This page is archived

Links to external sources may no longer work as intended. The content may not represent the latest thinking in this area or the Society’s current position on the topic.

Integrating Hebbian and homeostatic plasticity

19 - 20 April 2016 09:00 - 17:00

Scientific discussion meeting organised by Professor Kevin Fox FMedSci and Professor Michael Stryker.

This meeting brought together computational and experimental neuroscientists to discuss how interactions between Hebbian plasticity and homeostatic plasticity occur and how they can be detected and interrogated at the systems level.

Organiser and speaker biographies are available below, together with the schedule of talks and abstracts. Alternatively you can download the draft programme (PDF). 

Position papers

Position papers are available to download below. Three of the speakers contributed their thoughts on the state of the field and the major issues to be addressed at the meeting. The position papers are provided by the speakers and the Royal Society takes no responsibility for their content.

Attending this event

This meeting has taken place. Recorded audio of the presentations can be found below, and papers from the meeting will be published in a future issue of Philosophical Transactions B.

Enquiries: Contact the events team.

Organisers

  • Professor Kevin Fox FMedSci, Cardiff University, UK

    Kevin Fox PhD MAE FMedSci is a Professor of Neuroscience and a member of the School of Biosciences and the Neuroscience and Mental Health Research Institute at Cardiff University. He studied Electrical Engineering at the University of Bath before receiving his PhD in Neurophysiology from the University of London. His research programme is funded by the Medical Research Council in the UK. His lab is interested in the systems, cellular and molecular mechanisms of cortical plasticity. Work in the lab has focused on the visual and somatosensory cortex as model systems for understanding plasticity. Current research addresses structural plasticity mechanisms in layer 2/3 neurons and functional plasticity in layer 5 neurons, particularly the differences inherent in Regular spiking and Intrinsic bursting cells of cortical layer 5. The overall aim of his lab is to understand plasticity to the level where it can be safely manipulated for therapeutic benefit.

  • Professor Michael Stryker, University of California, San Francisco, USA

    Michael Stryker studied at Deep Springs College and the University of Michigan and did research in the laboratory of James Olds. After two years in East Africa as a hydraulics engineer with the Peace Corps, he entered Peter Schiller's laboratory at MIT for PhD studies, followed by a postdoctoral fellowship with David Hubel and Torsten Wiesel at the Harvard Medical School. He joined the Physiology Department and the nascent neuroscience programme at UCSF as an assistant professor in 1978, where he served as department chair for 12 years and has remained except for sabbaticals at Oxford and as the Galileo Professor of Science at Esculoa Normale Superiore in Pisa. He has also served as Co-director of the UCSF Neuroscience Program in and as Director of the Boyer Program in Biological Science overseeing all the UCSF basic science graduate programs.  He holds the W.F. Ganong Chair of Physiology at UCSF and has been honoured by election to the American Academy of Arts and Sciences and the US National Academy of Sciences and by the W. Alden Spencer Award and the Pepose Award in Vision Science.

Schedule

Chair

Professor John Lisman, Brandeis University, USA

09:05 - 09:30 Spine-size fluctuations enable stable cell assembly learning in recurrent circuit models

Cortical circuits rewire in an experience-dependent way. A major biological mechanism underlying this is Hebbian plasticity. In models of recurrently connected networks, ongoing Hebbian plasticity is often unstable in nature because of its positive feedback, eg a tightly coupled and coherently active group of neurons tends to drive other neurons well and expand the group (Kunkel et al., 2011). This typically fuses multiple memory patterns and results in a deficiency in learning/memory performance. The biological mechanism that stabilises Hebbian plasticity is unknown. Here we combine experimentally observed fluctuations of spine sizes (Yasumatsu et al., 2008) with spike-timing-dependent plasticity in recurrently connected neural networks. We show that an appropriate level of spine fluctuations is sufficient to stabilise memory patterns without fusing, and maintains a physiological volume distribution of spines in the presence of ongoing Hebbian plasticity. In addition to stabilising Hebbian plasticity, we posit that abnormal spine fluctuations impair learning/memory performance. Our theory explains how high spine turnover rates, experimentally observed in several animal models for autism (Isshiki et al., 2014), cause slow learning and impairs memory performance.

