hypotheses. Members of functional neuronal ensembles, in any of the proposed theories of cortical function, The basic anatomical pathways and connectivity are dispersed within neuronal populations, and no between cortical areas and cortical layers are systematic anatomical organization within these reasonably well characterized (Braitenberg and ensembles has yet been described. It is thus generally Schüz, 1991; Braitenberg et al., 1998). However, it thought that increasing the numbers of neurons is the firing of action potentials (APs) that defines from which simultaneous recordings are made will the functional cortical characteristics, moment by increase the chances of capturing many members of moment, through these pathways. APs propagate an ensemble (Grewe and Helmchen, 2009). Given throughout an individual ’s entire axonal that it is not clear how many members make up a arbor (Cox et al., 2000; Koester and Sakmann, 2000), neuronal ensemble or whether the same neurons are and probably influence all postsynaptic partners. involved from one trial to another, we suggest that Individual postsynaptic neurons receive and integrate this is only part of the picture. Making measurements a vast number of inputs from presynaptic neurons at from functional neuronal ensembles during cortical any moment (Hasenstaub et al., 2005; Waters and computation is likely to require multiple techniques Helmchen, 2006). Although individual neurons capable of recording from, locating, and potentially have considerable computational capacity (Larkum manipulating the activity of individual neurons and Nevian, 2008; Losonczy et al., 2008; Jia et al., embedded within large populations. Although a 2010), neuronal processing of sensory information multitude of new technical advances can be applied

© 2012 Kerr 44

NOTES to locating the neuronal assembly (improved • The ability to record from the same neurons over transynaptic tracing being an example), the question many days (Mank et al., 2008; Tian et al., 2009; arises: Are we any closer to recording from, or Andermann et al., 2010). understanding, the Hebbian cell assembly and its role in sensory coding? More recently, several groups have extended multiphoton population imaging to the study Lighting the Cortical Ensemble of several animal models: the awake head-fixed Multiphoton imaging (Greenberg et al., 2008), head-fixed and behaving (Andermann et al., 2010; Komiyama et al., 2010), One of the biggest recent advances in the ability to head-fixed but mobile (Dombeck et al., 2007, 2009), record activity simultaneously from many neurons and freely moving animal (Sawinski et al., 2009). If in vivo (Stosiek et al., 2003), with single-cell and detection of AP firing is central to detecting neurons single-AP accuracy (Kerr et al., 2005), has come from involved in a cell assembly, in which individual multiphoton imaging (Denk et al., 1990; Svoboda members could change with each trial, then imaging et al., 1997; Kerr and Denk, 2008). Its three main must be able to accurately resolve activity on a trial- advantages are as follows: by-trial basis to enable capture of activity related to neuronal assembly. • The ability to infer electrical activity from all neurons within a local area on a trial-by-trial basis (Kerr et al., 2007; Sato et al., 2007; Rothschild et Inferring action potential al., 2010); All activity-based population recordings have used either bolus loading of synthetic fluorescent Ca2+- • Known spatial location of all the recorded neurons indicator dyes (Stosiek et al., 2003) (Fig. 1a–b) or 2+ (Ohki et al., 2005; Mrsic-Flogel et al., 2007); infection of cells with genetically encoded Ca indicators (Hasan et al., 2004; Mank et al., 2008; • The capacity to record activity from neurons that Wallace et al., 2008; Tian et al,. 2009). These fire at low rates (Kerr et al., 2005; Greenberg et al., indicators typically label populations of hundreds of 2008); and neurons in areas covering ~500 × 500 × 500 μm (Kerr and Denk, 2008). Although the indicators report

