Information

Can we disambiguate competing molecular mechanisms of learning and plasticity by measuring electrical activity of neurons?

Can we disambiguate competing molecular mechanisms of learning and plasticity by measuring electrical activity of neurons?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I have been reading with fascination about the several molecular- and cellular-scale mechanisms and structural changes that underlie what we refer to as long-term plasticity. For instance, [1], [2], [3].

In particular, there seem to be at least: synaptic scaling, synaptic pruning, synaptogenesis, spinogenesis, neurogenesis.

My questions:

  1. How do each of the above mechanisms/ events affect the classical extracellular single/ multi-unit spikes, postsynaptic potentials (LFPs), and local electric fields? Does a mechanistic forward model exist, which specifies the biophysical transformation from synapse-level changes to spikes and LFPs? Are there spike simulators out there that model these properties?

  2. Using such a forward model, can we solve a statistical inverse problem to infer anything at all about these LTP/LTD-induced small-scale structural and functional changes?

  3. For instance, one metric of analysis could be: post-spike-time filters (self-terms and cross-terms) inferred using statistical point-process models. Another could be the shape of the action potential itself, which is typically used only for spike sorting, and then discarded.

  4. Has something along these lines been done? Or is it infeasible? If so, why?

  5. Finally, If anyone could give a brief overview of the gamut of synapse-scale physiological/ imaging techniques used to characterize these mechanisms, I would be very grateful!


Just adding a little bit here. Estimating changes in connectivity based on STDP is hard http://klab.smpp.northwestern.edu/wiki/images/2/2b/Stevenson_Inferring_Plasticity_2011.pdf

but yes - these questions are enough to keep a big field busy for a long time.


Thanks!

I was thinking along the following lines.

(1) Write down the cable equations for neurons within a network, given: anatomical layout of the network, spatial distribution of dendrites, spines, synapses, and synaptic strengths (forward model).

(2) Acquire joint microscopy + electrophysiology dataset.

(3) Invert the forward model with synaptic strengths as free parameters.

(4) Track these inferred parameters over time or compare them between conditions.


Can we disambiguate competing molecular mechanisms of learning and plasticity by measuring electrical activity of neurons? - Biology

Despite the ubiquity of sleep across phylogeny, its function remains elusive. In this review, we consider one compelling candidate: brain plasticity associated with memory processing. Focusing largely on hippocampus-dependent memory in rodents and humans, we describe molecular, cellular, network, whole-brain and behavioral evidence establishing a role for sleep both in preparation for initial memory encoding, and in the subsequent offline consolidation of memory. Sleep and sleep deprivation bidirectionally alter molecular signaling pathways that regulate synaptic strength and control plasticity-related gene transcription and protein translation. At the cellular level, sleep deprivation impairs cellular excitability necessary for inducing synaptic potentiation and accelerates the decay of long-lasting forms of synaptic plasticity. In contrast, rapid eye movement (REM) and non-rapid eye movement (NREM) sleep enhance previously induced synaptic potentiation, although synaptic de-potentiation during sleep has also been observed. Beyond single cell dynamics, large-scale cell ensembles express coordinated replay of prior learning-related firing patterns during subsequent NREM sleep. At the whole-brain level, somewhat analogous learning-associated hippocampal (re)activation during NREM sleep has been reported in humans. Moreover, the same cortical NREM oscillations associated with replay in rodents also promote human hippocampal memory consolidation, and this process can be manipulated using exogenous reactivation cues during sleep. Mirroring molecular findings in rodents, specific NREM sleep oscillations before encoding refresh human hippocampal learning capacity, while deprivation of sleep conversely impairs subsequent hippocampal activity and associated encoding. Together, these cross-descriptive level findings demonstrate that the unique neurobiology of sleep exerts powerful effects on molecular, cellular and network mechanisms of plasticity that govern both initial learning and subsequent long-term memory consolidation.


Findings

We have used dissociated neuronal cultures grown over MEA for 2𠄶 weeks to monitor the electrical activity from a population of neurons [9]. MEAs allow stable and long lasting recordings (hours to days) of extracellular signals from the entire population and permit to characterize and follow the properties of single spikes from identified neurons. In this way, it was possible to describe the global properties of the network, such as its overall electrical activity and to obtain a characterization of changes during neuronal plasticity of single identified spikes. This analysis could not be performed with hippocampal slices or organotypic cultures grown on MEAs or in vivo, because in these cases local field potentials (LFPs) are observed and a detailed investigation of neuronal plasticity at a single spike level is almost impossible. We increased synaptic efficacy and the overall electrical activity by treating hippocampal cultures for 30 min. with the GABAA receptor antagonist gabazine (GabT). After GabT, gabazine was washed out and the time course of evoked electrical activity was followed/studied. MEA's extracellular electrodes were used for recording and stimulation so to quantify changes of the evoked activity. Brief (200 μs) bipolar pulses were applied to a row of electrodes (black bar in the grid of Fig. ​ Fig.1A) 1A ) and the propagation of evoked spikes throughout the network was recorded. In order to avoid saturation, the lowest voltage pulse evoking at least one spike was used, before, during and after GabT (Fig. ​ (Fig.1). 1 ). Its amplitude varied between 200 to 450 mV depending on the culture. After GabT, the number of evoked spikes increased at all times in almost all trials at the level of the single electrode (Fig. ​ (Fig.1B) 1B ) and in the entire network (Fig. ​ (Fig.1C 1C ).

Potentiation of the evoked response induced by GabT. A, Examples of the evoked activity's time course for a single trial for three representative electrodes located at 0.5, 1 and 2 mm distance (see inset) from the stimulating electrode respectively (black bar in inset). The number of evoked spikes increases following the GabT. Hippocampal cultures were stimulated with the lowest intensity (200� mV) evoking a response. 40 pulses (trials) with an inter-pulse interval 4s were used. Inset, graphical representation of the MEA grid. Each square represents an electrode. Distance between electrodes is 500 μm. Black bar corresponds to the stimulated electrodes and grey numbered squares to the electrodes whose activity is shown in (A). B, Raster plots of the spikes evoked in one electrode analyzed at each time point. Each horizontal line represents the response to one trial recorded up to 250 ms after the stimulus onset. C, Time course of the NFR. The total number of evoked spikes was counted in 10 ms bins, following electrical stimulation. The increase of evoked spikes is especially pronounced in the first 100 ms following stimulation. Data from a single experiment shown in (A-C). D, Average time course of the NFR (n = 5 cultures). The NFR was normalized relative to control values. Bin size was 250 ms. E, Ten overlapping spike traces in control conditions (black traces) and 3 h after gabazine washout (red traces) of an individual electrode showing the decrease of the latency of the first evoked spike. Artifacts have been truncated for clarity. F, Time course of the latency (blue traces) and jitter (red traces) of the first spike for three neurons showing that latency and jitter decrease after GabT. Each symbol corresponds to a different neuron. G, Time course of propagation constant (λ see Methods) showing the increase of the activity spread following GabT. Inset, the number of spikes in a 100 ms time window was counted and averaged, for electrodes located at 0.5, 1 and 2 mm from the stimulated bar of electrodes, and fitted with an exponential function Ae -d/λ (grey line). Colours as in (E).

Changes of the evoked response were quantified by computing the total number of evoked spikes in a time window of 100 ms from the stimulus onset, referred to as the network firing rate of the evoked response NFR. NFR significantly increased between 1 and 6 h after GabT (Fig. 1C, D ): the evoked response was maximally potentiated 3 h after GabT and started to decline after 6 h, returning close to the control level 24 h after GabT.