Dr Taro Toyoizumi, RIKEN Brain Science Institute, Japan

09:45 - 10:15 Homeostasis and assembly formation in spiking networks

Homeostatic mechanisms homogenise activity within spiking networks, ensuring that any potential for unstable network activity is corrected. However, strict homogeneity in a recurrent cortical circuit precludes any rich dynamics, as well as the computations they support. Thus, there is a tension between the need for stability and the desire for rich circuit structures that often flirt with instability. We explore this problem by studying how homeostasis through plasticity of inhibitory connections interacts with plasticity of excitatory connections during assembly formation in models of cortex. We show that homeostatic inhibition is essential for stabilising assembly formation both with rate-based and timing-based plasticity rules. The key requirement is that homeostatic mechanisms should not operate on the timescale of learning – for rate based excitatory plasticity the homeostatic mechanisms should be slower, for spike-timing based mechanisms homeostasis should be faster.

Dr Brent Doiron, University of Pittsburgh, USA

11:00 - 11:30 Homeostasic control in recurrent networks

We present a systematic analysis of homeostatic control in networks of neurons. It reveals two important aspects of homeostatic control.

First, we consider networks of neurons with homeostasis and show that homeostatic control that is stable for single neurons, can destabilise activity in otherwise stable recurrent networks. This instability can be prevented by dramatically slowing down the homeostatic control. Next, we consider the case that homeostatic feedback is mediated via a cascade of multiple intermediate stages. Counter-intuitively, the addition of extra stages in the homeostatic control loop further destabilises activity in single neurons and networks. We thus reveal previously unconsidered constraints on homeostasis in biological networks, and provide a possible explanation for the slow time-constants of homeostatic regulation observed experimentally.

Dr Mark van Rossum, University of Edinburgh, UK

11:45 - 12:15 Homeostatic control during Hebbian changes: a question of time scales

When looking at the interaction of homeostatic and Hebbian plasticity we make a puzzling observation: while homeostasis of synapses reported in experiments is slow, homeostasis of synapses in most computational models is rapid, or even instantaneous. Even worse, most existing plasticity models cannot maintain stability in simulated networks with the slow homeostatic plasticity reported in experiments. To solve this paradox, we suggest that there are both fast and slow forms of homeostatic plasticity with distinct functional roles. Theory predicts that there must be a fast form of synaptic homeostasis (on the time scale of seconds or minutes), in order to render Hebbian plasticity intrinsically stable. This fast form might be experimentally observable as heterogeneous synaptic plasticity. Furthermore, slower forms of homeostatic plasticity are important for fine-tuning neural circuits. Taken together we suggest that learning and memory relies on an intricate interplay of diverse plasticity mechanisms on different timescales which jointly ensure stability and plasticity of neural circuits. 

1. F. Zenke, E.J. Agnes and W. Gerstner. 2015. Diverse synaptic plasticity mechanisms orchestrated to form and retrieve memories in spiking neural networks. Nature Communications 6, 6922.

2. F. Zenke, G. Hennequin and W. Gerstner. 2013. Synaptic Plasticity in Neural Networks Needs Homeostasis with a Fast Rate Detector. PLOS Computational Biology 9:e1003330 DOI:10.1371/journal.pcbi.1003330

KONICA MINOLTA DIGITAL CAMERA

Professor Wulfram Gerstner, École polytechnique fédérale de Lausanne, Switzerland

Chair

Professor Gina Turrigiano, Brandeis University, USA

13:30 - 14:00 Cocaine-induced synaptic plasticity in the striatum: Hebbian and homeostatic mechanisms

Drugs of abuse, such as cocaine, induce changes in reward circuitry, which manifest as long lasting changes in behavior. The synaptic changes are likely due to a combination of both Hebbian and homeostatic types of plasticity. As the pro-inflammatory cytokine tumor necrosis factor alpha (TNF) is required for some forms of homeostatic plasticity in the cortex and hippocampus but is not required for Hebbian forms of plasticity, we tested the TNF-dependence of cocaine-induced striatal plasticity. Repeated administration of cocaine results in an initial weakening of glutamatergic synapses in the nucleus accumbens, followed by a synaptic strengthening during abstinence from drug. We show that this biphasic plasticity is comprised of two parts. Cocaine substantially elevates dopamine, which directly potentiates synapses on D1-expressing accumbens neurons. However, cocaine also activates microglia in the nucleus accumbens and increases TNF production. TNF acts to depress synapses preferentially on D1-expressing neurons in the accumbens, antagonizing cocaine-induced synaptic plasticity and reducing behavioural sensitization. During abstinence, microglia de-activate and TNF levels drop, revealing the underlying potentiation. Importantly, a weak TLR4 agonist can re-activate microglia, increase TNF production, depress synaptic strength in the accumbens, and suppress cocaine-induced sensitization. Thus, microglia act adaptively to maintain circuit function in the face of cocaine-induced disruption, and suggests this is a homeostatic response to aberrant Hebbian-type plasticity induced by cocaine.