Figure 1. (see opposite page) Imaging neuronal activity. a, Two-photon image of a population of cortical cells labeled with the fluorescent Ca2+ indicator Oregon green BAPTA-1. Astrocytes are counterstained with Sulforhodamine 101 (yellow/red), while neurons appear green. b, Ca2+ transients simultaneously recorded from a population of 13 neurons in vivo. c, Simultaneous Ca2+ imaging and cell-attached electrophysiological recording in vivo showing Ca2+ transients associated with single APs and doublets. Simultaneous electrophysiological recording is essential to accurately calibrate algorithms designed to convert the Ca2+ traces observed in in vivo recordings from populations of neurons into accurate AP raster plots. The simultaneously recorded Ca2+ trace and extracellular electrophysiology are shown to the right, with the model output from the spike-detection algo- rithm (described in Greenberg et al., 2008) corresponding to the Ca2+ trace (green). d–e, Infrared images and Ca2+ transients recorded in neurons in the forelimb representation of the primary in a head-fixed mouse on a spherical treadmill. d, Infrared video images showing forepaw movements during typical grooming and running behavior. e, Baseline subtracted Ca2+ traces (black) with significant transients detected using the detection method employed (orange). In Dombeck et al. (2009), two independent analysis methods were used to provide compelling evidence for functional clustering of neurons preferentially active during running or grooming behaviors. f–h, High-speed in vivo two-photon imaging of neuronal activity using an AOD scanning system. f, Overview image of a field of labeled neurons, highlighting a group of 7 cells on an irregular scan path from which data were collected. g, Ca2+-imaging traces from the correspondingly numbered neurons in panel f. h, Graph showing the relationship between cellwise sampling rate and number of cells scanned when the number of acquisition points in each cell is varied from 3 to 9. Using this AOD scanning approach, ~100 cells can be scanned in vivo with a cellwise acquisition rate of ~100 Hz. i–l, data acquired using a miniaturized, head-mounted two-photon microscope. i, Image of a population of layer II/III neurons in the primary . The labeling of neurons and astrocytes in this image is the same as that shown in panel a. In this study, the animal was allowed to move freely around a raised C-shaped track around which 3 different visual stimuli were arranged. The layout of the track and stimuli is shown in j. Ca2+ traces simultaneously recorded in 3 neurons and the associated raster plots derived from these data are shown in l and k, respectively. Colored blocks in the background of the raster demarcate times during which the animal’s center of gaze was within one of the stimulus monitors. Color coding cor- responds to the visual-stimulus monitor outline colors shown in panel j. The colors of the individual raster ticks correspond to transients that the spike-detection algorithm used in this study and allocated 1 (black), 2 (red), or 3 (green) APs. The neuron from which the trace labeled as “i” in k and l was robustly activated while the animal’s gaze was on the stimulus marked as 3 in j. a J. Kerr, unpublished observations. b–c, Greenberg et al., 2008, their Fig. 1, adapted with permission. d–e, Dombeck et al., 2009, their Fig. 1, adapted with permission. f–h Grewe et al., 2010, their Fig. 2, adapted with permission. i–l, Sawinski et al.,

2009, their Fig. 3E and 4B,D, adapted with permission. Scale bars: a, 50 μm; b, 5 s and 40% ∆F/F0; c, 20% ∆F/F0 and 5 s; e, 30% ∆F/F0 and 15 s; f, 20 μm, 10% ∆F/F0, and 5 s; i, 20 μm; l, 30% ∆F/F0 and 10 s. © 2012 Kerr Chasing the Cortical Assembly 45