Neuronal plasticity induced by GabT not only modified synaptic efficacy but also several other network properties such as speed and reliability of the evoked spikes. Speed was evaluated by measuring the latency of the evoked response i.e. the delay from the stimulus of the evoked spike, while reliability was measured by the standard deviation of the latency (jitter). In some experiments it was possible to identify spikes produced by the same neuron (Fig. ​ (Fig.1E) 1E ) and therefore we could measure how its latency and jitter changed during neuronal plasticity. The latency in control conditions from stimulus onset varied between 6 and 9 ms, was reduced by 2𠄳 ms after GabT and its jitter similarly decreased (Fig. ​ (Fig.1F). 1F ). We also analyzed how spikes propagated in the network by measuring the space constant λ (total number of evoked spikes as a function of the distance from the stimulus) of the evoked activity. Collected data from 4 cultures showed that λ increased by about 25 % within 1 h after GabT and remained larger than the control values up to 24 h (Fig. ​ (Fig.1G). 1G ). These results show that increase of synaptic efficacy by exposure to gabazine alone in the absence of a concomitant strong or tetanic electrical stimulation, potentiated the electrical response propagating in the culture, inducing in this way a form of LTP, which we refer to as medium time LTP (M-LTP) because it was not identified as maximal after gabazine removal, developed 1 hour after GabT and lasted about 6 hours. Therefore, M-LTP is likely to be associated not only to local protein trafficking but also to changes of gene expression, occurring on a time scale of some hours. In order to understand the molecular events underlying M-LTP, we have analysed changes of gene expression induced by the same pharmacological treatment, i.e. a 30 min. exposure to gabazine with Affymetrix microarrays (RAT 230_2.0 Gene Chip) (Broccard et al., manuscript in preparation). We have identified 342 genes significantly up-regulated at the same times as when the evoked electrical activity was potentiated. Nevertheless, because gene profiles were obtained from the whole culture, we could not identify the type of neurons where up-regulated genes were expressed. Many of these genes are well known players in LTP such as Bdnf and its receptor TrkB [11], Arc [12], Egr1 [13] and Homer1 [14]. We hypothesized that the large majority of identified genes underlies induction and maintenance of LTP and that their activation orchestrates neuronal plasticity. In fact, a search in the PuBMed database indicates that 40% of the 284 annotated genes is, or could be, involved in changes of synaptic strength related to LTP. 43 genes have already been implicated in LTP, 25 genes have been classified as Structural genes for their structural role in cellular function and their up-regulation could underlie structural and morphological changes associated to LTP. Analogously, the 25 Pre-synaptic and the 24 Post-synaptic genes found in our screening could mediate changes of synaptic properties occurring during LTP.

ERK1/2 signalling plays an important role in several plasticity-related processes in the nervous system [15]. Therefore, we investigated the effect of the inhibition of ERK1/2 pathway by PD98059 and U0126 on the potentiation of the evoked response (Fig. ​ (Fig.2). 2 ). Application of these inhibitors to neuronal cultures decreased the spontaneous activity measured for all extracellular electrodes (Fig. ​ (Fig.2A). 2A ). The network firing rate was almost halved (Fig. ​ (Fig.2B) 2B ) in all tested cultures (n = 4), but periods of larger electrical activity could still be observed. In these cultures, inhibitors of the ERK1/2 were incubated for 45 minutes before GabT and changes of the evoked response were analyzed. In the presence of these inhibitors, GabT still potentiated the evoked response, although to a lesser extent (red trace in Fig. ​ Fig.2C), 2C ), with a time course similar to that observed in the absence of these inhibitors (black trace in Fig. ​ Fig.2C). 2C ). 3 h after GabT, the number of evoked spikes reached 198 ± 41 % in normal conditions, but increased only by 39 ± 15 % in the presence of PD98059 and U0126. Therefore, inhibition of the ERK1/2 pathway reduced but not abolished the potentiation of the evoked response caused by GabT.

Effect of inhibitors of the ERK1/2 pathway. A, Representative traces from two individual extracellular electrodes in control conditions (left) and in the presence of 50 μM PD98059 (right). B, Corresponding network firing rate computed with a bin width of 25 ms, in normal conditions (left) and in the presence of 50 μM PD98059 (right) showing a depression of the spontaneous electrical activity by PD98059. Dotted line indicates zero spike. C, The evoked activity is still potentiated when the ERK1/2 pathway was blocked before GabT (see Results). Dashed line corresponds to potentiation of the evoked activity without blocking the ERK1/2 pathway before GabT. Data for PD98059 (50 μM n = 3) and U0126 (20 μM n = 3) were pooled together as they affected network properties in a similar way. D, Changes of gene expression occurring 1.5 h after GabT measured by Real-Time PCR for Egr1, Egr2, Egr3, Nr4a1, Bdnf, Homer1a and Arc in the presence of gabazine and PD98059 (gray bars) relative to normalized expression in the presence of gabazine alone (black bar).

We analyzed with real-time PCR the effect of the ERK1/2 inhibitors on some of the LTP-related genes, up-regulated in our microarray screening: Egr1 [13],Egr2 [16],Egr3 [17],Nr4a1 [18],Bdnf [11],Homer1a [14] and Arc [12]. As shown in Figure ​ Figure2D, 2D , the up-regulation induced by GabT of genes of the EGR family, Nr4a1 and Arc was significantly reduced and almost blocked by inhibitors of the ERK1/2 pathway, but not the up-regulation of Bdnf and Homer1a.

The results described in the present investigation show that when following GabT a potentiation of the evoked electrical activity occurs at medium times (M-LTP). This form of chemically induced LTP is expected to modify the great majority of synapses present in the network and therefore to affect its global properties. When LTP is induced by a local electrical stimulation, only a limited number of synapses are expected to be modified.

As shown in Fig. ​ Fig.2, 2 , potentiation of the M-LTP was reduced, but not eliminated by inhibitors of the ERK1/2 pathway, in agreement with the notion that neuronal plasticity is mediated by several distinct pathways likely to be working in unison. These results allowed us to relate changes of electrical properties occurring during neuronal plasticity to specific underlying molecular events.

The present analysis combining MEA and DNA microarrays represents a simple system to study neuronal plasticity [4], but does not allow to identify the cellular origin of detected changes of gene expression. Given the large abundance of pyramidal neurons in hippocampal cultures and acute slices it is likely that detected changes of gene expression occur in these neurons, but it is possible that they occur also in interneurons and in glial cells. In order to resolve this issue it will be necessary to perform single cell gene profiling in the intact hippocampus. Preliminary experiments performed in our laboratory in intact organotypic slices show that treatment with gabazine induces very similar changes of gene expression in dissociated cultures, as those here used and in neuronal slices preserving the original physiological connectivity.


References

Marder, E. & Thirumalai, V. Cellular, synaptic and network effects of neuromodulation. Neural Netw. 15, 479–493 (2002).

Abbott, L. F. & Nelson, S. B. Synaptic plasticity: taming the beast. Nature Neurosci. 3 (suppl.), 1178–1183. (2000).

Sjostrom, P. J. & Nelson, S. B. Spike timing, calcium signals and synaptic plasticity. Curr. Opin. Neurobiol. 12, 305–314 (2002).

Kandel, E. R. The molecular biology of memory storage: a dialogue between genes and synapses. Science 294, 1030–1038 (2001).

Martin, K. C. & Kosik, K. S. Synaptic tagging — who's it? Nature Rev. Neurosci. 3, 813–820 (2002).

Dayan, P. & Abbott, L. F. in Theoretical Neuroscience (MIT, Cambridge, 2001).

Marder, E., Abbott, L. F., Turrigiano, G. G., Liu, Z. & Golowasch, J. Memory from the dynamics of intrinsic membrane currents. Proc. Natl Acad. Sci. USA 93, 13481–13486 (1996).

Zhang, W. & Linden, D. J. The other side of the engram: experience-driven changes in neuronal intrinsic excitability. Nature Rev. Neurosci. 4, 885–900 (2003).

Daoudal, G. & Debanne, D. Long-term plasticity of intrinsic excitability: learning rules and mechanisms. Learn. Mem. 10, 456–465 (2003).

Prinz, A. A., Abbott, L. F. & Marder, E. The dynamic clamp comes of age. Trends Neurosci. 27, 218–224 (2004).

Sharp, A. A., Skinner, F. K. & Marder, E. Mechanisms of oscillation in dynamic clamp constructed two-cell half-center circuits. J. Neurophysiol. 76, 867–883 (1996).

Cajal, R. S. Histologie du Système Nerveux de l'Homme et des Vertébrés (Maloine, Paris, 1909).

Rall, W. Distinguishing theoretical synaptic potentials computed for different soma-dendritic distributions of synaptic input. J. Neurophysiol. 30, 1138–1168 (1967).

Rall, W. Time constants and electrotonic length of membrane cylinders and neurons. Biophys. J. 9, 1483–1508 (1969).

Rall, W. & Rinzel, J. Branch input resistance and steady attenuation for input to one branch of a dendritic neuron model. Biophys. J. 13, 648–687 (1973).

Johnston, D., Magee, J. C., Colbert, C. M. & Cristie, B. R. Active properties of neuronal dendrites. Annu. Rev. Neurosci. 19, 165–186 (1996).

Migliore, M. & Shepherd, G. M. Emerging rules for the distributions of active dendritic conductances. Nature Rev. Neurosci. 3, 362–370 (2002).

Yuste, R. & Tank, D. W. Dendritic integration in mammalian neurons, a century after Cajal. Neuron 16, 701–716 (1996).