Dr David Stellwagen, McGill University, Canada

14:15 - 14:45 Cholinergic modulation of NMDA receptor function mediated through an astrocyte intermediate

A transformation in thinking about plasticity occurred a quarter of a century ago when it was demonstrated that astrocytes release chemical transmitters that can modulate synaptic transmission and plasticity. In this presentation I will summarise this work as well as highlight recent studies which have shown the importance of astrocyte derived D-serine in the control of NMDA receptor function and, as a consequence, learning and memory. While synaptic plasticity is an essential feature of pre and postsynaptic neurons, through gliotransmission at the tripartite synapse astrocytes tune the synaptic system for effective scaling of the magnitude of the plastic event.

Professor Philip Haydon, Tufts University School of Medicine, USA

15:30 - 16:00 Synaptic signalling of retinoic acid

A neuron’s ability to change its responsiveness to synaptic inputs based on prior experience is an essential feature of the nervous system. The known forms of long-term synaptic plasticity can be grossly divided into two main categories: Hebbian and homeostatic. Our work established a critical role of retinoic acid (RA) in a form of homeostatic synaptic plasticity that is induced by prolonged reduction in synaptic excitation. Acting through a distinct molecular mechanism, RA is capable of rapidly changing excitatory as well as inhibitory synaptic strength. One of the main open questions is whether and how homeostatic synaptic plasticity intersects with Hebbian synaptic plasticity, and how such interaction impacts animal learning. In this talk, I will describe a recent study in which we examined the impact of RA-dependent synaptic signalling on Hebbian plasticity and hippocampal-dependent learning. Exposure of adult animals to an enriched environment (EE) engaged hippocampal RA signalling, which altered excitatory synaptic strength in the hippocampal CA1 neurons. Unexpectedly, although EE exposure did not alter LTP magnitude in WT animals, it significantly enhanced LTP in RARalpha KO hippocampus. By contrast, EE exposure significantly enhanced LTD in WT hippocampus, and such enhancement was prevented by RARalpha deletion. We also examined the behavioural consequences of altered Hebbian plasticity and found that hippocampal-dependent learning was altered in EE-exposed RARalpha knockout animals. Taken together, our study for the first time established an in vivo function of synaptic RA signalling in adult animals and demonstrated that RA-dependent experience-induced synaptic modification acts as a form of meta-plasticity to impact Hebbian plasticity and learning.

Dr Lu Chen, Stanford University, USA

16:15 - 16:45 Homeostatic plasticity of inhibition

Inhibitory synaptic transmission is critical for normal cortical functions. In sensory cortices, changes in the functional connectivity between inhibitory interneurons and pyramidal cells dictate the maturation of cortical function in tune with the environment. In order for inhibitory network to provide appropriate level of control over neural activity, it is critical that they undergo homeostatic regulation depending on the demand of the cortical circuitry. We found that homeostatic regulation of inhibitory synapses occurs via two distinct mechanisms during development. In immature circuits, inhibitory synapses adapt to overall changes in sensory experience by postsynaptic mechanisms. In contrast, in mature circuits we uncovered a novel mechanism that allows selectively control of action-potential independent release without changes in evoked inhibition. Such adaptation endows the mature cortical circuit to provide homeostatic control of overall neural activity without compromising information coding capabilities.

Dr Hey-Kyoung Lee, Johns Hopkins University, USA

Chair

Dr Hey-Kyoung Lee, Johns Hopkins University, USA

09:05 - 09:30 Firing rate homeostasis is gated by sleep/wake states

Homeostatic mechanisms stabilise neural circuit function by keeping firing rates (FRs) within a set-point range, but whether individual neocortical neurons regulate firing around a cell-autonomous set-point, and whether this process is restricted to certain behavioural states such as sleep or wake, is unknown. I will start by discussing the mechanisms of synaptic scaling, a form of cell-autonomous homeostatic plasticity, and the role of this plasticity in generating firing rate set-points in vivo. I will then discuss new work in which we follow the process of FR homeostasis in individual visual cortical neurons in freely behaving rodents as they cycled between sleep and wake states. When FRs are perturbed by visual deprivation, over time they returned precisely to a cell-autonomous set-point, and this restoration of firing occurred selectively during periods of active waking and was suppressed by sleep. Longer natural waking periods result in more FR homeostasis, as does artificially extending the length of waking. This exclusion of FR homeostasis from sleep raises the possibility that memory consolidation or some other sleep-dependent process is vulnerable to interference from homeostatic plasticity mechanisms.