NOTES

© 2012 Kerr 46

NOTES AP-evoked changes in Ca2+, several research groups methods as those used with data from conventional have been able to accurately infer single spikes from galvanic scanning (Fig. 1f–h) but were able to resolve this signal (Kerr et al., 2005, 2007; Sato et al., 2007; transient peaks evoked from individual spikes firing Tian et al., 2009; Rothschild et al., 2010). They have at ~30 Hz. also been able to infer spike numbers during complex bursting activity (Greenberg et al., 2008) using Spatial resolution of neurons various spike-finding algorithm approaches (Fig. 1c). Several groups have been able to relate this activity back to the spatial location of the neurons. They 2+ Inferring APs from Ca transients with the aim have taken advantage of spatial resolution that of calculating firing rates or response rates for allows them to precisely locate recorded neurons individual neurons usually requires “ground truth” relative to both macrostructures (e.g., cortical to be simultaneously established using cell-attached layers and somatosensory columnar borders) and electrical recordings in order to establish both the microstructures (e.g., surrounding neurons). In these false-positive and false-negative rates of the search experiments, a strong relationship has emerged on algorithm (but see Dombeck et al., 2009 and a trial-by-trial basis between neuronal firing–based Vogelstein et al., 2009 for statistical approaches). correlations and distance, when studied on a fine For inferring spikes in multiple summated transients scale, during sensory stimulation (Kerr et al., 2007), (Greenberg et al., 2008; Dombeck et al., 2009), and in awake, head-fixed animals (Dombeck et which are generally found in awake animals al., 2009; Komiyama et al., 2010). A clear spatial (Fig. 1d–e), algorithms need to be sufficiently robust relationship has also emerged for average orientation to slow drifting baseline calcium levels and variable tuning preferences for cat visual cortex neurons, firing rates of the individual neurons (Greenberg though not in the rat (Ohki et al., 2005). et al., 2008). Accurate inference of APs from Ca2+ transients enables researchers to measure both Despite these promising applications, this approach spontaneous firing rates and stimulus response still has fundamental limitations. Limitations rates in neurons that have sparse activity or low include the small number of neurons that can be firing rates—something that is almost impossible simultaneously scanned (typically, 20–50 neurons for extracellular recording approaches using spike- using conventional galvanic raster-scanning and waveform separation (but see Girman et al., 1999 requiring single-AP-detection fidelity) and the and Ecker et al., 2010). imaging-depth constraints (Theer and Denk, 2006). Further, because the functional cell assembly most Although almost all in vivo two-photon imaging likely will involve all the cortical layers, several studies have been restricted to the cortical attempts have been made at imaging neuronal supragranular layers, where neuronal responses and populations in three dimensions. In vivo, this has firing rates are lower than for neurons in the deeper been achieved by moving the objective lens with granular layers (Girman et al., 1999; de Kock et al., a piezoelectric actuator while using conventional 2007; de Kock and Sakmann, 2009), a recent study galvanic scanners (Gobel et al., 2007). Although by Mittmann et al. (2011) has imaged activity from this approach allows researchers to scan many more layer Vb neurons labeled with GCaMP3 (Tian et neurons at a time, it is still limited by the duty- al., 2009). The question arises as to whether it will cycle speed and increases the number of potentially still be possible to accurately infer spiking activity sampled neurons only by a factor of ~10 (but see 2+ from Ca transients in neurons firing at higher rates. Reddy and Saggau, 2005 and Botcherby et al., 2008 Several lines of evidence show that this is possible. for promising approaches). All these improvements 2+ High firing rates have been inferred from Ca have their advantages and limitations (Grewe and transients post hoc using a deconvolution method Helmchen, 2009). Also, although current imaging in mitral cells at firing rates of ~100 Hz (Yaksi approaches are limited to relatively small populations and Friedrich, 2006). A promising approach that of neurons in the upper cortical layers, making has, until recently, been employed mainly in vitro recordings from substantially larger populations from involves using acoustic-optical devices (AODs) to most cortical layers will likely be achievable within improved temporal resolution and to optimize scan the next few years in the awake behaving rodent. paths (Iyer et al., 2006; Vucinic and Sejnowski, 2007; Duemani Reddy et al., 2008). A recent study (Grewe Ultimately, the probability of finding a functional et al., 2010) has successfully applied the AOD two- cell assembly will potentially increase. To achieve dimensional scanning method in vivo to detect single this goal, strategies will aim to reduce the number APs and AP bursts. The group used similar detection of neurons that have to be sampled from (e.g., to

