Nerve cells, called neurons, are one of the core components of the brain and form complex networks by connecting to other neurons via long, thin ‘wire-like’ processes called axons. Axons can extend across the brain, enabling neurons to form connections—or synapses—with thousands of others. It is through these complex networks that incoming information from sensory organs, such as the eye, is propagated through the brain and encoded.

The basic unit of communication between neurons is the action potential, often called a ‘spike’, which propagates along the network of axons and, through a chemical process at synapses, communicates with the postsynaptic neurons that the axon is connected to. These action potentials excite the neuron that they arrive at, and this excitatory process can generate a new action potential that then propagates along the axon to excite additional target neurons. In the visual areas of the cortex, neurons respond with action potentials when they ‘recognize’ a particular feature in a scene—a process called tuning. How a neuron becomes tuned to certain features in the world and not to others is unclear, as are the rules that enable a neuron to change what it is tuned to. What is clear, however, is that to understand this process is to understand the basis of sensory perception.

Memory storage and formation is thought to occur at synapses. The efficiency of signal transmission between neurons can increase or decrease over time, and this process is often referred to as synaptic plasticity. But for these synaptic changes to be transmitted to target neurons, the changes must alter the number of action potentials. Although it has been shown in vitro that the efficiency of synaptic transmission—that is the strength of the synapse—can be altered by changing the order in which the pre- and postsynaptic cells are activated (referred to as ‘Spike-timing-dependent plasticity’), this has never been shown to have an effect on the number of action potentials generated in a single neuron in vivo. It is therefore unknown whether this process is functionally relevant.

Now Pawlak et al. report that spike-timing-dependent plasticity in the visual cortex of anaesthetized rats can change the spiking of neurons in the visual cortex. They used a visual stimulus (a bar flashed up for half a second) to activate a presynaptic cell, and triggered a single action potential in the postsynaptic cell a very short time later. By repeatedly activating the cells in this way, they increased the strength of the synaptic connection between the two neurons. After a small number of these pairing activations, presenting the visual stimulus alone to the presynaptic cell was enough to trigger an action potential (a suprathreshold response) in the postsynaptic neuron—even though this was not the case prior to the pairing.

This study shows that timing rules known to change the strength of synaptic connections—and proposed to underlie learning and memory—have functional relevance in vivo, and that the timing of single action potentials can change the functional status of a cortical neuron.

The formation of sensory percepts in the mammalian cortex is thought to involve the propagation of action potentials through populations of connected neurons in response to sensory stimulation (Hebb, 1949; Harris, 2005). Rules that dictate changes in the strength, or efficacy, of these connections rely on persistent correlated activity between the connected neurons (for recent review, see Abraham, 2008). For these changes in synaptic efficacy to be functional and recognized by the down-stream neurons targeted by the axon, the changes need to influence AP patterns. Plasticity induction involving precisely timed action-potentials, known as spike-timing-dependent plasticity (Markram et al., 1997; Bi and Poo, 1998), has been suggested as a candidate mechanism to determine which stimulus driven inputs are strengthened and weakened (out of the many possible sensory driven inputs). What makes STDP rules so appealing for modification of synaptic strength in the cortex is that the low numbers of APs and low repetition frequencies used to induce plastic changes in STDP protocols (Feldman, 2012) are consistent with low AP firing rates observed in awake freely moving animals exploring a novel environment (Lee et al., 2006; Sawinski et al., 2009). In addition, cortical neurons are capable of AP-responses that are temporally precise over multiple trials (Bair and Koch, 1996; Marsalek et al., 1997; Tchumatchenko et al., 2011), similar to the precision typically used during STDP pairing protocols (for recent review see, Feldman, 2012). Lastly, most modeling studies of synaptic plasticity using STDP rules assume that synaptic modification leads to changes in spiking (Gerstner et al., 1996; Song and Abbott, 2001) with a recent study suggesting that STDP underlies sensory map plasticity in the young and adult cortex (Young et al., 2007). However, evidence for single-neuron STDP in vivo remains scarce and confined to subthreshold changes (Meliza and Dan, 2006; Jacob et al., 2007) and remains controversial (Lisman and Spruston, 2005, 2010; Fregnac et al., 2010; Shouval et al., 2010). While extracellular studies have suggested STDP can modify spiking patterns in anesthetized (Fu et al., 2002) and freely moving animals (Celikel et al., 2004), this evidence remains difficult to interpret without information on single-cell membrane potential dynamics.

The visual cortex is perfectly suited to study STDP since visual cortex neuronal responses and receptive fields are known to be highly plastic (Gilbert, 1998), and sensory maps can undergo plastic changes with experience (Daw and Wyatt, 1976; Li et al., 2008) or after removal of visual input (Wiesel and Hubel, 1963; Kaas et al., 1990; Gilbert and Wiesel, 1992). Finally during visual stimulation, the action potential output for many visual cortex neurons is tuned to specific locations in the visual field and to specific stimulus features (Hubel and Wiesel, 1959), even though each neuron receives synaptic inputs corresponding to surrounding locations or many different stimulus features (Ferster, 1986; Jia et al., 2010). Together these conditions make a perfect system to test whether STDP rules can, with a few tens of spikes, reliably change a subtreshold responding neuron into a neuron responding to the same stimulus with APs and change a neuron's visually driven AP patterns in a timing dependent manner.