Stuart, G., Spruston, N. & Hausser, M. Dendrites (MIT, Cambridge, Massachusetts, 2000).

Schwindt, P. C. & Crill, W. E. Amplification of synaptic current by persistent sodium conductance in apical dendrite of neocortical neurons. J. Neurophysiol. 74, 2220–2224 (1995).

Magee, J. C. Dendritic Ih normalizes temporal summation in hippocampal CA1 neurons. Nature Neurosci. 2, 508–514 (1999).

Williams, S. R. & Stuart, G. J. Site independence of EPSP time course is mediated by dendritic I(h) in neocortical pyramidal neurons. J. Neurophysiol. 83, 3177–3182 (2000).

Magee, J. C. & Johnston, D. A synaptically controlled, associative signal for Hebbian plasticity in hippocampal neurons. Science 275, 209–213 (1997).

Markram, H., Lübke, J., Frotscher, M. & Sakmann, B. Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs. Science 275, 213–215 (1997).

Hines, M. L. & Carnevale, N. T. The NEURON simulation environment. Neural Comput. 9, 1179–1209 (1997).

Koch, C. & Segev, I. Methods in Neuronal Modeling (MIT, Cambridge, 1998).

Bower, J. & Beeman, D. The Book of GENESIS (Springer, Berlin, 1994).

Golding, N. L., Staff, N. P. & Spruston, N. Dendritic spikes as a mechanism for cooperative long-term potentiation. Nature 418, 326–331 (2002).

Turrigiano, G. G. & Nelson, S. B. Homeostatic plasticity in the developing nervous system. Nature Rev. Neurosci. 5, 97–107 (2004).

Marder, E. & Prinz, A. A. Modeling stability in neuron and network function: the role of activity in homeostasis. Bioessays 24, 1145–1154 (2002).

Turrigiano, G. G., Leslie, K. R., Desai, N. S., Rutherford, L. C. & Nelson, S. B. Activity-dependent scaling of quantal amplitude in neocortical neurons. Nature 391, 892–896 (1998).

Desai, N. S., Rutherford, L. C. & Turrigiano, G. G. Plasticity in the intrinsic excitability of cortical pyramidal neurons. Nature Neurosci. 2, 515–520 (1999).

Goldman, M. S., Golowasch, J., Marder, E. & Abbott, L. F. Global structure, robustness, and modulation of neuronal models. J. Neurosci. 21, 5229–5238 (2001).

MacLean, J. N., Zhang, Y., Johnson, B. R. & Harris-Warrick, R. M. Activity-independent homeostasis in rhythmically active neurons. Neuron 37, 109–120 (2003).

Golowasch, J., Abbott, L. F. & Marder, E. Activity-dependent regulation of potassium currents in an identified neuron of the stomatogastric ganglion of the crab Cancer borealis. J. Neurosci. 19, RC33 (1999).

Selverston, A. I. Are central pattern generators understandable? Behav. Brain Sci. 3, 535–571 (1980).

Getting, P. A. Emerging principles governing the operation of neural networks. Annu. Rev. Neurosci. 12, 185–204 (1989).

Friesen, W. O. Reciprocal inhibition: a mechanism underlying oscillatory animal movements. Neurosci. Biobehav. Rev. 18, 547–553 (1994).

Wang, X. -J. & Rinzel, J. Alternating and synchronous rhythms in reciprocally inhibitory model neurons. Neural Comput. 4, 84–97 (1992).

Van Vreeswijk, C., Abbott, L. F. & Ermentrout, G. B. When inhibition not excitation synchronizes neural firing. J. Comput. Neurosci. 1, 313–321 (1994).

White, J. A., Chow, C. C., Ritt, J., Soto-Trevino, C. & Kopell, N. Synchronization and oscillatory dynamics in heterogeneous, mutually inhibited neurons. J. Comput. Neurosci. 5, 5–16 (1998).

Marder, E. & Calabrese, R. L. Principles of rhythmic motor pattern generation. Physiol. Rev. 76, 687–717 (1996).

Nusbaum, M. P. & Beenhakker, M. P. A small-systems approach to motor pattern generation. Nature 417, 343–350 (2002).

Harris-Warrick, R. M. et al. Distributed effects of dopamine modulation in the crustacean pyloric network. Ann. N Y Acad. Sci. 860, 155–167 (1998).

Katz, P. S. & Frost, W. N. Intrinsic neuromodulation: altering neuronal circuits from within. Trends Neurosci. 19, 54–61 (1996).

Steriade, M., Jones, E. G. & McCormick, D. A. Thalamus (Elsevier, Amsterdam, 1997).

Destexhe, A. & Sejnowski, T. J. Interactions between membrane conductances underlying thalamocortical slow-wave oscillations. Physiol. Rev. 83, 1401–1453 (2003).

McCormick, D. A. Cholinergic and noradrenergic modulation of thalamocortical processing. Trends Neurosci. 12, 215–221 (1989).

Mountcastle, V. B. in The Neurosciences: Fourth Study Program (eds Schmidt, F. O. & Worden, F. G.) 21–42 (MIT Press, Cambridge, 1979).

Hubel, D. H. & Wiesel, T. N. Shape and arrangement of columns in cat's striate cortex. J. Physiol. 165, 559–568 (1963).

Douglas, R. J. & Martin, K. A. A functional microcircuit for cat visual cortex. J. Physiol. 440, 735–769 (1991).

Szentagothai, J. The modular architectonic principle of neural centers. Rev. Physiol. Biochem. Pharmacol. 98, 11–61 (1983).

Barlow, H. in Models of the Visual Cortex (eds Rose, D. & Dobson, V.) 37–46 (Wiley, Chichester, 1985).

Braitenberg, V. & Schuz, A. Cortex: statistics and geometry of neuronal connectivity (Springer, Berlin, 1998).

Destexhe, A., Rudolph, M. & Pare, D. The high-conductance state of neocortical neurons in vivo. Nature Rev. Neurosci. 4, 739–751 (2003).

Ho, N. & Destexhe, A. Synaptic background activity enhances the responsiveness of neocortical pyramidal neurons. J. Neurophysiol. 84, 1488–1496 (2000).

Wiesenfeld, K. & Moss, F. Stochastic resonance and the benefits of noise: from ice ages to crayfish and SQUIDs. Nature 373, 33–36 (1995).

Chance, F. S., Abbott, L. F. & Reyes, A. D. Gain modulation from background synaptic input. Neuron 35, 773–782 (2002).

Shu, Y., Hasenstaub, A., Badoual, M., Bal, T. & McCormick, D. A. Barrages of synaptic activity control the gain and sensitivity of cortical neurons. J. Neurosci. 23, 10388–10401 (2003).

Rudolph, M. & Destexhe, A. A fast-conducting, stochastic integrative mode for neocortical neurons in vivo. J. Neurosci. 23, 2466–2476 (2003).

Wei, D. S. et al. Compartmentalized and binary behavior of terminal dendrites in hippocampal pyramidal neurons. Science 293, 2272–2275 (2001).

Shepherd, G. M. & Brayton, R. K. Logic operations are properties of computer-simulated interactions between excitable dendritic spines. Neuroscience 21, 151–165 (1987).

Mel, B. W. Information processing in dendritic trees. Neural Comput. 6, 1031–1085 (1994).

Softky, W. Sub-millisecond coincidence detection in active dendritic trees. Neuroscience 58, 13–41 (1994).

Shalden, M. N. & Newsome, W. T. The variable discharge of cortical neurons: implications for connectivity, computation, and information coding. J. Neurosci. 18, 3870–3896 (1998).

Moore, G. P., Perkel, D. H. & Segundo, J. P. Statistical analysis and functional interpretation of neuronal spike data. Annu. Rev. Physiol. 28, 493–522 (1966).

Rao, R., Olshausen, B. & Lewicki, M. Probabilistic Models of the Brain (MIT, Cambridge, 2002).

Hopfield, J. J. Neural networks and physical systems with emergent collective computational abilities. Proc. Natl Acad. Sci. USA 79, 2554–2558 (1982).

Roskies, A. The binding problem: special issue. Neuron 24, 7–125 (1999).

von der Malsburg, C. & Schneider, W. A neural cocktail-party processor. Biol. Cybern. 54, 29–40 (1986).

Engel, A. K., Fries, P. & Singer, W. Dynamic predictions: oscillations and synchrony in top-down processing. Nature Rev. Neurosci. 2, 704–716 (2001).

Abeles, M. Corticonics: Neuronal Circuits of the Cerebral Cortex (Cambridge University Press, Cambridge, 1991).