© 2012 Kerr Chasing the Cortical Assembly 47

neurons known a priori to be synaptically coupled) that head-fixed mice perform the discrimination NOTES and narrow the time window in which the neurons task under the two-photon microscope (albeit are active (e.g., by carefully designing behavioral with slightly reduced behavioral performance) experiments). and will perform enough trials in a day to allow the construction of full psychometric functions. Recording Awake Behavior Using a GECI, the researchers were able to record Given the complexity of the associating spikes responses from the same neurons for periods of up with neuronal assemblies, recording in the primary to several months. While adaptation of animals sensory areas provides a possible timing link to the to head-fixation has enabled imaging of neuronal forming of a percept in the awake animal. Because ensemble activity using conventional two-photon during movement light penetration into tissue is microscopes, another approach (not without its less mechanically damaging than the placement own complications) is to miniaturize the microscope of physical electrodes (but see Fee and Leonardo, sufficiently to allow it to be carried by the animal 2001), imaging allows recording activity in the (Flusberg et al., 2008; Sawinski et al., 2009). Using 2+ awake animal where movements within the image a miniaturized microscope, functional Ca imaging can be successfully corrected offline, with very little from populations of layer II/III neurons in the visual data loss (Dombeck et al., 2007; Greenberg et al., cortex of freely moving rats was recently demonstrated 2008; Greenberg and Kerr, 2009). The development (Sawinski et al., 2009) (Fig. 1i–l). This study opens of genetically encoded calcium indicators (GECIs) another door to investigating neuronal ensemble is a further advance that has made imaging neurons activity in freely behaving animals. Image stability in trained animals a more realistic goal (Hasan et al., was suitably high to allow continuous recordings 2004; Mank et al., 2008; Wallace et al., 2008; Tian of neuronal activity for prolonged periods (hours) et al., 2009). GECIs now allow neuronal activity to while the animal moved around a raised C-shaped be recorded from the same set of neurons for up to track, around which three differently orientated several months (Tian et al., 2009; Andermann et al., visual stimuli were arranged. Within the imaged 2010) without having to open the recording chamber neuronal populations, some neurons were found to and reapply the Ca2+ indicator. be preferentially activated as the animal’s gaze swept across one of the three monitors, consistent with the Several recent studies have begun to make use known orientation preferences observed in rat visual of these new opportunities to measure activity in cortex (Girman et al., 1999; Ohki et al., 2005). ensembles of neurons and how behavioral outcomes are reflected by these ensembles’ activity (Dombeck Imaging in head-fixed animals retains several et al., 2009; Komiyama et al., 2010). The recent substantial advantages and is likely to remain the development of a spherical treadmill (Styrofoam preeminent imaging method for studying awake ball) on which an animal was free to run while behaving animals. However, the recent finding head-fixed provides great opportunities for imaging that peak visual responses in head-fixed mice on a in awake behaving animals (Dombeck et al., 2007). treadmill were significantly greater if the stimuli were Most recently, this treadmill was used to provide presented as the animal was running, compared with compelling evidence for functional clustering of when it was motionless (Niell and Stryker, 2010), neurons into ensembles correlated with either suggests that free movement may have a greater running or grooming movements of the contralateral impact on recorded neuronal responses than may forelimb (Dombeck et al., 2009) (Fig. 1d–e). otherwise have been anticipated. Another recent study identified two cortical areas involved in licking in mice (Komiyama et al., 2010). Conclusion The group demonstrated different populations of The Hebbian cell assembly hypothesis was published neurons whose activity reflected different behavioral more than 60 years ago. Since then, a large amount choices the animals made. They also showed that of detailed knowledge about the neurons that closely neighboring neurons (within ~150 µm)— make up the cortex has been gathered. The next particularly those with the same response type— challenge is placing this knowledge back into often showed substantial synchronous activity and the context of the behaving animal. The recent that these temporal correlations increased as the emergence of techniques that enable optical animals’ behavioral performance improved. Another imaging of activity from large neuronal populations, recent advance has demonstrated population Ca2+ optical manipulation of activity, and large-scale imaging in the visual cortex of awake head-fixed reconstruction of neuronal circuits offers new mice performing a visual orientation discrimination opportunities to accurately correlate behavior and task (Andermann et al., 2010). This study found activity with circuit anatomy. Although the ability to © 2012 Kerr 48