We addressed these questions by pairing precisely timed postsynaptic spikes to presynaptic inputs arising from visual stimulation. We used STDP protocols similar to those previously applied in vivo (Meliza and Dan, 2006), but with the difference that all ongoing membrane potential voltage fluctuations and spiking at the cell soma were measured and taken into account when presenting the stimulus, since large membrane potential fluctuations (Cowan and Wilson, 1994) can alter a neuron's spiking responses (Petersen et al., 2003; Sachdev et al., 2004) and ongoing spiking may influence plasticity outcome (Sjostrom et al., 2001). We targeted whole-cell recordings (Margrie et al., 2003) to Layer 2/3 (L2/3) neurons in the binocular primary visual cortex (n = 58 rats) using intrinsic imaging followed by two-photon microscopy as using stereotaxic coordinates alone can be inaccurate (Slotnick and Brown, 1980; Kline and Reid, 1984). All neurons (n = 71, yielding 72 data points, see ‘Materials and methods’) were identified as L2/3 pyramidal neurons, based on firing pattern, location below the pia and morphology after post hoc histological staining. For visual stimulation, a bar was flashed (500 ms) (Figure 1A) every 3–4 s at one of four non-overlapping spatial positions (Figure 1B). To measure ongoing spiking and membrane potential fluctuations, and to reduce variance and maximize consistency in the evoked responses (Petersen et al., 2003; Sachdev et al., 2004), we recorded in current clamp mode and always presented stimuli after the membrane potential had transitioned from an up-state into a down-state (Figure 1C), during which far fewer spontaneous synaptic inputs occur (Waters and Helmchen 2006).

Figure 1

Download asset Open asset

All four stimulus positions evoked either subthreshold or suprathreshold responses, consistent with the broad receptive fields of neurons in younger rats (Fagiolini et al., 1994). Mean subthreshold responses reached an initial peak 50–150 ms after stimulus onset, then decayed to a trough within 100–200 ms (Figure 1D), though considerable variability (Hirsch et al., 2002) was observed in the peak time, response shape, and amplitude across neurons (Figure 1D). After a stable baseline period (9.0 ± 1.3 min, 32.4 ± 1.9 presentations of each stimulus, n = 32 recordings), we applied an STDP induction protocol by pairing single-APs, evoked by brief depolarizing current pulses, with visual stimulation at only one of the four stimulus positions (42.0 ± 1.5 pairing trials, 1.3 ± 0.16 APs/trial, n = 32, Figure 1E). Since each neuron's response time was different, the timing of the pairing spikes was individually chosen for each neuron to occur either before (‘negative timing inductions’, n = 17, example in Figure 1E, left) or after the initial peak (‘positive timing inductions’, n = 15, example in Figure 1E, right). As a result, by taking all recordings together based on whether the pairing spike was delivered before (red, negative timing) or after (green, positive timing) the initial peak, the time of pairing spikes ranged from 82 ms before the response peak to 62 ms after (Figure 1F). After pairing, we again recorded the neuron's responses to stimuli at each of the four positions without injecting current pulses (51.9 ± 3.6 presentations of each stimulus over 31.8 ± 2.4 min, starting 1.0 ± 0.4 min after the last paired stimulus, n = 32, Figure 2A). Responses at the paired position were changed, with the direction of the change strongly dependent on the whether the pairing spike appeared before or after the response peak (Figure 2B,C). As expected (for review see, Bi and Poo, 2001) in neurons that received negative timing inductions the peak voltage amplitude significantly decreased by 26.4 ± 4.6% (n = 17) or 3.5 ± 0.7 mV (Figure 2C, p<0.0001), while positive timing inductions increased the peak voltage amplitude by 22.5 ± 8.7 % (n = 15) or 1.4 ± 0.6 mV (Figure 2C, p=0.025). The difference between what was observed here compared to previous in vitro (Bi and Poo, 2001) and in vivo studies (Fregnac et al., 2010) was that although the direction of the plasticity observed in responses to visual stimulation depended on spike timing, the largest initial peak changes were not around t = 0 but were scattered from 0–50 ms on either side of t = 0.

Figure 2

Download asset Open asset

The small changes in membrane potential observed here and in other studies raise the issue of whether the observed plasticity is sufficient to modify spiking responses in the postsynaptic neuron. Synaptic plasticity of a given neuron's inputs is functionally effective only if it alters neuronal spiking, so that information is passed on to downstream synaptic targets. To link STDP to changes in spiking at the single cell level, we therefore tested whether a sensory-driven but subthreshold-responding neuron (Figure 2D, left) could be changed through STDP induction into a suprathreshold-responding neuron (Figure 2D, right). In most neurons that received positive timing inductions sensory-evoked spiking rates increased after the pairing protocol (Figure 2E). Specifically, of the 15 neurons tested, 11 showed an increase in the average number of APs evoked by stimulus presentation, 1 showed a decrease, and 3 never discharged a stimulus-evoked AP (Figure 2G). Strikingly, in eight neurons the paired stimulus evoked spiking only after induction, that is over 50% transitioned from non-spiking to spiking. Over the 15 neurons tested, an additional 0.11 ± 0.06 APs per stimulus were observed for the paired stimulus position after induction (p=0.01, initial response 0.06 ± 0.04 APs per stimulus). In four cases increases in AP firing were significant at the single-neuron level (p<0.05), despite the modest numbers of stimulus presentation trials and low firing rates (see ‘Materials and methods’).