Rudolph, M. & Destexhe, A. Correlation detection and resonance in neural systems with distributed noise sources. Phys. Rev. Lett. 86, 3662–3665 (2001).

Barlow, H. B. in Sensory Communications (ed. Rosenblith, W.) Ch. 13, 217–234 (MIT, Cambridge, 1961).

Barlow, H. & Foldiak, P. in The Computing Neuron Ch. 4 (eds Durbin, R., Miall, C. & G, M.) 54–72 (Addison-Wesley, New York, 1989).

Srinivasan, M. V., Laughlin, S. B. & Dubs, A. Predictive coding: a fresh view of inhibition in the retina. Proc. R. Soc. Lond. B 216, 427–459 (1982).

Ito, M. Cerebellar long-term depression: characterization, signal transduction, and functional roles. Physiol. Rev. 81, 1143–1195 (2001).

Bell, C. C., Han, V. Z., Sugawara, Y. & Grant, K. Synaptic plasticity in a cerebellum-like structure depends on temporal order. Nature 387, 278–281 (1997).

Foldiak, P. Forming sparse representations by local anti-Hebbian learning. Biol. Cybern. 64, 165–170 (1990).

Wang, X. J., Liu, Y., Sanchez-Vives, M. V. & McCormick, D. A. Adaptation and temporal decorrelation by single neurons in the primary visual cortex. J. Neurophysiol. 89, 3279–3293 (2003).

Goldman, M. S., Maldonado, P. & Abbott, L. F. Redundancy reduction and sustained firing with stochastic depressing synapses. J. Neurosci. 22, 584–591 (2002).

Peters, A. & Yilmaz, E. Neuronal organization in area 17 of cat visual cortex. Cereb. Cort. 3, 49–68 (1993).

Olshausen, B. A. & Field, D. J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images. Nature 381, 607–609 (1996).

Perez-Orive, J. et al. Oscillations and sparsening of odor representations in the mushroom body. Science 297, 359–365 (2002).

Hahnloser, R. H., Kozhevnikov, A. A. & Fee, M. S. An ultra-sparse code underlies the generation of neural sequences in a songbird. Nature 419, 65–70 (2002).

Baranyi, A. & Feher, O. Conditioned changes of synaptic transmission in the motor cortex of the cat. Exp. Brain Res. 33, 283–298 (1978).

Kirkwood, A. & Bear, M. F. Hebbian synapses in visual cortex. J. Neurosci. 14, 1634–1645 (1994).

Thomson, A. M. Facilitation, augmentation and potentiation at central synapses. Trends Neurosci. 23, 305–312 (2000).

Hebb, D. O. The Organization of Behavior (Wiley, New York, 1949).

Gupta, A., Wang, Y. & Markram, H. Organizing principles for a diversity of GABAergic interneurons and synapses in the neocortex. Science 287, 273–278 (2000).

Silberberg, G., Gupta, A. & Markram, H. Stereotypy in neocortical microcircuits. Trends Neurosci. 25, 227–230 (2002).

Buonomano, D. V. & Merzenich, M. M. Temporal information transformed into a spatial code by a neural network with realistic properties. Science 267, 1028–1030 (1995).

Maass, W., Natschlager, T. & Markram, H. Real-time computing without stable states: a new framework for neural computation based on perturbations. Neural Comput. 14, 2531–2560 (2002).

Monier, C., Chavane, F., Baudot, P., Graham, L. J. & Fregnac, Y. Orientation and direction selectivity of synaptic inputs in visual cortical neurons: a diversity of combinations produces spike tuning. Neuron 37, 663–680 (2003).

Steriade, M. & McCarley, R. W. Brainstem Control of Wakefulness and Sleep (Plenum, New York, 1990).

Bear, M. F. & Singer, W. Modulation of visual cortical plasticity by acetylcholine and noradrenaline. Nature 320, 172–176 (1986).

Shulz, D. E., Sosnik, R., Ego, V., Haidarliu, S. & Ahissar, E. A neuronal analogue of state-dependent learning. Nature 403, 549–553 (2000).

Stickgold, R., Hobson, J. A., Fosse, R. & Fosse, M. Sleep, learning, and dreams: off-line memory reprocessing. Science 294, 1052–1057 (2001).

Hinton, G. E., Dayan, P., Frey, B. J. & Neal, R. M. The ‘wake-sleep’ algorithm for unsupervised neural networks. Science 268, 1158–1161 (1995).

Frégnac, Y. in Handbook of Brain Theory and Neural Networks (ed. Arbib, M. A.) 515–522 (MIT, Cambridge, 2002).


Discussion

By performing a comprehensive electrophysiological study, computational modeling and molecular analysis, we found that hyperpolarization-activated cyclic nucleotide-gated (HCN) and T-type calcium (CaV3) channels play important role in the homeostatic plasticity of primary dissociated neurons as well as hippocampal organotypic slice cultures induced by chronic activity-deprivation.

Previous investigations showed that these channels regulate synaptic integration and intrinsic excitability under normal conditions and contribute to dysfunctional network activity in pathological conditions, especially in epilepsy 26 – 29 . In the present work we provide evidence that h- and T-currents are upregulated during homeostatic adaptations following 48 h of TTX treatment. Our data show that all subunits of the HCN and T-type calcium channels are present in primary dissociated hippocampal cultures. Blocking spike-mediated transmission increased CaV3.1 mRNA expression, but we did not detect a change in the quantity or plasma membrane localization of the subunit protein. It should be noted that changes in the mRNA levels cannot be interpreted as direct correlates of functional protein expression 30 . Nevertheless, a feasible explanation is that spatial redistribution and/or post-translational modification of T-type Ca channels were induced in a homeostatic manner, leading to increased T-type channel conductances and the establishment of a larger rebound potential, a characteristic physiological signature of T-type Ca-currents.

The action of HCN channels on intrinsic excitability is somewhat bidirectional: on the one hand, HCN channels reduce excitatory postsynaptic potential (EPSP) summation through shunting of the dendritic postsynaptic currents 31 , 32 . On the other hand, they contribute to the depolarization of the resting membrane (RMP) potential 10 . Additionally, Dyhrfjeld-Johnsen et al. 33 showed that during fever-induced seizures, upregulated Ih evokes dendritic hyperexcitability, which can promote epilepsy. In agreement, activity-deprivation in our cultured neurons induced depolarization of voltage sag positive neurons. Interestingly, we failed to detect TTX-induced changes in HCN mRNA or protein levels in culture lysates, despite of the higher number of cells possessing voltage sag. Notably, we also detected a significant increase in TRIP8b (1a-4) isoform levels, which regulates surface expression of HCN channels in the hippocampus 19 , 20 . It is known that HCN channel localization primarily determines the extent of voltage sag responses measured in the soma. In agreement, we detected a more proximal HCN1 channel redistribution in CA1 neurons within organotypic hippocampal slices following 48 h of activity-deprivation. Thus, our results show a specific redistribution of HCN channels that is considered as a very effective way of homeostatic upregulation of intrinsic excitability.

It is known that network activity gradually develops in dissociated hippocampal cultures and from DIV12, neurons exhibit robust firing and mature integrative properties 34 . Our data show that two days of TTX treatment started between DIV12 to 14 increased burst activity. This is in agreement with earlier observations on treating immature cultures with TTX for 9ꃚys, starting from DIV5 until DIV14 35 . Additionally, our pharmacological experiments showed that HCN and T-type Ca channels participate in the formation of global network bursting activity amplified by TTX-evoked homeostatic regulation. In accordance with previous reports showing that these channels have fundamental role in bursting 36 – 39 , we found that selective blocking of these channels decreased burst frequency. At the same time, we found no changes in other burst parameters or the mean firing rate. It is known that T-type voltage-gated channels and HCN channels work together and regulate the firing frequency during post-inhibitory rebound, first spike latency and spike precision in the deep cerebellar nuclear neurons 40 . In addition, HCN1 channels were reported to create signaling complexes with certain CaV3 subunits 41 . Nevertheless, we cannot rule out the possibility that sodium channels are also affected by activity-deprivation as this has been verified by earlier reports using similar model systems 11 . It is likely that differential regulation of sodium, HCN and T-type calcium channels all contribute to the formation of the bursting phenotype during homeostatic plasticity.