NOTES simultaneously record activity in progressively larger Dombeck DA, Graziano MS, Tank DW (2009) neuronal populations is advantageous, because of the Functional clustering of neurons in motor cortex vast number of the potential member neurons in a determined by cellular resolution imaging in awake functional cell assembly, testing the cell assembly behaving mice. J Neurosci 29:13751–13760. hypothesis will most likely require the combination Douglas RJ, Martin KA (2004) Neuronal circuits of of all these approaches. the neocortex. Annu Rev Neurosci 27:419–451. Acknowledgment Douglas RJ, Martin KA (2007) The butterfly and the loom. Res Rev 55:314–328. This chapter is a vastly redacted version of an article that was originally published as “Chasing the Duemani Reddy G, Kelleher K, Fink R, cell assembly” in Current Opinion in Neurobiology Saggau P (2008) Three-dimensional random 2010;20(3):296–305. access multiphoton microscopy for functional imaging of neuronal activity. Nat Neurosci References 11:713–720. Andermann ML, Kerlin AM, Reid RC (2010) Ecker AS, Berens P, Keliris GA, Bethge M, Chronic cellular imaging of mouse visual cortex Logothetis NK, Tolias AS (2010) Decorrelated during operant behavior and passive viewing. neuronal firing in cortical microcircuits. Science Front Cell Neurosci 4:3. 327:584–587. Botcherby EJ, Juskaitis R, Booth MJ, Wilson T Fee MS, Leonardo A (2001) Miniature motorized (2008) An optical technique for remote focusing microdrive and commutator system for chronic inmicroscopy. Opt Comm 281:880–887. neural recording in small animals. J Neurosci Braitenberg V (1978) Cell assemblies in the cerebral Methods 112:83–94. cortex. In: Theoretical approaches to complex Flusberg BA, Nimmerjahn A, Cocker ED, systems, Lecture notes in biomathematics Mukamel EA, Barretto RP, Ko TH, Burns LD, (Heim R, Palm G, eds). Berlin and New York: Jung JC, Schnitzer MJ (2008) High-speed, Springer:171–188. miniaturized fluorescence microscopy in freely Braitenberg V, Schüz A (1991) Anatomy of the moving mice. Nat Methods 5:935–938. cortex: Statistics and geometry. Berlin and New Gerstein GL, Bedenbaugh P, Aertsen MH (1989) York: Springer. Neuronal assemblies. IEEE Trans Biomed Eng Braitenberg V, Schüz A (1998) Cortex: Statistics 36:4–14. and geometry of neuronal connectivity, 2nd ed. Gerstner W, Ritz R, van Hemmen JL (1993): Berlin and New York: Springer. Why spikes? Hebbian learning and retrieval of Cox CL, Denk W, Tank DW, Svoboda K (2000) time-resolved excitation patterns. Biol Cybern, Action potentials reliably invade axonal arbors of 69:503–515. rat neocortical neurons. Proc Natl Acad Sci USA Girman SV, Sauve Y, Lund RD (1999) Receptive 97:9724–9728. field properties of single neurons in rat primary de Kock CP, Bruno RM, Spors H, Sakmann B (2007) visual cortex. J Neurophysiol 82:301–311. Layer- and cell-type-specific suprathreshold Gobel W, Kampa BM, Helmchen F (2007) Imaging stimulus representation in rat primary cellular network dynamics in three dimensions somatosensory cortex. J Physiol 581:139–154. using fast 3D laser scanning. Nat Methods 4:73– de Kock CP, Sakmann B (2009) Spiking in primary 79. somatosensory cortex during natural whisking in Greenberg DS, Kerr JN (2009) Automated correction awake head-restrained rats is cell-type specific. of fast motion artifacts for two-photon imaging of Proc Natl Acad Sci USA 106:16446–16450. awake animals. J Neurosci Methods 176:1–15. Denk W, Strickler JH, Webb WW (1990) Two- Greenberg DS, Houweling AR, Kerr JN (2008) photon laser scanning fluorescence microscopy. Population imaging of ongoing neuronal activity Science 248:73–76. in the visual cortex of awake rats. Nat Neurosci Dombeck DA, Khabbaz AN, Collman F, Adelman 11:749–751. TL, Tank DW (2007) Imaging large-scale neural Grewe BF, Helmchen F (2009) Optical probing of activity with cellular resolution in awake, mobile neuronal ensemble activity. Curr Opin Neurobiol mice. Neuron 56:43–57. 19:520–529.