We next tested whether the STDP protocol could decrease the sensory-evoked spiking rates using negative timing inductions, in which pairing spikes occurred before the initial peak. For this to occur, we identified a stimulus position that evoked spiking in the recorded neuron (successful in 15 out of 17 recordings, 0.84 ± 0.23 APs per stimulus), and then proceeded to pair spiking with stimulus presentation at this position. After pairing, the average number of APs evoked by stimulus presentation decreased (0.18 ± 0.10 APs per stimulus after induction, p=0.04, n = 15). Of these 15 neurons 12 showed a decrease in spiking and 3 an increase after induction (Figure 2H). These changes were statistically significant at the single-neuron level (p<0.05) in four cases, all of which showed a decrease in spiking. Together this shows that an STDP protocol that induces small bidirectional changes in the membrane potential can both increase and decrease stimulus-evoked spiking depending on the relative timing of the single pairing spikes.

These changes in spiking responses observed after STDP induction could not be accounted for by neurons becoming more depolarized (down-state membrane potential −64.8 ± 1.5 mV before vs −64.2± 1.5 mV after, p=0.61, n = 32) or by input resistance changes (323 ± 39 MΩ before vs 419 ± 54 MΩ after, p=0.16), and neither depolarization nor input resistance changes were correlated with changes in spiking (p>0.05). In a control group, where visual stimuli were presented but no APs were paired, stimulus-driven spiking rates in the second half of the recording were not significantly different from the first half (p=0.20, four neurons with four positions each), indicating that the increase in spiking was due to the STDP protocol. Finally, we note that increases in AP firing at the paired position for positive timing inductions remained significant even after removing the two neurons exhibiting the greatest increases (p=0.036, n = 13). Together, these results provide a robust demonstration that repeated pairing of a single AP to a visual stimulus ∼40 times can initiate spiking responses in non-spiking neurons and can increase or diminish existing spiking responses. This is consistent with an STDP-driven change in functional status and provides experimental validation that STDP could provide a plasticity mechanism for changing receptive fields (Young et al., 2007).

We next measured whether induced spiking changes also extended within the receptive field to non-paired stimulus positions (Figure 3A). For positive timing experiments, we observed diverse changes in the spiking responses measured at non-paired positions as spiking elicited at positions directly adjacent to the paired position (±1) increased in some neurons (Figure 3B–D) but decreased in others (Figure 3E), and the position eliciting the maximum response often changed after pairing (Figure 3D,E). Compared to the average 0.11 ± 0.05 additional APs per stimulus evoked at the pairing position after induction, stimulus positions immediately adjacent to the pairing position also showed an increase nearly as large (0.10 ± 0.05 APs per stimulus, p=0.02, n = 15). This increase in AP firing diminished for more distant positions (Figure 3G) and was no longer significant for stimulus positions not immediately adjacent to the paired position (p>0.05). In negative timing experiments, changes in AP firing were not significant at any non-paired positions (p>0.05).

Figure 3

Download asset Open asset

We further examined whether these changes displayed any common pattern, and might admit some simple description. To test this we considered neurons' spiking responses as functions of stimulus position, and calculated the ‘stimulus response centers’ of these functions in the space of stimulus positions (Figure 3D,E, dashed lines). Stimulus response centers were calculated as weighted averages of the four stimulus positions, with each position weighted by the rate of stimulus-evoked spiking. This analysis showed a simple but robust trend in the positive timing experiments: after stimulus-AP pairing, the stimulus response center moved toward the pairing position (Figure 3F). The shift in the suprathreshold response center was strongly correlated with the difference between pairing position and the response center before pairing (R = 0.81, p=0.008). For the negative timing inductions the majority of response centers moved away from the pairing position, but these changes were small and the trend was not significant (p=0.70).

We next examined whether the shifts in response centers observed for suprathreshold responses were also reflected in subthreshold responses. Subthreshold response centers, calculated by weighting each position by the peak amplitude, also shifted toward the pairing position for positive timings, albeit much less robustly than the suprathreshold shifts (R = 0.53, p=0.04, Figure 4A–C). As with the suprathreshold response centers, for negative timings subthreshold response centers tended to move away from the paired position, but these changes were too small to be significant (p>0.05, Figure 4C). This consistency between the suprathreshold and the subthreshold response center changes brought about by an STDP induction protocol shows that plasticity of stimulus responses was timing-specific and spatially structured, with transfer to neighboring regions of the receptive field. These results show that STDP protocols induce a rearrangement of the sub- and suprathreshold receptive field structure, especially for positive timing protocols.

Figure 4

Download asset Open asset

A key difference between the present study and previous in vitro STDP experiments (Bi and Poo, 1998), where activation of presynaptic inputs is temporally precise and brief, is that in vivo during sensory stimulation cortical L2/3 neurons receive a more complex and longer-lasting synaptic input barrage (Figures 1D, 2B) with variable timing of individual inputs (Fagiolini et al., 1994; Niell and Stryker, 2008). Despite this in vivo variability, traditional STDP rules dictate that inputs consistently arriving before the pairing spike should be potentiated and increase the response amplitude before the spike, while those arriving consistently after should be depressed and reduce the response amplitude after the spike. Finally, if each response is, on average, made up of the same synaptic inputs occurring in the same relative temporal order, then the stimulus-AP pairing should not simply change the peak amplitude, but rather induce a complex change that varies continuously over the time course of the subthreshold response relative to the pairing spike (Figure 5A, same data as Figure 2B). We therefore calculated the time course of changes in the subthreshold response relative to the time of the pairing spike for all neurons (positive and negative timing inductions, n = 32, Figure 5B, grey lines). Thus, in contrast to traditional STDP analysis where a neuron provides only a single data point on a graph of time difference vs response change (Figure 2C), both potentiation and depression occur for each neuron as a continuous function of time that on average is biphasic (Figure 5B, black line) with an increase before the spike (maximum increase 0.84 mV at 64.5 ms before, p=0.04) and a decrease thereafter (maximum decrease 1.4 mV at 43.5 ms after, p=0.01). This showed that STDP protocols in vivo can simultaneously induce potentiation and depression in the membrane potential of the same neuron (Figure 2B). Changes in membrane potential were strongest at the paired position, with similarly shaped but smaller changes at adjacent positions consistent with overlap in synaptic inputs, and nondescript changes at more distant positions (Figure 5C). While the shape of these membrane potential changes was qualitatively similar to the biphasic STDP curve observed in vitro (Markram et al., 1997; Bi and Poo, 1998), the positive and negative peaks were temporally further away from the pairing spike and shape of the curve were much broader in vivo.