Professor Gina Turrigiano, Brandeis University, USA

09:45 - 10:10 Hebbian and homeostatic plasticity and the role of TNF-α in the visual cortex

Professor Michael Stryker, University of California, San Francisco, USA

10:50 - 11:15 Diversity of homeostatic and Hebbian plasticity properties in cortical neurones

Not all neurones are created equal when it comes to plasticity. Either by virtue of their different afferent pathways or their different synaptic receptors, neurones in different layers of the cerebral cortex show different degrees and types of synaptic plasticity. For principal neurons in the adult somatosensory cortex, layer 4 is relatively aplastic while layer 2/3 and layer 5 are highly plastic. Recent studies show that the diversity does not end there. The two major subdivisions of layer 5 pyramidal cells, regular spiking (RS) and intrinsic bursting (IB) cells also show different types of plasticity and rely on Hebbian and homeostatic mechanisms to different degrees. RS neurones show experience-dependent depression following whisker trimming, which slowly recovers homeostatically back to baseline despite the maintained deprivation. The homeostatic rebound is TNFa dependent. Potentiation of spared whisker responses is absent in these cells. In contrast, IB cells do show potentiation of spared whisker responses comprising both TNF a and a-CaMKII-autophosphorylation dependent components1. As LTP is absent in aCaMKII-autophosphorylation mutants2 and synaptic upscaling is absent in TNF a knockouts3, these findings suggest the two synaptic mechanisms are distributed differently between the two cell types. In the light of these results, we looked again at plasticity in layer 2/3. Previous studies had shown that LTP and experience-dependent potentiation are absent in the barrel cortex of aCaMKII-autophosphorylation mutants2. Our present studies show that in addition, following depression of deprived whisker responses, layer 2/3 cells show a homeostatic rebound that is prevented by a soluble TNF a scavenger. In summary, all three cortical cell types show varying degrees of TNF a dependent homeostatic plasticity, but, while layer 2/3 cells show both Hebbian depression and potentiation, in layer 5 Hebbian depression and potentiation are segregated between RS and IB cells respectively.

1. S. D. Greenhill et al. 2015. Neuron 88, 539-552.

2. N. Hardingham et al. 2003.J Neurosci 23, 4428-4436.

3. D. Stellwagen, R. C Malenka. 2006. Nature 440, 1054-1059.

Professor Kevin Fox FMedSci, Cardiff University, UK

11:30 - 11:55 Rapid homeostasis by control of inhibition

Cortical plasticity involves both Hebbian mechanisms that alter neural tuning in response to experience, and multiple homeostatic mechanisms that maintain cortical firing rates within a stable operating regime. We have investigated the circuit and cellular mechanisms for homeostasis in response to brief sensory manipulations in whisker somatosensory cortex (S1). Whisker deprivation induces a rapid homeostatic increase in local circuit excitability in L2/3 of S1 that preserves sensory-evoked spike rates despite loss of excitatory synaptic input. This occurs within 1 day of deprivation, and precedes classical changes in receptive fields and maps. These changes do not involve synaptic scaling, but instead reflect a preferential reduction in whisker-evoked inhibition in L2/3 pyramidal cells, due to reduced activation of parvalbumin (PV) interneuron circuits. Deprivation-induced disinhibition occurs in both feedforward and recurrent inhibitory networks.  Sustained deprivation drives disinhibition by reducing excitatory synaptic input to L2/3 PV neurons. In contrast, brief deprivation drives rapid disinhibition by reducing intrinsic excitability of PV neurons. This is evident in elevated spike threshold, increased spike width and damped near-threshold excitability. 

These findings add to a growing body of evidence that PV interneurons are a critical nexus for homeostatic plasticity in sensory cortex. This single site of plasticity in the cortical microcircuit can control average firing rate in local cortical networks, regulate sensory gain, and gate subsequent Hebbian plasticity for reorganization of the whisker map.

Dr Daniel Feldman, University of California, Berkeley, USA

12:10 - 12:35 Homeostatic mechanisms in the mouse visual cortex in vivo

Homeostatic synaptic scaling is thought to occur cell-wide. We used repeated in vivo two-photon imaging in mouse visual cortex after sensory deprivation to investigate the spatial extent of synaptic scaling. We used increases in spine size as a proxy for synaptic scaling in vivo in both excitatory and inhibitory neurons and found that after sensory deprivation, increases in spine size are restricted to a subset of dendritic branches, which we confirmed using immunohistochemistry. We found that the branches that had increases in spine size also had a lower spine density. Within a given dendritic branch, the degree of spine size increases was proportional to recent spine loss within that branch. Using simulations, we showed that this compartmentalised form of synaptic scaling better retained the previously established input-output relationship in the cell.