© 2012 Kerr Chasing the Cortical Assembly 49

NOTES Grewe BF, Langer D, Kasper H, Kampa BM, Lorente de No R (1949) : architecture, Helmchen F (2010) High-speed in vivo calcium intracortical connections, motor projections. In: imaging reveals spike trains in neuronal networks Physiology of the , 3rd ed. (Fulton with near-millisecond precision. Nat Methods JF, ed). New York: Oxford UP: 288–330. 7:399–405. Losonczy A, Makara JK, Magee JC (2008) Harris KD (2005) Neural signatures of cell assembly Compartmentalized dendritic plasticity and input organization. Nat Rev Neurosci 6:399–407. feature storage in neurons. Nature 452:436–441. Hasan MT, Friedrich RW, Euler T, Larkum ME, Mank M, Santos AF, Direnberger S, Mrsic-Flogel TD, Giese G, Both M, Duebel J, Waters J, Bujard H, Hofer SB, Stein V, Hendel T, Reiff DF, Levelt C, Griesbeck O, Tsien RY, Nagai T, Miyawaki A, Borst A, Bonhoeffer T, Hübener M, Griesbeck O Denk W (2004) Functional fluorescent Ca2+ (2008) A genetically encoded calcium indicator indicator proteins in transgenic mice under TET for chronic in vivo two-photon imaging. Nat control. PLoS Biol 2:e163. Methods 5:805–811. Hasenstaub A, Shu Y, Haider B, Kraushaar U, Mittmann W, Wallace DJ, Czubayko U, Herb JT, Duque A, McCormick DA (2005) Inhibitory Schaefer AT, Looger LL, Denk W, Kerr JN (2011) postsynaptic potentials carry synchronized Two-photon calcium imaging of evoked activity frequency information in active cortical networks. from L5 somatosensory neurons in vivo. Nat Neuron 47:423–435. Neurosci 14:1089–1093. Hebb DO (1949) The organization of behavior: a Mountcastle VB (1978) An organizing principle for neuropsychological theory. New York: Wiley. cerebral function. In: The mindful brain. (Edelman Iyer V, Hoogland TM, Saggau P (2006) Fast functional GM, Mountcastle VB, eds) Cambridge, MA: MIT imaging of single neurons using random-access Press:7–50. multiphoton (RAMP) microscopy. J Neurophysiol Mountcastle VB (1997) The columnar organization 95:535–545. of the neocortex. Brain 120(Pt 4):701–722. Jia H, Rochefort NL, Chen X, Konnerth A (2010) Mountcastle VB (2003) Introduction. Computation Dendritic organization of sensory input to cortical in cortical columns. Cereb Cortex 13:2–4. neurons in vivo. Nature 464:1307–1312. Mrsic-Flogel TD, Hofer SB, Ohki K, Reid RC, Kerr JN, Denk W (2008) Imaging in vivo: Watching Bonhoeffer T, Hubener M (2007) Homeostatic the brain in action. Nat Rev Neurosci 9:195–205. regulation of eye-specific responses in visual Kerr JN, Greenberg D, Helmchen F (2005) Imaging cortex during ocular dominance plasticity. Neuron input and output of neocortical networks in vivo. 54:961–972. Proc Natl Acad Sci USA 102:14063–14068. Niell CM, Stryker MP (2010) Modulation of visual Kerr JN, de Kock CP, Greenberg DS, responses by behavioral state in mouse visual Bruno RM, Sakmann B, Helmchen F (2007) cortex. Neuron 65:472–479. Spatial organization of neuronal population Ohki K, Chung S, Ch'ng YH, Kara P, Reid RC (2005) responses in layer 2/3 of rat barrel cortex. J Neurosci Functional imaging with cellular resolution reveals 27:13316–13328. precise micro-architecture in visual cortex. Nature Koester HJ, Sakmann B (2000) Calcium dynamics 433:597–603. associated with action potentials in single nerve Palm G (1982) Neural assemblies: An alternative terminals of pyramidal cells in layer 2/3 of the approach to artificial intelligence. Berlin and New young rat neocortex. J Physiol 529(Pt 3):625–646. York: Springer. Komiyama T, Sato TR, O'Connor DH, Zhang YX, Reddy GD, Saggau P (2005) Fast three-dimensional Huber D, Hooks BM, Gabitto M, Svoboda K laser scanning scheme using acousto-optic (2010) Learning-related fine-scale specificity deflectors. J Biomed Opt 10:064038. imaged in motor cortex circuits of behaving mice. Rothschild G, Nelken I, Mizrahi A (2010) Nature 464:1182–1186. Functional organization and population dynamics Larkum ME, Nevian T (2008) Synaptic clustering in the mouse primary auditory cortex. Nat Neurosci by dendritic signalling mechanisms. Curr Opin 13:353–360. Neurobiol 18:321–331. Sato TR, Gray NW, Mainen ZF, Svoboda K (2007)