Figure 5 with 3 supplements see all

Download asset Open asset

We hypothesized that this peak shift and broadening effect results from variable timing of synaptic inputs from L4 and L2/3 (Fagiolini et al., 1994; Niell and Stryker, 2008), since potentiating or depressing an input with more variable jitter in the timing will potentially affect a greater range of the subthreshold stimulus response. To test this, we built an STDP model with a plasticity rule (Gerstner et al., 1996; Song et al., 2000) consistent with previous in vitro studies (Figure 5D–K) to examine STDP effects for various presynaptic activity patterns (Figure 5E,H). The membrane voltage in the neuron model is the sum of excitatory postsynaptic potentials (EPSPs), each described as a double exponential with rise time 2 ms and decay time 10 ms. Amplitudes of EPSPs were chosen from an exponential distribution with mean amplitude of 2 mV. With this set of parameters and 400 presynaptic inputs, the summed EPSPs had amplitudes and time courses similar to those measured in the experiment. Since the model was designed to describe experiments where spiking is induced artificially by the experimenter, firing is not an intrinsic property of the model but firing times were set ‘by hand’ at the target times designated for current injection by the experimental protocol (Figure 5D). Modeling each input's stimulus response as fixed and temporally precise produced membrane potential changes similar to those reported for in vitro STDP timing curves (Figure 5F). However, introducing jitter into the presynaptic activity (Figure 5G) caused the positive and negative peaks of the change in membrane potential to broaden and become more distant from the pairing spike with increasing jitter (Figure 5G). With a presynaptic jitter of 40 ms the model produced a biphasic change in the mean membrane potential (Figure 5H,I) that closely resembled the in vivo experimental results (Figure 5B), with a maximal potentiation 50 ms before and maximal depression 74 ms after the postsynaptic spike. The model also reproduced partial transfer of STDP effects to neighboring stimulus positions (Figure 5J, red) in a spike timing dependent manner (Figure 5K) and given the assumption that a subset of presynaptic neurons respond with similar spiking patterns for multiple stimulus positions. These modeling results show that the membrane potential changes observed in vivo are broader than those observed in vitro and that the broader time course can be explained by variably-timed inputs (but see Lin et al., 2003; Froemke et al., 2005; Zhang et al., 2009 for alternative sources of broader STDP windows). Additionally, the slight transfer of changes in synaptic strength from one stimulus position to the next within a receptive field can be explained by partial overlap of the inputs activated by different stimuli as previously suggested (Meliza and Dan, 2006).

Our results demonstrate that an STDP protocol using less than 50 paired APs can convert a neuron's visually evoked responses from non-spiking to spiking, can increase and diminish existing spiking responses, and can induce a rearrangement of the neurons sub- and suprathreshold responses to stimuli presented at locations surrounding the paired location. Our current findings extend our knowledge about STDP-induced subthreshold changes in vivo (Meliza and Dan, 2006; Jacob et al., 2007) to the suprathreshold domain. Our study also provides a bridge between the changes in spiking observed across many neurons' receptives fields after alteration of sensory input for example through retinal lesions (Wiesel and Hubel, 1963; Kaas et al., 1990; Gilbert and Wiesel, 1992) and subtle changes in subthreshold responses induced by STDP at the single neuron level both in vitro (Magee and Johnston, 1997; Markram et al., 1997; Bi and Poo, 1998) and in vivo (Meliza and Dan, 2006; Jacob et al., 2007). Such a connection between receptive field plasticity, rearrangement of a neuron's tuning properties and STDP has been suggested by theoretical studies (Song and Abbott, 2001; Young et al., 2007), but to date has not been shown experimentally. As the rates of AP activity used in the pairing protocols was similar to activity reported for upper cortical layer populations in the visual cortex of both headfixed awake (Greenberg et al., 2008) and freely exploring rodents (Sawinski et al., 2009), this form of plasticity could represent a more general plasticity principle that relies on low firing rates and is central to individual neurons joining or leaving cell-assemblies, as proposed by Donald Hebb (Hebb, 1949).