Dr Tara Keck, University College London, UK

Chair

Dr Daniel Feldman, University of California, Berkeley, USA

Professor Mark Hübener, Max Plank Institute of Neurobiology, Germany

13:50 - 14:15 LTP, STP, and scaling: electrophysiological, biochemical, and structural mechanisms

Synapses are complex because they perform multiple functions, including at least six mechanistically different forms of plasticity (STP, early LTP, late LTP, LTD, distance-dependent scaling, and homeostatic scaling). The ultimate goal of neuroscience is to provide an electrophysiologically, biochemically, and structurally specific explanation of the underlying mechanisms. This review summarises the still limited progress towards this goal. Several areas of particular progress will be highlighted: 1) STP, a Hebbian process that requires small amounts of synaptic input, appears to make strong contributions to some forms of working memory; 2) The rules for LTP induction in the stratum radiatum of the hippocampus have been clarified: induction does not depend obligatorily on backpropagating Na spikes but, rather, on dendritic branch-specific NMDA spikes. Thus, computational models based on STDP need to be modified; 3) Late LTP, a process that requires a dopamine signal (neoHebbian), is mediated by trans-synaptic growth of the synapse, a growth that occurs about an hour after LTP induction; 4) There is no firm evidence for cell-autonomous homeostatic synaptic scaling; rather, homeostasis is likely to depend on a) cell-autonomous processes that are not scaling, b) synaptic scaling that is not cell autonomous but instead depends on population activity, or c) metaplasticity processes that change the propensity of LTP vs LTD; 5) The evidence for distance-dependent scaling along the primary dendrite is now firm, and a plausible structural-based mechanism is suggested; 6) Recent super-resolution studies indicate that glutamatergic synapses are modular (module size 70-80nm), as predicted by theoretical work. Modules are trans-synaptic structures and have high concentrations of PSD-95 and AMPAR. These modules function as quasi-independent loci of AMPA-mediated transmission and can probably be independently modified during plasticity (eg, some modules may be silent but become functional after LTP induction). These new discoveries open the door for understanding the structure/function relationships that underlie the multiple forms of synaptic plasticity at individual synapses.

Professor John Lisman, Brandeis University, USA

14:30 - 14:55 Effects of dark exposure on mouse visual cortex plasticity

Changes in the excitatory-inhibitory (E/I) balance, mediated by maturation of parvalbumin positive (PV) GABAergic interneurons, have been identified as a key factor controlling the critical period of experience-dependent plasticity. Interventions that slow PV cell maturation, such as dark rearing, also delay the time course of the critical period. In both rats and cats brief dark exposure (DE) later in life can restore plasticity and enable recovery of vision through a previously deprived eye. 

We studied the effects of DE on ocular dominance and single-cell responses in the binocular zone of mouse V1, during recovery from long-term monocular deprivation. We employed chronic intrinsic signal imaging and two-photon calcium imaging. We also examined the density of PV neurons and of perineuronal nets (PNNs), a known structural brake on plasticity. After reopening of the deprived eye one group was placed in a dark room for 7 days and subsequently transferred to standard cages in a 12h day/night cycle while the other group was placed immediately in standard cages. Animals were imaged 5 times at weekly intervals. In terms of ocular dominance, mice that had experienced DE exhibited more rapid recovery (within 1 week) than those that had, not but the end points were not significantly different. Single cells calcium signals revealed greater recovery of orientation selectivity after DE. The proportion of PV cells surrounded by PNNs was smaller in mice that had experienced DE.

Our results show that DE boosts experience-dependent plasticity by restoring the visual cortex to a more juvenile-like state.

Professor Frank Sengpiel, Cardiff University, UK

15:30 - 15:55 Cell-specific restoration of stimulus preference after monocular deprivation in visual vortex

Monocular deprivation (MD) evokes a prominent shift of neuronal responses in the visual cortex towards the open eye, accompanied by functional and structural synaptic rearrangements. This shift is reversible, but it is unknown whether the recovery happens at the level of individual neurons or whether it reflects a population effect. We used ratiometric Ca2+ imaging to follow the activity of the same excitatory layer 2/3 neurons in mouse visual cortex over months during repeated episodes of ocular dominance (OD) plasticity. We observed robust shifts towards the open eye in most neurons. Nevertheless, these cells faithfully returned to their pre-deprivation OD during binocular recovery. Moreover, the initial network correlation structure was largely recovered, suggesting that functional connectivity may be regained despite prominent experience-dependent plasticity.

Dr Tobias Rose, Max Plank Institute for Neurobiology, Germany

16:10 - 17:00 General summary and discussion

Dr Tara Keck, University College London, UK

Dr Taro Toyoizumi, RIKEN Brain Science Institute, Japan