Homeostatic regulation of intrinsic excitability has been described as an efficient way to prevent neurons from becoming hyper- or hypoactive in case of sustained loss of neuronal inputs or highly elevated neuronal activity, respectively. Initial research on activity-deprived neurons identified several voltage-dependent membrane conductances, including mostly potassium currents 42 , which are associated with spike generation and are subjects of homeostatic regulation. Indeed, up- or downregulation of such conductances can have a profound effect on the excitability of neurons often manifesting as changes in the rheobase or input–output gain parameters. Our previous experimental data on extended amygdala neurons 43 , as well as present computational models, have shown that homeostatic regulation of other voltage-gated currents including T-type and HCN channels can dramatically alter the neuron’s integrative properties while standard parameters of their static excitability appear to be less influenced. Considering these, even those intrinsic changes that introduce minor shifts in the static input–output functions of neurons (e.g. LTP of intrinsic excitability) can have a robust functional impact when neurons operate under fluctuating synaptic inputs. Our present model simulations also support this idea by showing that upregulation of both the h- and T-currents boosts firing under the action of excitatory synaptic inputs more than firing under standard current step stimulation.

In summary, our data show that in hippocampal neurons, HCN and T-type voltage-gated calcium channels participate in the activity-dependent homeostatic regulation of their integrative properties facilitating synchronization of network activity and bursting.


Abstract

Addiction is caused, in part, by powerful and long-lasting memories of the drug experience. Relapse caused by exposure to cues associated with the drug experience is a major clinical problem that contributes to the persistence of addiction. Here we present the accumulated evidence that drugs of abuse can hijack synaptic plasticity mechanisms in key brain circuits, most importantly in the mesolimbic dopamine system, which is central to reward processing in the brain. Reversing or preventing these drug-induced synaptic modifications may prove beneficial in the treatment of one of society's most intractable health problems.


Supporting Information

Figure S1

Morphology of adult Panx1 +/+ and Panx1 −/− mouse brains. Photographs of fixed 9 month-old male control (Panx1 +/+ , left panels) and Panx1 −/− (right panels) brains showing no macroscopic differences in dorsal (A, B) and ventral views (C, D). Scale Bars A𠄽 =𠂥 mm.

Figure S2

Comparison of frontal brain sections Panx1 +/+ and Panx1 −/− mouse brains. (A, B) Overview representing slices (1.5 mm) at the level of the dorsal hippocampus reveals no obvious regional differences between cortex (Cx), hippocampus, thalamus, (Th), hypothalamus (Hy) and amygdala (Amy) between Panx1 +/+ (left panels) and Panx1 −/− mice (right panels). All animals were 9 months old. (C–H) Toluidine-blue stained semithin sections (0.8 µm) of the hippocampus showing regions CA1 (C, D), CA3 (E, F) and dentate gyrus (DG in G, H). All regions display normal cellular and dendritic composition in both genotypes. Cp, cerebral peduncle GR, granule cell layer, PO, polymorph layer PY, pyramidal cell layer RAD, stratum radiatum SLU, stratum lucidum Scale bars in A, B =𠂠.5 mm bar in C–H =� µm.

Figure S3

Calbindin and Parvalbumin immunohistochemistry of hippocampus in Panx1 +/+ and Panx1 −/− mice. (A, B) Frontal overview vibratome sections, 50 µm thick, show the characteristic calbindin staining pattern of the hippocampal subregions CA1, stratum lucidum (SLU) of CA3 with the positive mossy fibers and the strongly stained dentate gyrus (DG) in Panx1 +/+ (left panels) and Panx1 −/− mice (right panels). Enlargements of the CA1 area exhibit no difference in staining of the pyramidal cell layer (PY), stratum oriens (OR) and stratum radiatum (RAD) between both genotypes. Overview micrograph of Parvalbumin immunostaining displays similar staining of Panx1 +/+ (E) and Panx1 −/− mice (F). Enlargements of CA1 (G, H) show immunpositive somata of interneurons in the pyramidal cell layer and stratum oriens with the dendrites spanning all layers. Bar in A, B, E, F =� µm bar in B, D, G, H =� µm.

Figure S4

Expression of selected glutamate receptor family genes. Real Time PCR was performed to analyze relative expression changes of grm1 (mGlu family I), grm2 (mGlu family II), grm4 (mGlu family III), grin1 (AMPA receptor family) and gria1 (NMDA receptor family). HSP90 and 18 sRNA expression was used for normalization.

Table S1

Summary of the mean ± SEM values for Panx1 +/+ and Panx1 −/− derived early phase LTP and late phase LTP. Data are listed according to the bath applied pharmacological treatment. Note, that wash in of pharmacology was performed at least 10 min in advance to measurements and was kept upright during the whole phase of LTP recordings. P-Values reveal significances for Holm-Sidack post hoc comparisons, which were performed following one-way ANOVA analyses. ns = non-significant

Table S2

Summary of RNA expression profiling data. Summary of data analysis using PCRArrayDataAnalysis_V3.3 software, version August 2010, (http://www.sabiosciences.com/pcrarraydataanalysis.php). Metabotropic glutamate receptor 4 (GRM4) is highlighted in red. Samples in wells H01–H05 were used for normalization of the data set.

Methods S1

This section describes methods used to obtain the data described in the supporting information section.

Movie S1

Typical behavior of trained control (Panx1 +/+ cage on the left) and knock-out (Panx1 −/− cage on the right) mice in cookie finding assay. The “X” presented in the lower left corner signifies where the cookie was stored during training sessions. Control mice immediately try to find the cookie in this corner, and when unsuccessful choose the top right hand corner instead. The Panx1 −/− explores the cage, however, lack indications of a systematic, memory-based exploration strategy.


Biography

Louis-Solal Giboin obtained his PhD in France under the supervision of Véronique Marchand-Pauvert, with whom he studied the cortical control of cervical propriospinal neurons in humans. He is now a post doctorant in charge of the Sensorimotor Performance Lab in the Human Performance Research Centre lead by Markus Gruber at the University of Konstanz in Germany. He is currently working on the neural adaptations induced by motor training and on the role of self-control in the regulation of strenuous physical activity.


Results

Previous results have shown that PKA anchoring is important for striatal LTP [26] and striatal dependent behavior [72], but these experiments did not delineate whether PKA needs to be anchored near its target molecules, such as DARPP-32 and the AMPA receptor GluA1 subunit, or near a source of activator molecules, such as adenylate cyclase. Though disruption of PKA anchoring prevents phosphorylation of targets such as sodium channels [23] or adenylate cyclase 5 [73] in response to a general elevation in cAMP, the role of anchoring in response to spatially constrained cAMP has not been investigated. In order to clarify which of these two associations is critical to PKA function, a model of the signaling pathways implicated in striatal synaptic plasticity ( Fig. 1A ) was implemented using NeuroRD [57], in a 7.75 µm long dendrite with four spines ( Fig. 1B ).

Validation of the Model

Prior to evaluating the role of PKA and adenylate cyclase location, the model was validated by comparison with experiments that measured phosphorylation of PKA targets in response to bath application of agonists. For this validation, both PKA and adenylate cyclase were located in the spine and dendrite in equal amounts.

The first validation simulates the response to bath application of 10 µM dopamine. Fig. 2A shows that a simulated increase in dopamine from 10 nM (basal level) to 10 µM leads to a 6-fold increase in phosphoThr34 DARPP-32 within 2 min, and a decrease of phosphoThr75 DARPP-32 to 60% of basal ( Fig. 2B ) comparable to the changes measured experimentally using dopamine D1R agonists [74]. In addition, simulated phosphoSer845 GluA1 increases by 8 fold within 5 to 10 minutes ( Fig. 2C ), comparable to measurements performed in the absence of D2 receptor stimulation [48].

(A) Change in phosphoThr34 DARPP-32 is similar to that observed experimentally in response to 10 µM dopamine alone, 600 nM calcium alone (inset), and the combination of dopamine with calcium. (B) Decrease in phosphoThr75 DARPP-32 for same three conditions as (A). (C) Change in phosphoSer845 GluA1 for same three conditions as (A). All responses are similar to experimental measurements.

The second validation simulates the response to a sustained increase in intracellular calcium concentration, as occurs due to experimental activation of NMDA receptors and voltage dependent calcium channels [75]. In response to this calcium increase, simulated phosphorylation of both DARPP-32 residues decreases, to half (Thr34) or two-thirds (Thr75) the basal level ( Fig. 2A and 2B ), in agreement with experimental data [56], [76]. Furthermore, the simulated calcium elevation leads to dephosphorylation of phosphoSer845 GluA1 ( Fig. 2C ) as previously reported [75].