In the present study we manipulated the timing of spikes relative to an avalanche of stimulus-triggered presynaptic events and observed significant changes in the firing properties of the postsynaptic neuron. In other words, we focused on manifestations of Hebbian plasticity that depended upon spike timing. We cannot exclude, however, that purely subthreshold manipulation of the postsynaptic voltage could have caused similar, or even stronger effects of Hebbian plasticity (Lisman and Spruston, 2005, 2010; Fregnac et al., 2010; Shouval et al., 2010). While STDP highlights the timing-dependence on a millisecond time scale, Hebbian synaptic plasticity also manifests itself in experimental protocols manipulating firing rates and postsynaptic voltage (Artola et al., 1990; Sjostrom et al., 2001). In fact, previous models assuming that Hebbian synaptic plasticity depends critically on postsynaptic voltage, have allowed general theories of plasticity with both rate- and timing-dependent plasticity as special cases (Clopath et al., 2010). Optimality theories of STDP developed over the last 10 years, have proposed that the fundamental reason why pre-before-post induces LTP is to strengthen the causal link between input and output. Thus, the optimal timing between pre- and post should be related to the rise time or width of the PSP caused by the presynaptic input (Pfister et al., 2006; Bohte and Mozer, 2007; Toyoizumi et al., 2007; Parra et al., 2009; Hennequin et al., 2010; Pool and Mato, 2011). There is no similar optimality theory as to why synaptic depression should be spike-triggered, as unspecific forms of depression of synapses (not temporally linked to the output spike) would work just as well from a theoretical point of view (Pfister et al., 2006; Bohte and Mozer, 2007; Toyoizumi et al., 2007; Hennequin et al., 2010; Pool and Mato, 2011). What our study proposes is that timing based post-before-pre LTD (negative timing) is a robust mechanism for manipulating neuronal spiking responses along with the previously proposed pre-before post LTP (positive timing) (Gerstner et al., 1996; Song and Abbott, 2001).

The present study shows a robust timing effect for the change in voltage of the initial peak, where negative timing protocols produced decreased synaptic efficacy and positive timing protocols resulted in an increased synaptic efficacy, similar to that seen in vitro. What is not consistent with previous in vitro STDP studies, where synaptic efficacy changes were derived from post-synaptic potentials, is our reported lack of dependence of the peak changes on the milisecond timescale. From in vitro studies it would be expected that a pre- before post pairing with a delay during pairing of around 40 ms would produce almost no change and a pre- before post with a delay of around 2 ms would produce the most change (Bi and Poo, 1998). Our reported initial peak changes were not at t = 0 but were seen for a diverse range of timings within a time range of ∼100 ms (Figure 2C). We suggest that this discrepancy with previous studies (also see Fregnac et al., 2010) is due to the complexity of in vivo responses to visual stimulation which are made up of an avalanche of both inhibitory and excitatory synaptic events (Hirsch et al., 1998; Okun and Lampl, 2008). It is therefore an oversimplification to capture such complex responses by measuring only the peak of the initial response. Here we measured the evoked voltage across the entire time course of the response and aligned the voltage changes around the timing of pairing APs. With this approach we were able to show a very robust timing dependence, not just of an isolated synaptic input, but of the integrated inputs made up of multiple synaptic events (Figure 5A–C). The peak response of this voltage change was not always at t = 0 as would have been expected from in vitro experimental studies (Markram et al., 1997; Bi and Poo, 1998; Froemke et al., 2005) and assumed in STDP timing models. We found peak response changes at about 30–50 ms before and after the pairing spike. Our model suggests that this discrepancy is due to the very precise pre-synaptic timing during the pairing protocol from one trial to the next used in in vitro STDP studies which is contrary to the variable L4 and L2/3 visual response latencies found in vivo (Fagiolini et al., 1994; Niell and Stryker, 2008). In our STDP model, variability of input timing alone was sufficient to explain this broadening effect in our data compared to that expected from in vitro STDP studies.

The absolute voltage changes brought about by STDP in our study were on the order of 2 mV, yet such a change was sufficient to significantly alter the stimulus evoked spiking probability. Hence, our study suggests that small voltage changes have a dramatic effect on spiking output. We can only speculate on the mechanism because the dependence of spike generation on synaptically driven depolarizations is very complex (Shu et al., 2007), and somewhat unknown. This is especially the case in vivo (Hasenstaub et al., 2007) where combinations of various non-linear currents over the sub-threshold voltage range are present (Waters and Helmchen, 2006) and the spatial location of active synaptic inputs during evoked sensory stimulation (Varga et al., 2011) are not known. Despite these in vivo uncertainties, complex membrane-potential waveforms at the soma in vitro can reliably generate spikes (Mainen and Sejnowski, 1995) and recent simulations paired with experiments (Tchumatchenko et al., 2011) show that cortical pyramidal neurons can rapidly and reliably change firing rates in response to very small changes in somatic depolarization and hyperpolarization (twofold spiking difference induced with a 40 pA current injection).

In the current study we show that spike-timing-dependent plasticity rules using very few pairing APs can turn a neuron that hitherto did not emit a spike in response to a visual stimulus into one that frequently spikes, while pairing APs with a different timing can reduce established spiking. We propose that this is a fundamental mechanism by which neurons join and leave cell-assemblies (Wallace and Kerr, 2010) as previously proposed (Hebb, 1949) in which very few precisely timed spikes select specific temporal synaptic sequences for long-term storage from the many that are activated during a sensory learning process. Since these spikes will be transmitted to hundreds of other downstream neurons, such a change in tuning has direct implications for the function of neuronal circuits in which the respective neuron is embedded.

1. 1. Abraham WC

   (2008)

   Metaplasticity: tuning synapses and networks for plasticity

   Nat Rev Neurosci 9:387.

   • Google Scholar
2. 1. Artola A
   2. Brocher S
   3. Singer W

   (1990)

   Different voltage-dependent thresholds for inducing long-term depression and long-term potentiation in slices of rat visual cortex

   Nature 347:69–72.

   • Google Scholar
3. 1. Baddeley R
   2. Abbott LF
   3. Booth MC
   4. Sengpiel F
   5. Freeman T
   6. Wakeman EA

   et al. (1997)

   Responses of neurons in primary and inferior temporal visual cortices to natural scenes

   Proc Biol Sci 264:1775–1783.

   • Google Scholar
4. 1. Bair W
   2. Koch C

   (1996)

   Temporal precision of spike trains in extrastriate cortex of the behaving macaque monkey

   Neural Comput 8:1185–1202.