© 2012 Kerr 50

NOTES The functional microarchitecture of the mouse Vogelstein JT, Watson BO, Packer AM, Yuste R, barrel cortex. PLoS Biol 5:e189. Jedynak B, Paninski L (2009) Spike inference Sawinski J, Wallace DJ, Greenberg DS, Grossmann S, from calcium imaging using sequential Monte Denk W, Kerr JN (2009) Visually evoked activity Carlo methods. Biophys J 97:636–655. in cortical cells imaged in freely moving animals. von Economo K, Koskinas GN (1925) Die Proc Natl Acad Sci USA 106:19557–19562. Cytoarchitektonik der Hirnrinde des erwachsenen Spruston N (2008) Pyramidal neurons: dendritic Menschen. Berlin: Springer. structure and synaptic integration. Nat Rev Vucinic D, Sejnowski TJ (2007) A compact Neurosci 9:206–221. multiphoton 3D imaging system for recording fast Stosiek C, Garaschuk O, Holthoff K, Konnerth A neuronal activity. PLoS One 2:e699. (2003) In vivo two-photon calcium imaging of Wallace DJ, Meyer zum Alten Borgloh S, Astori S, neuronal networks. Proc Natl Acad Sci USA Yang Y, Bausen M, Kugler S, Palmer AE, Tsien RY, 100:7319–7324. Sprengel R, Kerr JN, Denk W, Hasan MT (2008) Svoboda K, Denk W, Kleinfeld D, Tank DW (1997) Single-spike detection in vitro and in vivo with a 2+ In vivo dendritic calcium dynamics in neocortical genetic Ca sensor. Nat Methods 5:797–804. pyramidal neurons. Nature 385:161–165. Waters J, Helmchen F (2006) Background synaptic Theer P, Denk W (2006) On the fundamental activity is sparse in neocortex. J Neurosci imaging-depth limit in two-photon microscopy. 26:8267–8277. J Opt Soc Am A Opt Image Sci Vis 23:3139–3149. Wickens JR, Miller RA (1997) A formalisation of Tian L, Hires SA, Mao T, Huber D, Chiappe the neural assembly concept 1. Constraints on ME, Chalasani SH, Petreanu L, Akerboom J, neural assembly size. Biol Cybern 77:351–358. McKinney SA, Schreiter ER, Bargmann CI, Yaksi E, Friedrich RW (2006) Reconstruction of Jayaraman V, Svoboda K, Looger LL (2009) firing rate changes across neuronal populations Imaging neural activity in worms, flies and mice by temporally deconvolved Ca2+ imaging. Nat with improved GCaMP calcium indicators. Nat Methods 3:377–383. Methods 6:875–881.

© 2012 Kerr