The last model validation evaluates the change in phosphorylation of Thr34 and Thr75 in response to paired dopamine and calcium elevations. As demonstrated experimentally [75], [76], the addition of calcium reduces the increase in phosphoThr34 caused by dopamine ( Fig. 2A ), and enhances the decrease in phosphoThr75 ( Fig. 2B ). Furthermore, paired elevation of dopamine and calcium completely blocks the simulated phosphorylation of GluA1 Ser845 ( Fig. 2C ).

PKA Activity is Greater when Anchored near Adenylate Cyclase

Using this validated model, we explore the function of PKA anchoring relative to the location of both adenylate cyclase that activates it and PKA targets such as AMPA receptors ( Fig. 1C ). The D1 receptor is a metabotropic G protein coupled receptor functionally associated with adenylate cyclase [30], [31]. Immunohistochemical studies have shown D1 receptors located in spine heads and necks [77]. On the other hand, studies using immunogold labeling suggest D1 receptors are localized in extra synaptic sites homogeneously distributed in dendritic shafts and spines [78]. Therefore, prior to evaluating the effect of colocalization, we first examine the cAMP distribution when the adenylate cyclase/G protein/D1 receptor complex is located either in the spine head or in the submembrane region of the dendritic shaft.

Simulations show that when the adenylate cyclase-D1 receptor complex is located in the spine head, cAMP concentration is significantly higher in the spine head compared to the dendrite ( Fig. 3A ). This gradient is independent of PDE10 distribution (Suppl Fig S1). Conversely, when the complex is located in the submembrane compartment of the dendrite, cAMP concentration is higher in the dendrite than in the spine head ( Fig. 3B ). In summary, regardless of adenylate cyclase location, a cAMP gradient develops, with the highest concentration near the source (adenylate cyclase-D1 receptor complex) therefore, subsequent evaluations of the role of colocalization take into account the local cAMP concentration.

(A) When adenylate cyclase (AC) is in the spine head, there is a large difference (gradient) between spine cAMP and dendrite cAMP. (B) When adenylate cyclase is in the dendrite, there is a small gradient from dendrite to spine.

PKA location is regulated by interaction with A-Kinase anchoring proteins [19]. Various AKAPs anchor PKA to different locations, such as to the spine (e.g. AKAP5) [79], [80] or the dendritic shaft (e.g. MAP2) [81]. Recent evidence suggests that AKAP5 also anchors adenylate cyclase [22], [24], which would colocalize adenylate cyclase, PKA and GluA1 receptors. Alternatively, a pool of adenylate cyclase may be in the dendrite together with the dopamine D1 receptors found there [78]. In order to evaluate whether the critical function of anchoring is to place PKA near adenylate cyclase or near phosphoprotein targets, PKA is localized either in the spine heads or in a small submembrane region of the dendrite. For both of these spatial variations of PKA, the adenylate cyclase/D1 receptor complex is placed either in the spine heads or in the small submembrane region of the dendrite. Thus, four different spatial configurations are simulated ( Fig. 1C ). Note that the volume of the small dendrite region equals the volume of the spine head thus, both global and local concentrations of these molecules are equal for all simulations. This 2൲ experimental design ( Fig. 1C ) allows assessment of the role of PKA location relative to adenylate cyclase - the source of cAMP, or a non-diffusible target - the AMPA receptor GluA1 subunit. Note that membrane associated molecules have limited mobility in the membrane [62] and pools of PKA are anchored in multiple regions within the neuron nonetheless, our approach serves to enhance the effect of colocalization and facilitates evaluation of distinct functions of anchoring.

Simulations show that PKA colocalization with adenylate cyclase, either in the dendrite or the spine, produces a higher quantity of the active catalytic subunit than when PKA is apart from the adenylate cyclase ( Fig. 4A, B ). The large fluctuations in catalytic subunit are caused by its low quantity, which is due to a high affinity for the regulatory subunit and DARPP-32. Both the large fluctuations and the slow dynamics of PKA activation obscure the individual pulses that are visible in the cAMP traces. Colocalization in the spine produces greater PKA activity than colocalization in the dendrite because local spine cAMP is greater when adenylate cyclase is in the spine due to impeded cAMP diffusion. The effect of local cAMP (measured at the site of PKA anchoring) and colocalization (whether PKA was colocalized with adenylate cyclase or not) were evaluated using the SAS procedure GLM. Analysis revealed that local cAMP concentration predicted PKA activity (Pπ.0001, R 2  =𠂠.91), whereas colocalization did not reach significance (P =𠂠.083) thus the role of colocalization is primarily to place PKA in a microdomain of high cAMP concentration.

(A, B) Concentration of free catalytic subunit (PKAc) is greater for the two colocalized cases. (A) Traces are averaged over 4 trials nonetheless the stochastic fluctuations are so large that the traces overlay each other and are difficult to distinguish. (B) Mean and S.E.M. of the total PKA activity (PKAc summed between 50 and 350 s). (C,D) Concentration of phosphoThr34 DARPP-32 is greater for the two colocalized cases. (C) Traces are the average over four trials. (D) Mean and S.E.M. of the concentration of phosphoThr34 DARPP-32 averaged between 50 and 350 s. (E, F) Percent of phosphoSer845 GluA1 is greater for the two colocalized cases. (E) Traces are the average over four trials. (F) Mean and S.E.M. of the percent of phosphoSer845 GluA1 averaged between 50 and 300 s.

The effect of colocalization is propagated to downstream targets: Colocalization of PKA and adenylate cyclase (in spine or dendrite) leads to increased phosphoThr34 DARPP-32 ( Fig. 4C,D ) and phosphoSer845 GluA1 ( Fig. 4E,F ) compared to non-colocalized cases. In fact, both phosphoThr34 DARPP-32 and phosphoSer845 GluA1 are completely predicted by PKA activity (P =𠂠.0001, R 2  =𠂠.999 and P =𠂠.0001, R 2  =𠂠.93, respectively). The concentration of phosphoThr34 DARPP-32 (total quantity divided by volume of entire morphology) is illustrated because the gradients are much smaller than those of cAMP (Suppl Fig. S1). Evaluation of GluA1 in single spines on individual trials (Suppl Fig. S2) reveals that GluA1 phosphorylation varies considerably, exceeding 20% more often when PKA and adenylate cyclase are colocalized in the spine. This variability would not be evident in deterministic simulations.

Previous experiments have shown that disruption of PKA anchoring using Ht31 peptide blocks L-LTP induction in hippocampus [25] and LTP induction in the striatum [26]. To evaluate this experimental observation, we performed simulations with PKA uniformly distributed in the morphology, mimicking the disruption of anchoring by Ht31 peptide [23], [25]. The simulation shows that phosphoThr34 DARPP-32 and phosphoSer845 GluA1 are reduced (ratioρ) when PKA is uniformly distributed ( Fig. 5 ), especially for the case of adenylate cyclase located in the spine. This decreased PKA activity supports experimental studies showing that PKA anchoring is required for LTP induction.

When adenylate cyclase (AC) is in the spine, disruption of PKA anchoring reduces by 30% both phosphoThr34 DARPP-32 (p =𠂠.0006) and phosphoSer845 GluA1 (p =𠂠.0036). When adenylate cyclase is in the dendrite, disruption of PKA anchoring produces a significant decrease in phosphoThr34 DARPP-32 (p =𠂠.0005), but not phosphoSer845 GluA1 (p =𠂠.16). Most of the diffusely distributed PKA is in the dendrite thus, cAMP diffusion out of the spine (when adenylate cyclase is in the spine) to reach the PKA is more difficult than cAMP diffusion within the dendrite (when adenylate cyclase is in the dendrite). Consequently, disruption of PKA anchoring has a larger effect when adenylate cyclase is in the spine. PhosphoThr34 DARPP-32 is averaged between 50 and 350 s and phosphoSer845 GluA1 is averaged between 50 and 300 s.

Spine Neck Length Enhances the Effect of Colocalization

Recent studies have demonstrated that synaptic function and plasticity are correlated with synaptic structure and spine morphology [82], [83]. In particular, spine-neck geometry is an important determinant of NMDA receptor-dependent calcium signaling in the dendrite [84], [85]. To evaluate how spine morphology affects the interaction between PKA and cAMP, simulations are repeated using a spine with either a longer (1.0 µm) or shorter (0 µm) spine neck, representing the range of experimentally measured values in striatal dendrites [86]. Keeping the spine head the same size and shape maintains both the same quantity and concentration of molecules localized in the spine head. For these simulations, the four configurations of PKA and adenylate cyclase introduced previously are used.