   • Google Scholar
5. 1. Bi G
   2. Poo M

   (2001)

   Synaptic modification by correlated activity: Hebb's postulate revisited

   Annu Rev Neurosci 24:139–166.

   • Google Scholar
6. 1. Bi GQ
   2. Poo MM

   (1998)

   Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type

   J Neurosci 18:10464–10472.

   • Google Scholar
7. 1. Bohte SM
   2. Mozer MC

   (2007)

   Reducing the variability of neural responses: a computational theory of spike-timing-dependent plasticity

   Neural Comput 19:371–403.

   • Google Scholar
8. 1. Celikel T
   2. Szostak VA
   3. Feldman DE

   (2004)

   Modulation of spike timing by sensory deprivation during induction of cortical map plasticity

   Nat Neurosci 7:534–541.

   • Google Scholar
9. 1. Clopath C
   2. Busing L
   3. Vasilaki E
   4. Gerstner W

   (2010)

   Connectivity reflects coding: a model of voltage-based STDP with homeostasis

   Nat Neurosci 13:344–352.

   • Google Scholar
10. 1. Cowan RL
    2. Wilson CJ

    (1994)

    Spontaneous firing patterns and axonal projections of single corticostriatal neurons in the rat medial agranular cortex

    J Neurophysiol 71:17–32.

    • Google Scholar
11. 1. Daw NW
    2. Wyatt HJ

    (1976)

    Kittens reared in a unidirectional environment: evidence for a critical period

    J Physiol 257:155–170.

    • Google Scholar
12. 1. Fagiolini M
    2. Pizzorusso T
    3. Berardi N
    4. Domenici L
    5. Maffei L

    (1994)

    Functional postnatal development of the rat primary visual cortex and the role of visual experience: dark rearing and monocular deprivation

    Vision Res 34:709–720.

    • Google Scholar
13. 1. Feldman DE

    (2012)

    The spike-timing dependence of plasticity

    Neuron 75:556–571.

    • Google Scholar
14. 1. Ferster D

    (1986)

    Orientation selectivity of synaptic potentials in neurons of cat primary visual cortex

    J Neurosci 6:1284–1301.

    • Google Scholar
15. 1. Fregnac Y
    2. Pananceau M
    3. Rene A
    4. Huguet N
    5. Marre O
    6. Levy M

    et al. (2010)

    A Re-examination of Hebbian-covariance rules and spike timing-dependent plasticity in cat visual cortex in vivo

    Front Synaptic Neurosci 2:147.

    • Google Scholar
16. 1. Friedberg MH
    2. Lee SM
    3. Ebner FF

    (1999)

    Modulation of receptive field properties of thalamic somatosensory neurons by the depth of anesthesia

    J Neurophysiol 81:2243–2252.

    • Google Scholar
17. 1. Froemke RC
    2. Poo MM
    3. Dan Y

    (2005)

    Spike-timing-dependent synaptic plasticity depends on dendritic location

    Nature 434:221–225.

    • Google Scholar
18. 1. Fu YX
    2. Djupsund K
    3. Gao H
    4. Hayden B
    5. Shen K
    6. Dan Y

    (2002)

    Temporal specificity in the cortical plasticity of visual space representation

    Science 296:1999–2003.

    • Google Scholar
19. 1. Gerstner W
    2. Kempter R
    3. van Hemmen JL
    4. Wagner H

    (1996)

    A neuronal learning rule for sub-millisecond temporal coding

    Nature 383:76–81.

    • Google Scholar
20. 1. Gilbert CD

    (1998)

    Adult cortical dynamics

    Physiol Rev 78:467–485.

    • Google Scholar
21. 1. Gilbert CD
    2. Wiesel TN

    (1992)

    Receptive field dynamics in adult primary visual cortex

    Nature 356:150–152.

    • Google Scholar
22. 1. Greenberg DS
    2. Houweling AR
    3. Kerr JN

    (2008)

    Population imaging of ongoing neuronal activity in the visual cortex of awake rats

    Nat Neurosci 11:749–751.

    • Google Scholar
23. 1. Grinvald A
    2. Lieke E
    3. Frostig RD
    4. Gilbert CD
    5. Wiesel TN

    (1986)

    Functional architecture of cortex revealed by optical imaging of intrinsic signals

    Nature 324:361–364.

    • Google Scholar
24. 1. Harris KD

    (2005)

    Neural signatures of cell assembly organization

    Nat Rev Neurosci 6:399–407.

    • Google Scholar
25. 1. Hasenstaub A
    2. Sachdev RN
    3. McCormick DA

    (2007)

    State changes rapidly modulate cortical neuronal responsiveness

    J Neurosci 27:9607–9622.

    • Google Scholar
26. Book

    1. Hebb DO

    (1949)

    The organization of behavior. A neuropsychological theory

    New York: John Wiley & Sons, Inc.

    • Google Scholar
27. 1. Hennequin G
    2. Gerstner W
    3. Pfister JP

    (2010)

    STDP in adaptive neurons gives close-to-optimal information transmission

    Front Comput Neurosci 4:143.

    • Google Scholar
28. 1. Hirsch JA
    2. Alonso JM
    3. Reid RC
    4. Martinez LM

    (1998)

    Synaptic integration in striate cortical simple cells

    J Neurosci 18:9517–9528.

    • Google Scholar
29. 1. Hirsch JA
    2. Martinez LM
    3. Alonso JM
    4. Desai K
    5. Pillai C
    6. Pierre C

    (2002)

    Synaptic physiology of the flow of information in the cat's visual cortex in vivo

    J Physiol 540:335–350.

