Proc. Natl. Acad. Sci. USAVol. 94, pp. 10403–10408, September 1997NeurobiologyContext-sensitive synaptic plasticity and temporal-to-spatialtransformations in hippocampal slicesDEAN V. BUONOMANO*, PETER W. HICKMOTT, AND MICHAEL M. MERZENICHKeck Center for Integrative Neuroscience, University of California, San Francisco, San Francisco, CA 94143Edited by Lily Yeh Jan, University of California, San Francisco, San Francisco, CA, and approved July 16, 1997 (received for review May 27, 1997)ABSTRACT Hippocampal slices are used to show that, as atemporal input pattern of activity flows through a neuronallayer, a temporal-to-spatial transformation takes place. That is,neurons can respond selectively to the first or second of a pair ofinput pulses, thus transforming different temporal patterns ofactivity into the activity of different neurons. This is demon-strated using associative long-term potentiation of polysynapticCA1 responses as an activity-dependent marker: by depolarizinga postsynaptic CA1 neuron exclusively with the first or second ofa pair of pulses from the dentate gyrus, it is possible to ‘‘tag’’different subpopulations of CA3 neurons. This technique allowssampling of a population of neurons without recording simul-taneously from multiple neurons. Furthermore, it reflects abiologically plausible mechanism by which single neurons maydevelop selective responses to time-varying stimuli and permitsthe induction of context-sensitive synaptic plasticity. These ex-perimental results support the view that networks of neurons areintrinsically able to process temporal information and that it isnot necessary to invoke the existence of internal clocks or delaylines for temporal processing on the time scale of tens tohundreds of milliseconds.Sensory stimuli produce spatio-temporal patterns of activityon the peripheral sensory layers of the nervous system. De-pending on the nature of the stimuli and of the task, the spatialandyor temporal features of these activity patterns are used bythe central nervous system to make perceptual and behavioraljudgments. In temporal processing, information is encoded inthe temporal pattern of activity on the sensory layer (i.e., byinterval, duration, and order of different stimulus features).Determining whether two flashes of light or two brief tones areseparated by 100 or 150 ms is an example of a temporal task(1, 2). Temporal processing is important for sensory processingin most sensory modalities but is perhaps most important inspeech processing. Speech is rich in temporal structure, par-ticularly temporal features on the time scale of tens to hun-dreds of milliseconds (3, 4). Indeed, recent data suggest thatcertain elements of speech can be identified based primarily ontemporal cues (5). Furthermore, deficits in temporal process-ing may underlie certain types of learning disabilities (6–8).In spatial tasks, information is encoded in the spatial patternof active sensory afferents. Orientation selectivity or theformation of topographic maps are examples that require theformation of selective neuronal responses based on the spatialarrangement of input fibers. We have a relatively good under-standing of how neurons develop responses to specific orien-tations (e.g., ref. 9). In contrast, we have little understandingof how neurons develop selective responses to simple temporalpatterns, such as two tones separated by either a 100- or 150-msinterval. Both stimuli will activate the same hair cells in thecochlea but with a different temporal pattern. For centralneurons to respond selectively to each stimulus or to generatetwo different behavioral responses, the nervous system mustperform a temporal-to-spatial transformation. That is, infor-mation initially encoded in the temporal pattern of inputs mustultimately be encoded by different populations of neurons.We have previously proposed that networks of neurons areintrinsically able to implement temporal-to-spatial transforma-tions as a result of short-term forms of plasticity and slow synapticevents (10). Specifically, if a 100-ms interval is bounded by twoinput pulses, the first pulse will activate a subpopulation of unitsand trigger a series of processes such as paired-pulse facilitation(PPF) and slow inhibitory postsynaptic potentials. Even thoughthe second pulse is identical to the first, it will activate a differentsubpopulation of units, because the network is in a different stateas a result of the occurrence of the first pulse; some synapses willbe facilitated while some cells will be inhibited. Differences inpopulation responses to the first and second pulses can be viewedas a temporal-to-spatial transform and used to code intervals ororder. A prediction that emerges from this model is that it shouldbe possible to record differences in the activity of a population ofneurons in response to the first versus a second pulse. In thepresent paper we tested this prediction by using hippocampalslices to study the transformations that take place as activity flowsthrough a network of neurons.MATERIALS AND METHODSExperiments were performed on 500-mm-thick transverse hip-pocampal slices from Sprague–Dawley rats (15–28 days). Thehippocampus was removed following anesthesia with pentobar-bital and decapitation. Slices were cut and submerged in oxygen-ated medium composed of 119 mM NaCl, 2.5 mM KCl, 1.3 mMMgSO4, 1.0 mM NaH2PO4, 26.2 mM NaCO3, 2.5 mM CaCl2, and10 mM glucose. After an equilibrium period of at least 1 hr, sliceswere transferred to a recording chamber, perfused at a rate of 2mlymin, and maintained at a temperature of 30–31°C. Intracel-lular recordings from CA1 or CA3 pyramidal neurons were madewith 40–100 MV electrodes filled with 3 M KAc.Polysynaptic postsynaptic potentials were recorded in CA1pyramidal neurons in response to paired-pulse stimulation of thedentate gyrus (DG). Stainless-steel bipolar electrodes wereplaced in the granule or molecular layer of the DG, generally inthe suprapyramidal blade. The interpulse interval for paired-pulse stimulation was 100 or 200 ms. Stimulation intensity wasbetween 50 and 200mA, with 100-ms-long pulses. By stimulatingthe DG and recording in CA1 we tapped into the hippocampaltrisynaptic circuit: the perforant path projects to the DG, whichprojects to CA3, which in turn projects to CA1. Because westimulated the DG and recorded activity in a single CA1 neuron,the CA3 region essentially represents a ‘‘hidden layer,’’ and theCA1 neuron