Simulations show that spine neck enhances the effect of colocalization through control of cAMP concentration. An increase in spine neck length increases the cAMP gradient due to reduced diffusional coupling. Thus, an increase in neck length produces a larger cAMP in the spine when adenylate cyclase is in the spine, and a smaller cAMP in the spine when adenylate cyclase is in the dendrite. The effect of neck length on cAMP concentration gradient propagates downstream to PKA targets ( Fig. 6A,B ), yielding a larger effect of colocalization for the longer spine neck. Furthermore, local cAMP concentration predicts PKA activity regardless of neck length (Pπ.0001, R 2  =𠂠.93), and PKA activity predicts both phosphoThr34 DARPP-32 (Pπ.0001, R 2  =𠂠.99) and phosphoSer845 GluA1 (Pπ.0001, R 2  =𠂠.93). These results reinforce the observation that colocalization of PKA with its source of cAMP functions to locate PKA in a microdomain of high cAMP concentration, and also show that the effect of colocalization is robust to variation in spine neck length.

(A) The difference in phosphoThr DARPP-32 between colocalized and non-colocalized cases increases with a longer spine neck. (B) The difference in phosphoSer845 GluA1 between colocalized and non-colocalized cases increases with spine neck length. (C) Increase in diffusion constant decreases the activity of PKA when colocalized in the spine with adenylate cyclase (AC), but maintains enhancement over the non-colocalized case. (D) Changes in rate of GluA1 phosphorylation changes the level of phosphoSer845 GluA1, but maintains the difference between colocalized and non-colocalized cases. PhosphoThr34 DARPP-32 is averaged between 50 and 350 s and phosphoSer845 GluA1 is averaged between 50 and 300 s.

Several other simulations were performed to demonstrate that the results were robust to variations in parameters. The effect of colocalization was evaluated for different values of diffusion constants, rate of GluA1 dephosphorylation, and PDE10 location. For all these parameter variations, simulations demonstrate that colocalization of PKA with its activator still yields the greatest PKA activity. GluA1 dephosphorylation rate had no effect on PKA activity (results not shown), whereas an increase in diffusion constant ( Fig. 6C ) predictably decreased, but did not eliminate the effect of colocalization on PKA activity. Similarly, the GluA1 dephosphorylation rate changed the level of GluA1 Ser845 ( Fig. 6D ), but maintained the enhancement in phosphoSer845 GluA1 due to colocalization. Fig. S1 illustrates the effect of distributing PDE10 throughout the morphology (instead of locating it in the spine and submembrane region) for the two cases with adenylate cyclase in the spine. Though cAMP concentration and phosphoThr34 DARPP-32 are slightly increased when PKA is in the dendrite, PKA colocalization with adenylate cyclase still produces a much greater activity.

Effect of Calcium on Dopamine Response

Experimental evidence demonstrates that calcium influx through NMDA receptors is required for plasticity [87], [88] however the molecular mechanisms of calcium action are less clear. One possibility suggested by a previous model [27] is that transient calcium elevation (in contrast with prolonged calcium elevation) paired with dopamine enhances phosphoThr34 DARPP-32 compared with dopamine alone. Another possibility is that multiple kinases are required for synaptic plasticity, and, in particular, that calcium is required for CaMKII activation [45]. Therefore to explore the interaction between dopaminergic and glutamatergic transient signals in the medium spiny projection neuron, the model was stimulated with simultaneous dopamine and calcium pulses.

Transient calcium influx to spines was applied at 100 Hz for 1 sec, which approximates the calcium influx through NMDA receptors during LTP protocols [59]. This calcium influx generates a transient increase in calcium concentration that is constrained to the stimulated spines, as observed experimentally. Peak calcium levels in the stimulated spines (1500� nM) are much higher than both the non-stimulated spines (500� nM) and the dendrite (� nM), regardless of PKA and adenylate cyclase location.

Independent of whether PKA and adenylate cyclase were colocalized in the spine or dendrite, the PKA activity, phosphoThr34 DARPP-32, and phosphoSer845 were 10% smaller in response to calcium plus dopamine than in response to dopamine alone ( Fig. 7 ). The inhibition in PKA activity is caused by three pathways: inhibition of adenylate cyclase 5, activation of phosphodiesterase 1B and activation of calcineurin that dephosphorylates phosphoThr34 DARPP-32. The small increase in PKA activity in response to calcium alone is due to enhanced dephosphorylation of phosphoThr75 DARPP-32 by calcium bound PP2A ( Fig. 7B ) which cannot compensate for the other inhibitory actions of calcium, regardless of PP2A distribution or affinity for calcium (results not shown).

(A) Calcium alone produces a large increase in CaMKII phosphorylation and a small increase in PKA activity. Calcium together with dopamine leads to CaMKII phosphorylation only 10% lower than that observed with calcium alone, and PKA activity only 10% lower than that observed with dopamine alone. Both PKA and CaMKII activity are summed between 50 and 350 sec. (B) The calcium induced increase in PKA activity is caused by a small activation in PP2A, leading to a small decrease in phosphoThr75 (not shown).

To evaluate the possibility that calcium is needed for activation of other kinases, we measured the quantity of phosphorylated CaMKII, implicated in striatal LTP [89], in response to calcium, dopamine, and calcium plus dopamine. Fig. 7A shows that calcium stimulation alone causes a tremendous increase in phosphoCaMKII when compared with dopamine stimulation alone, while the combination of dopamine and calcium produces only a 10% reduction compared to calcium alone. This suggests that the increase in phosphoCaMKII more than compensates for the small decrease in PKA activity due to calcium. Though PKA and CaMKII have distinct targets in the model, extracellular signal-regulated kinase type II, important for dopamine signaling in the striatum [90], is phosphorylated by several kinases [91], and other points of PKA and CaMKII convergence are possible. In summary, these results suggest that both dopamine and calcium are required for LTP due to the need for activation of multiple kinases.

Colocalization of Dopamine Terminals and Receptors

There are two conflicting hypotheses on the relationship between the extracellular concentration of dopamine and the activation of dopamine receptors [92]: either dopamine concentration exhibits a spatial gradient and preferentially activates receptors near terminals, or dopamine concentration does not exhibit a spatial gradient and activates receptors homogenously through volume transmission. The latter hypothesis is supported by the microscopic anatomy of dopamine receptors and terminals [78], [93]. On the other hand, voltammetry experiments coupled with modeling demonstrate a spatial gradient of dopamine concentration, suggesting that nearby low affinity receptors (e.g. D1R) are activated preferentially [66], [94], [95]. Therefore the next set of simulations evaluates the extent to which an extracellular microdomain in dopamine would produce an intracellular microdomain of D1R activated second messengers.

To approximate the spatio-temporal profile of dopamine observed in voltammetry experiments [66], dopamine diffusion was decreased by a factor of 3. This approach avoided the need for dopamine uptake via the dopamine transporter and subsequent degradation, which was beyond the scope of the current paper that focuses on post-synaptic mechanisms. The quantity of dopamine released at the spine was adjusted so that dopamine concentration at the release site remained the same, which yielded no change in PKA activity when PKA, D1R and adenylate cyclase were colocalized with the dopamine terminals in the spine. The effect of the resulting dopamine spatial gradient was evaluated by measuring the change in PKA activity when PKA, D1R and adenylate cyclase were colocalized in the dendrite, but 3 µm from the dopamine terminals.

Simulations show that the dopamine spatial gradient ( Fig. 8B1 inset) leads to a small but significant decrease in amplitude of cAMP concentration (p =𠂠.0012) and phosphoThr34 DARPP-32 (p =𠂠.046), compared to no dopamine gradient ( Fig. 8 ). In contrast, the time of the peak cAMP concentration ( Fig. 8A1 ), phosphoThr34 DARPP-32 ( Fig. 8B1 ) and phosphoSer845 GluA1 ( Fig. 8C1 ) were not altered. This small change in intracellular signaling molecules caused by the dopamine spatial gradient suggests that small distances between receptor and terminal may not be important, especially if the intracellular molecules do not exhibit a spatial gradient. This idea was further evaluated with the next simulation, in which spatial gradients of intracellular signaling molecules were evaluated in a longer dendrite.

A spatial gradient of dopamine decreases (A) cAMP and PKA activity as measured by (B) phosphoThr34 DARPP-32 and (C) phosphoSer845 GluA1, but no delay in time course. A1, B1 and C1 show the time course of a single simulation, A2 show mean and stdev (n =𠂤) of cAMP averaged between 50 and 200 s B2 and C2 show mean and stdev (n =𠂤) of phosphoThr34 DARPP-32 averaged between 50 and 350 s, and phosphoSer845 GluA1 averaged between 50 and 300 s, respectively. The inset of B1 shows the gradient in dopamine concentration from the PSD to the dendrite. Traces are the average of four simulations.