    • Google Scholar
30. 1. Hubel DH
    2. Wiesel TN

    (1959)

    Receptive fields of single neurones in the cat's striate cortex

    J Physiol 148:574–591.

    • Google Scholar
31. 1. Jacob V
    2. Brasier DJ
    3. Erchova I
    4. Feldman D
    5. Shulz DE

    (2007)

    Spike timing-dependent synaptic depression in the in vivo barrel cortex of the rat

    J Neurosci 27:1271–1284.

    • Google Scholar
32. 1. Jia H
    2. Rochefort NL
    3. Chen X
    4. Konnerth A

    (2010)

    Dendritic organization of sensory input to cortical neurons in vivo

    Nature 464:1307–1312.

    • Google Scholar
33. 1. Kaas JH
    2. Krubitzer LA
    3. Chino YM
    4. Langston AL
    5. Polley EH
    6. Blair N

    (1990)

    Reorganization of retinotopic cortical maps in adult mammals after lesions of the retina

    Science 248:229–231.

    • Google Scholar
34. 1. Kerr JN
    2. Greenberg D
    3. Helmchen F

    (2005)

    Imaging input and output of neocortical networks in vivo

    Proc Natl Acad Sci USA 102:14063–14068.

    • Google Scholar
35. 1. Kerr JN
    2. Plenz D

    (2002)

    Dendritic calcium encodes striatal neuron output during up-states

    J Neurosci 22:1499–1512.

    • Google Scholar
36. 1. Kline J
    2. Reid KH

    (1984)

    Variability of bregma in 300 gram Long-Evans and Sprague-Dawley rats

    Physiol Behav 33:301–303.

    • Google Scholar
37. 1. Kubicki S
    2. Rieger H

    (1968)

    The EEG during acute intoxication with hypnotics

    Electroencephalogr Clin Neurophysiol 25:94.

    • Google Scholar
38. 1. Lee AK
    2. Manns ID
    3. Sakmann B
    4. Brecht M

    (2006)

    Whole-cell recordings in freely moving rats

    Neuron 51:399–407.

    • Google Scholar
39. 1. Li Y
    2. Van Hooser SD
    3. Mazurek M
    4. White LE
    5. Fitzpatrick D

    (2008)

    Experience with moving visual stimuli drives the early development of cortical direction selectivity

    Nature 456:952–956.

    • Google Scholar
40. 1. Lin YW
    2. Min MY
    3. Chiu TH
    4. Yang HW

    (2003)

    Enhancement of associative long-term potentiation by activation of beta-adrenergic receptors at CA1 synapses in rat hippocampal slices

    J Neurosci 23:4173–4181.

    • Google Scholar
41. 1. Lisman J
    2. Spruston N

    (2005)

    Postsynaptic depolarization requirements for LTP and LTD: a critique of spike timing-dependent plasticity

    Nat Neurosci 8:839–841.

    • Google Scholar
42. 1. Lisman J
    2. Spruston N

    (2010)

    Questions about STDP as a general model of synaptic plasticity

    Front Synaptic Neurosci 2:140.

    • Google Scholar
43. 1. Magee JC
    2. Johnston D

    (1997)

    A synaptically controlled, associative signal for Hebbian plasticity in hippocampal neurons

    Science 275:209–213.

    • Google Scholar
44. 1. Mainen ZF
    2. Sejnowski TJ

    (1995)

    Reliability of spike timing in neocortical neurons

    Science 268:1503–1506.

    • Google Scholar
45. 1. Margrie TW
    2. Meyer AH
    3. Caputi A
    4. Monyer H
    5. Hasan MT
    6. Schaefer AT

    et al. (2003)

    Targeted whole-cell recordings in the mammalian brain in vivo

    Neuron 39:911–918.

    • Google Scholar
46. 1. Markram H
    2. Lubke J
    3. Frotscher M
    4. Sakmann B

    (1997)

    Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs

    Science 275:213–215.

    • Google Scholar
47. 1. Marsalek P
    2. Koch C
    3. Maunsell J

    (1997)

    On the relationship between synaptic input and spike output jitter in individual neurons

    Proc Natl Acad Sci USA 94:735–740.

    • Google Scholar
48. 1. Meliza CD
    2. Dan Y

    (2006)

    Receptive-field modification in rat visual cortex induced by paired visual stimulation and single-cell spiking

    Neuron 49:183–189.

    • Google Scholar
49. 1. Niell CM
    2. Stryker MP

    (2008)

    Highly selective receptive fields in mouse visual cortex

    J Neurosci 28:7520–7536.

    • Google Scholar
50. 1. Nimmerjahn A
    2. Kirchhoff F
    3. Kerr JN
    4. Helmchen F

    (2004)

    Sulforhodamine 101 as a specific marker of astroglia in the neocortex in vivo

    Nat Methods 1:31–37.

    • Google Scholar
51. 1. Okun M
    2. Lampl I

    (2008)

    Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities

    Nat Neurosci 11:535–537.

    • Google Scholar
52. 1. Parra LC
    2. Beck JM
    3. Bell AJ

    (2009)

    On the maximization of information flow between spiking neurons

    Neural Comput 21:2991–3009.

    • Google Scholar
53. 1. Petersen CC
    2. Hahn TT
    3. Mehta M
    4. Grinvald A
    5. Sakmann B

    (2003)

    Interaction of sensory responses with spontaneous depolarization in layer 2/3 barrel cortex

    Proc Natl Acad Sci USA 100:13638–13643.

    • Google Scholar
54. 1. Pfister JP
    2. Toyoizumi T
    3. Barber D
    4. Gerstner W

    (2006)

    Optimal spike-timing-dependent plasticity for precise action potential firing in supervised learning

    Neural Comput 18:1318–1348.

    • Google Scholar
55. 1. Pool RR
    2. Mato G

    (2011)

    Spike-timing-dependent plasticity and reliability optimization: the role of neuron dynamics

    Neural Comput 23:1768–1789.

    • Google Scholar
56. 1. Sachdev RN
    2. Ebner FF
    3. Wilson CJ

    (2004)

    Effect of subthreshold up and down states on the whisker-evoked response in somatosensory cortex

    J Neurophysiol 92:3511–3521.

    • Google Scholar
57. 1. Sawinski J
    2. Wallace DJ
    3. Greenberg DS
    4. Grossmann S
    5. Denk W
    6. Kerr JN

    (2009)

    Visually evoked activity in cortical cells imaged in freely moving animals

    Proc Natl Acad Sci USA 106:19557–19562.

    • Google Scholar
58. 1. Shouval HZ
    2. Wang SS
    3. Wittenberg GM

    (2010)

    Spike timing dependent plasticity: a consequence of more fundamental learning rules

    Front Comput Neurosci 4:1–13.

    • Google Scholar
59. 1. Shu Y
    2. Duque A
    3. Yu Y
    4. Haider B
    5. McCormick DA

    (2007)

    Properties of action-potential initiation in neocortical pyramidal cells: evidence from whole cell axon recordings

    J Neurophysiol 97:746–760.

    • Google Scholar
60. 1. Sjostrom PJ
    2. Turrigiano GG
    3. Nelson SB

    (2001)

    Rate, timing, and cooperativity jointly determine cortical synaptic plasticity

    Neuron 32:1149–1164.

    • Google Scholar
61. 1. Slotnick BM
    2. Brown DL

    (1980)

    Variability in the stereotaxic position of cerebral points in the albino rat

    Brain Res Bull 5:135–139.

    • Google Scholar
62. 1. Song S
    2. Abbott LF

    (2001)

    Cortical development and remapping through spike timing-dependent plasticity

    Neuron 32:339–350.

    • Google Scholar
63. 1. Song S
    2. Miller KD
    3. Abbott LF

    (2000)

    Competitive Hebbian learning through spike-timing-dependent synaptic plasticity

    Nat Neurosci 3:919–926.

    • Google Scholar
64. 1. Tchumatchenko T
    2. Malyshev A
    3. Wolf F
    4. Volgushev M

    (2011)

    Ultrafast population encoding by cortical neurons

    J Neurosci 31:12171–12179.

    • Google Scholar
65. 1. Toyoizumi T
    2. Pfister JP
    3. Aihara K
    4. Gerstner W

    (2007)

    Optimality model of unsupervised spike-timing-dependent plasticity: synaptic memory and weight distribution

    Neural Comput 19:639–671.

    • Google Scholar
66. 1. Varga Z
    2. Jia H
    3. Sakmann B
    4. Konnerth A

    (2011)

    Dendritic coding of multiple sensory inputs in single cortical neurons in vivo

    Proc Natl Acad Sci USA 108:15420–15425.

    • Google Scholar
67. 1. Wallace DJ
    2. Kerr JN

    (2010)

    Chasing the cell assembly

    Curr Opin Neurobiol 20:296–305.

    • Google Scholar
68. 1. Waters J
    2. Helmchen F

    (2006)

    Background synaptic activity is sparse in neocortex

    J Neurosci 26:8267–8277.

    • Google Scholar
69. 1. Wiesel TN
    2. Hubel DH

    (1963)

    Single-cell responses in striate cortex of kittens deprived of vision in one eye

    J Neurophysiol 26:1003–1017.

    • Google Scholar
70. 1. Young JM
    2. Waleszczyk WJ
    3. Wang C
    4. Calford MB
    5. Dreher B
    6. Obermayer K

    (2007)

    Cortical reorganization consistent with spike timing-but not correlation-dependent plasticity

    Nat Neurosci 10:887–895.

    • Google Scholar
71. 1. Zhang JC
    2. Lau PM
    3. Bi GQ

    (2009)

    Gain in sensitivity and loss in temporal contrast of STDP by dopaminergic modulation at hippocampal synapses

    Proc Natl Acad Sci USA 106:13028–13033.

    • Google Scholar

Author details

1. Verena Pawlak

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

   Contribution

   VP, Performed all in vivo experiments and histology, Conception and design, Analysis and interpretation of data

   Competing interests

   The authors have declared that no competing interests exist

2. David S Greenberg

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

   Contribution

   DSG, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors have declared that no competing interests exist

3. Henning Sprekeler

   School of Computer and Communication Sciences and School of Life Sciences, Brain Mind Institute, Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland

   Contribution

   HS, Devised model and performed simulations

   Competing interests

   The authors have declared that no competing interests exist

4. Wulfram Gerstner

   1. School of Computer and Communication Sciences and School of Life Sciences, Brain Mind Institute, Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland
   2. Bernstein Center for Computational Neuroscience, Tübingen, Germany

   Contribution

   WG, Devised model and performed simulations

   Competing interests

   The authors have declared that no competing interests exist

5. Jason ND Kerr

   1. Network Imaging Group, Max Planck Institute for Biological Cybernetics, Tübingen, Germany
   2. Bernstein Center for Computational Neuroscience, Tübingen, Germany

   Contribution

   JNDK, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   jason@tuebingen.mpg.de

   Competing interests

   The authors have declared that no competing interests exist

• 3,014

  Page views

• 437

  Downloads

• 21

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.