Microdomains and Spatial Specificity

The effect of close proximity between dopamine terminals and receptors can only be important when accompanied by mechanisms for limiting the spread of intracellular signaling molecules. Spatial specificity of intracellular signaling is critical for information processing, in particular for a neuron to discriminate between different patterns of input. To evaluate if spatial specificity of dopamine signaling would propagate to downstream targets, we performed simulations using a longer dendrite with an average spine density of 1/µm, as experimentally measured [86]. These simulations use the slow dopamine diffusion which produces a gradient in dopamine to ensure that only the dopamine receptors at one end are activated.

Fig. 9A shows that dopamine stimulation of one end of the dendrite produces a spatial gradient in cAMP concentration, reaching � nM at the stimulated end, and decaying to basal concentration within 13 µm of the stimulation site. The spatial profile of cAMP is well described by a single spatial decay constant of 4.69ଐ.14 µm ( Fig. 9B ). This gradient does not propagate to PKA activity, as measured by phosphoThr34 DARPP-32 activity ( Fig. 9C ), or phosphoSer845 GluA1 ( Fig. 9E ). The lack of intracellular gradient of PKA activity supports the ineffectiveness of extracellular gradients in dopamine. This result further suggests that PKA anchoring is not sufficient by itself to produce PKA microdomains, in part because when PKA is activated, the catalytic subunit diffuses.

(A) cAMP concentration versus time and distance from dopamine release site. (B) cAMP concentration, averaged from 50 to 150 sec, is well fit by single exponential decay. (C) phosphoThr34 DARPP-32 concentration versus time and distance from dopamine release site exhibits minimal spatial gradient. (D) Concentration of phosphoThr34 DARPP-32, averaged between 100 and 250 sec, exhibits a spatial gradient when diffusion of all DARPP-32 forms is zero (red), or diffusion of PKA bound DARPP-32 is zero (blue). Blocking the phosphoThr75-PKA interaction does not change the gradient that appears when diffusion of all DARPP-32 forms is zero (black). All three cases overlap and have the same decay space constant thus, they are difficult to distinguish in the figure. The inset shows the fits alone, which also overlap. (E) Percent of GluA1 phosphorylated on Ser845, averaged between 100 and 250 sec, versus distance from dopamine release site. (F) Percent of GluA1 phosphorylated on Ser845, averaged between 100 and 250 sec, exhibits a spatial gradient when diffusion of the DARPP-32 forms is zero (red), or diffusion of PKA bound DARPP-32 is zero (blue). Blocking the phosphoThr75-PKA interaction does not change the gradient that appears when diffusion of all DARPP-32 forms is zero (black).

The decay length of a molecule's concentration gradient is governed not only by its diffusion constant but also by its inactivation rate [17], [96]. Though PKA is inactivated by the large quantity (� µM) of phosphoThr75 DARPP-32, this inactivation is reversible, similar to calcium binding to calcium buffers [97]. In addition, diffusion of phosphoThr34 DARPP-32 could obscure the gradient of PKA activity. Therefore, to evaluate the role of DARPP-32, simulations were repeated with diffusion constants for various forms of DARPP-32 set to zero.

Fig. 9D and 9F show that a gradient in phosphoThr34 DARPP-32 and phosphoSer845 GluA1 is observed in the absence of diffusion of any form of DARPP-32 (red traces). The gradient has a spatial decay constant of 14.7ଘ.5 µm for phosphoThr34 DARPP-32, which does not differ significantly from the spatial decay constant of 16.1넒.2 µm for phosphoSer845 GluA1 (P =𠂠.84). To demonstrate that DARPP-32 acts similar to calcium buffers in spreading PKA activity, simulations were repeated with diffusion for only the PKA-bound forms set to zero (both the PKA-DARPP-32 complex and the inhibited PKA-phosphoThr75 DARPP-32 complex), but diffusion of phosphoThr34 DARPP-32 enabled. Figs. 9D and 9E show that a gradient in phosphoThr34 DARPP-32 and phosphoSer845 GluA1 are observed in these conditions (blue traces). More importantly, the spatial decay constants of the gradient do not differ between these two cases (P =𠂠.8 for phosphoThr34 DARPP-32 and P =𠂠.4 for phosphoSer845 GluA1). As a further demonstration that PKA binding to phosphoThr75 DARPP-32 does not act as an inactivation mechanism, simulations were repeated with this reaction blocked, in the absence of DARPP-32 diffusion. Fig. 9D and 9F (black traces) show that blocking this reaction does not decrease the gradient, and thus the reaction does not act as an inactivation mechanism at this spatial scale.


Methods

Neuronal culture preparation

Hippocampal neurons dissociated from Wistar rats (P0–P2) were plated on polyorhitine/matrigel pre-coated MEA at a concentration of 8 × 10 5 cells/cm 2 and maintained in a neuron medium as previously described [9]. After 48 h, 5 μM cytosine-β- D-arabinofuranoside (Ara-C) was added to the culture medium in order to block glial cell proliferation. Neuronal cultures were kept in an incubator providing a controlled level of CO2 (5%), temperature (37°C) and moisture (95%).

Electrical recordings and electrode stimulation

Multi electrode array (MEA) recordings were carried out with a MEA60 system (Multi Channel Systems, Reutlingen, Germany). Stimulations and recordings were carried out after 21–35 days in vitro. Synchronous network bursting was induced by 30 min. treatment with GABAA receptor antagonists, gabazine (20 μM). In some experiments, blockers of the ERK1/2 pathway (50 μM PD98059 and 20 μM U0126) were used. These drugs were pre-incubated before application of gabazine for 45 min. The voltage pulse was bipolar with a duration of 200 μsec. For a given culture, the same amplitude was used as before, during and after GabT. The pattern of stimulation consisted of a train of 40 bipolar pulses separated by an inter-pulse interval of 4s and applied to a bar of six neighbouring electrodes.

Data analysis

Acquired data were analyzed using MATLAB (The Mathworks, Inc.) as previously described [10, 19]. The network firing rate (NFR) is defined as the sum of all electrodes firing rates (i.e. the number of all spikes recorded in the network for each bin). The total number of spikes as a function of the distance d was fitted by the exponential function from which λ was obtained (inset Fig. 1G).

Quantitative RT-PCR

RNA (250 ng) was reverse transcribed using SuperScript II reverse transcriptase and random hexamer (Invitrogen, Milan, Italy). Real-time PCR was performed using iQ SYBR Green supermix (Biorad, Milan, Italy) and the iQ5 LightCycler. Gene specific primers were designed using Beacon Designer (Premier Biosoft, Palo Alto, CA, USA). The thermal cycling conditions comprised 3 min at 95°C, and 40 cycles of 10 seconds for denaturation at 95°C and 45 sec for annealing and extension at 58°C. The expression level of the target mRNA was normalized to the relative ratio of the expression of Gapdh mRNA. The forward primer for Gapdh was 5'-CAAGTTCAACGGCACAGTCAAGG-3', the reverse primer was 5'-ACATACTCAGCACCAGCATCACC-3'. Fold changes calculations were made between treated and untreated samples at each time point using the 2 -ΔΔCT method. The forward primer for Egr1 was 5'-AAGGGGAGCCGAGCGAAC-3', the reverse primer was 5'-GAAGAGGTTGGAGGGTTGGTC-3' forward primer for Egr2 was 5'-CTGCCTGACAGCCTCTACCC-3', reverse primer was 5'-ATGCCATCTCCAGCCACTCC-3' forward primer for Egr3 was 5'-ACTCGGTAGCCCATTACACTCAG-3', reverse primer was 5'-GTAGGTCACGGTCTTGTTGCC-3' forward primer for Nr4a1 was 5'-GGTAGTGTGCGAGAAGGATTGC-3', reverse primer was 5'-GGCTGGTTGCTGGTGTTCC-3' forward primer for Arc was 5'-AGACTTCGGCTCCATGACTCAG-3', reverse primer was 5'-GGGACGGTGCTGGTGCTG-3' forward primer for Homer1a was 5'-GTGTCCACAGAAGCCAGAGAGGG-3', reverse primer was 5'-CTTGTAGAGGACCCAGCTTCAGT-3' forward primer for Bdnf was 5'-CGATTAGGTGGCTTCATAGGAGAC-3', reverse primer was 5'-GAAACAGAACGAACAGAAACAGAGG-3'.